Mechanical water meter data acquisition method and device and electronic equipment

Abstract

A mechanical water meter data acquisition method, device and electronic equipment, the mechanical water meter data acquisition method comprising: bidirectional tracing binding of pixels of a first water meter image and three-layer physical space positions of a water meter to output a second water meter image; performing directional light field component separation operation on superimposed optical signals in the second water meter image based on a three-layer light field propagation model to output three-layer physical structure effective metering light field signals after stripping interference light field components; performing cross-space layer double-constraint feature fusion processing on the second water meter image based on the three-layer physical structure effective metering light field signals to output a third water meter image containing a feature-fused metering light field signal; outputting light field data with completed semantic reconstruction according to the third water meter image and associated light field distribution and topology information of the three-layer physical structure; and acquiring a water meter reading in the third water meter image in combination with a residual light field reverse closed-loop verification mechanism, the light field data and the third water meter image. Through the above setting, the accuracy of acquiring mechanical water meter data counting can be improved.

Description

Technical Field

[0001] This application relates to the field of water meter technology, and in particular to a method, apparatus and electronic device for acquiring data from a mechanical water meter. Background Technology

[0002] Existing mechanical water meters are generally installed in harsh conditions such as dim lighting, dampness, water accumulation, exposure to direct sunlight, and confined spaces. The meter crystal is prone to problems such as glare, internal fogging, and perspective distortion. Long-term accumulation of scale and grease on the dial directly causes blurred images, incomplete digits, broken pointer lines, and significant loss of metering features, making accurate water meter readings difficult. Current manual meter reading relies on visual inspection, which is prone to subjective errors and inefficient. General image recognition only identifies the digits and pointer values ​​independently based on pixel features, failing to consider the mechanical structure of the water meter's gear transmission and step-by-step carry-over mechanism. This cannot correct for recognition errors caused by residual environmental interference, frequently resulting in incorrect readings and decimal point distortion. General image processing only uses global uniform noise reduction, which cannot peel away multiple layers of interference from the meter crystal or repair metering features obscured by dirt, resulting in extremely poor image quality and making it impossible to accurately obtain mechanical water meter data.

[0003] Therefore, there is an urgent need for a method to obtain readings from mechanical water meters to solve the existing problems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this application is to provide a method, apparatus, and electronic device for acquiring data from mechanical water meters, which can improve the accuracy of acquiring mechanical water meter data.

[0005] To achieve the above objectives, this application adopts the following technical solution: A method for acquiring data from a mechanical water meter, characterized in that it includes: The original water meter reading image is obtained as the first water meter image. Based on the three-layer physical structure of the water meter's outer glass, internal sealed water vapor cavity, and dial base, a three-layer light field propagation model adapted to the optical characteristics of the water meter cavity is constructed. Based on the first water meter image and the three-layer light field propagation model, the pixels of the first water meter image are bidirectionally traced and bound to the three-layer physical spatial position of the water meter to output the second water meter image. Any pixel of the second water meter image is accurately mapped to the three-layer physical spatial position of the water meter and its position is bound. Based on the three-layer optical field propagation model, directional optical field component separation calculation is performed on the optical signal superimposed in the second water meter image. According to the optical characteristics of the specular reflection light field corresponding to reflection, the cavity scattering light field corresponding to water mist, and the occlusion attenuation light field corresponding to dirt, the independent optical field components corresponding to the three types of interference are synchronously decoupled within the three-layer physical structure. During the calculation of the independent optical field components, the dial topology prior rule and the three-layer optical field propagation law are loaded to form a dual constraint, so as to lock the metering characteristic optical field boundary of the water meter's dial, pointer, and scale, and output the effective metering optical field signal of the three-layer physical structure after stripping the interference optical field components. Based on the effective measurement optical field signal of the three-layer physical structure, the second water meter image is subjected to cross-spatial layer dual-constraint feature fusion processing to compensate for the light field energy attenuation and feature distortion caused by reflection and water mist, so as to output a third water meter image containing the measurement optical field signal of feature fusion. Based on the associated light field distribution and topological information of the third water meter image and the three-layer physical structure, a three-dimensional spatial light field-driven global semantic reconstruction is performed on the metering feature defect area caused by dirt in the water meter to restore the complete digit wheel outline and pointer lines in the water meter, and output the light field data after semantic reconstruction. By combining the residual optical field reverse closed-loop verification mechanism, optical field data, and the third water meter image, the optical field data is compared with the standard optical field model of the water meter to correct the optical field residual generated in the third water meter image during the layered operation and feature fusion process, thereby obtaining the water meter reading in the third water meter image.

[0006] Furthermore, the execution steps of the three-layer optical field propagation model include: Offline sample collection and layered optical parameter extraction: For the three independent physical layers of mechanical water meters—outer glass, internal sealed water vapor cavity, and dial base—interference-free standard light field samples, single-interference light field samples, and multi-coupled interference light field samples were collected respectively. Basic optical parameters such as light field propagation path, light field energy attenuation coefficient, and optical reflection / scattering response characteristics were extracted for each layer. Based on the optical parameters extracted from each layer, a mathematical model of the mirror reflection light field is constructed for the outer glass, a mathematical model of the cavity scattering light field is constructed for the internal sealed water vapor cavity, and a mathematical model of the substrate imaging light field is constructed for the dial substrate. The calculation rules for the light field of each layer as a function of spatial position and interference type are clarified. Combining the optical crosstalk characteristics of the sealed cavity of the water meter, rules for light field transmission, energy transfer, and morphological linkage between the three physical spaces are established. The three single-layer light field mathematical models are coupled and fused to form a complete three-layer light field propagation mathematical model, and the overall operation logic of the model is solidified. The three-layer optical field propagation model that has been coupled is pre-stored in the computing unit. When the mechanical water meter data acquisition method is executed, the pre-stored model is retrieved to complete online initialization and establish a mapping relationship between the model and the first water meter image and the three-layer physical space of the water meter. In each computational stage, including pixel bidirectional source binding, directional light field component separation, cross-spatial layer feature fusion, three-dimensional semantic reconstruction, and residual light field closed-loop verification, the three-layer light field propagation model is continuously invoked. Based on the optical propagation laws and inter-layer linkage rules built into the model, it participates in light field calculation, feature constraints, and deviation judgment until the entire water meter data acquisition process is completed.

[0007] Furthermore, the steps for bidirectionally binding the pixels of the first water meter image with the three-layer physical spatial location of the water meter include: Perform a full-domain pixel traversal on the first water meter image, and combine the optical imaging response characteristics of the three physical structures of the water meter's outer glass, internal sealed water vapor cavity, and dial base to determine the actual target physical layer attached to each pixel. Establish a positive mapping relationship between the two-dimensional coordinates of image pixels and the three-dimensional coordinates of the three-layer physical space of the water surface, and complete the positive tracing from image pixels to physical spatial location; Based on the three-layer light field propagation model, the theoretical imaging position of the corresponding pixel is derived from the three-dimensional coordinates of the three-layer physical space of the water surface, thus completing the reverse tracing from the physical space position to the image pixel. By comparing the forward and reverse tracing results and completing position calibration, the unique binding relationship between the pixel and the three-layer physical spatial position of the water meter is locked. After binding, the pixel's position remains constant and does not shift in all subsequent calculation processes to generate the second water meter image.

[0008] Furthermore, the steps for separating directional optical field components and decoupling the three types of interfering independent optical field components include: Using the second water meter image and the three-layer optical field propagation model as common inputs, and based on the three-layer physical structure of the water meter—outer glass, internal sealed water vapor cavity, and dial substrate—and considering the adhesion characteristics of reflection, water mist, and dirt, the optical field action areas corresponding to the specular reflection light field, cavity scattering light field, and occlusion attenuation light field are divided. Based on the three-layer optical field propagation model, optical feature parameters corresponding to the three types of optical fields are predefined, while the dial topology prior rules and the three-layer optical field propagation law are applied throughout the process to form a dual constraint. In the three-layer physical space, directional optical field component separation operations are performed in parallel, simultaneously extracting and separating the three types of independent interfering optical field components: specular reflection light field, cavity scattering light field, and occlusion attenuation light field. During the separation process, the measurement feature light field boundaries of the dial wheel, pointer, and scale are locked in real time according to the dual constraints to avoid damage to the effective measurement features. Finally, the calculation results of the three-layer physical space are integrated, all interfering optical field components are stripped away, and the effective measurement optical field signal of the three-layer physical structure is output.

[0009] Furthermore, the dual constraints formed by the dial topological prior rules and the three-layer light field propagation law, in the process of real-time locking of the metering feature light field boundaries corresponding to the dial wheel, hands, and scale during the entire decoupling process of light field components, include: Retrieve the pre-stored prior rules of dial topology and the propagation law of the three-layer light field, and combine them with the three-layer physical structure of the water meter corresponding to the second water meter image to build a dual constraint control system that links dial topology and optical field properties. Based on the preset parameters of the character wheel arrangement, pointer shape, scale distribution and contour range in the dial topology prior rules, the theoretical geometric boundaries corresponding to the character wheel, pointer and scale are calibrated in the three-layer physical space; Based on the propagation law of the three-layer light field, and combined with the light field propagation path and energy distribution characteristics of the outer glass, the internal sealed water vapor cavity, and the dial base, the aforementioned theoretical geometric boundary is transformed into a metrological characteristic light field boundary threshold that can be used for calculation and judgment. Throughout the process of separating directional light field components and synchronously decoupling three types of interference independent light field components, real-time light field operation data in the three-layer physical space is collected, and the real-time light field position is continuously compared with the preset metrological characteristic light field boundary threshold. If interfering light fields such as mirror reflection light field, cavity scattering light field, and occlusion attenuation light field are detected to intrude into the boundary of the metrological characteristic light field, the calculation range of light field separation is immediately corrected through dual constraints to achieve forced isolation between the interfering light field and the effective metrological light field. Throughout the process, the boundary of the measurement feature light field corresponding to the dial wheel, pointer, and scale remains stable until the decoupling operation of the light field components is completed, and finally the effective measurement light field signal of the three-layer physical structure without damage is output.

[0010] Furthermore, the cross-spatial-layer dual-constraint feature fusion processing uses the effective metering optical field signal of the three-layer physical structure as the fusion object, specifically compensating for the optical field energy loss caused by reflection and the image feature distortion caused by water mist, so that the fused metering optical field signal completely retains the original metering information of the water meter; the specific steps of the cross-spatial-layer dual-constraint feature fusion processing include: The effective metering optical field signal of the three-layer physical structure is received as input. At the same time, the dial topology prior rule and the propagation law of the three-layer optical field are loaded again to form a dual constraint. Based on the three-layer physical structure of the water meter, the outer glass, the inner sealed water vapor cavity, and the dial base, the effective metering optical field signal of the three-layer physical structure is split into independent optical field subsets corresponding to each physical layer. Under dual constraints and control, state detection is performed on each light field subset to identify the light field energy loss area caused by reflection in the outer glass layer, the characteristic distortion area caused by water mist in the inner sealed water vapor cavity layer, and the normal metering light field area of ​​the dial base layer. Based on the propagation law of the three-layer light field, directional energy gain compensation is performed on the light field region with energy loss in the outer glass layer, and optical morphology correction is performed on the light field region with characteristic distortion in the inner sealed water vapor cavity layer. During the compensation and correction process, the dial topology prior rules are continuously referenced to ensure that the geometric shape of the dial wheel, pointer and scale corresponding to each layer of light field does not change. Based on the global coordinate system defined by the dial topology prior rules, cross-spatial layer coordinate registration is performed on the three independent light field subsets after compensation and correction to eliminate positional and temporal deviations between light field signals of different physical layers. With dual constraints throughout the process, the three-layer optical field subsets that have been registered are spliced ​​and fused across the entire domain to form a complete metrological optical field signal. The integrity of the fused overall metering optical field signal is verified. After confirming that there is no missing optical field energy or metering features, a third water meter image containing the fused metering optical field signal is output.

[0011] Furthermore, the three-dimensional spatial light field-driven global semantic reconstruction abandons the traditional two-dimensional local pixel interpolation image restoration method. Based on the associated light field distribution and topological information of the three-layer physical space, it performs global reconstruction on the damaged areas of the digit wheel and pointer caused by stains, completely restoring the outline of the digit wheel and the lines of the pointer. The specific steps of the three-dimensional spatial light field-driven global semantic reconstruction include: Based on the third water meter image and the corresponding feature fusion metering light field signal, the associated light field distribution data and dial topology information of the three physical spaces of the water meter are loaded simultaneously as the basic data source for semantic reconstruction. By combining the dial topology information and the light field distribution characteristics of the three-layer physical structure, the third water meter image is scanned across the entire area to accurately identify and mark the incomplete areas of the wheel outline and the broken areas of the pointer lines caused by dirt obscuring them, and to determine the location, range and type of each defect area. Centered on the defective area, the global correlation light field features of the three physical spaces of the water meter's outer glass, internal sealed water vapor cavity, and dial base are extracted, including light field direction, energy gradient, and inter-layer correlation rules. Combined with the geometric constraints of the dial topology information, the optical constraints of the three-layer light field propagation rules, and the extracted global correlation light field features, a three-dimensional semantic reconstruction benchmark model adapted to the water meter's metering features is constructed. Based on the three-dimensional semantic reconstruction benchmark model, light field-driven semantic completion is performed in different regions. For the missing area of ​​the character wheel, the complete character outline is restored according to the topological contour and the direction of the three-dimensional light field. For the broken area of ​​the pointer, the pointer lines are completed according to the inherent shape of the pointer and the continuity of the light field between layers. The completed local areas are subjected to cross-three-layer physical space feature joint adjustment to ensure that the light field morphology of the reconstructed word wheel and pointer is continuous and the geometric morphology is unified across different physical layers. All reconstructed areas are integrated with the original normal measurement light field areas to form a complete overall light field, and finally the semantically reconstructed light field data is output.

[0012] Furthermore, after correcting the optical field residual through the residual optical field reverse closed-loop verification mechanism, a standardized dial image with accurate physical location, pure optical field signal, and complete metering characteristics is generated. This standardized dial image is then incorporated into the subsequent water meter mechanical transmission rule verification process. The final steps for determining the water meter's final reading based on the verification results include: Based on the semantically reconstructed light field data and the third water meter image, the pre-stored standard light field model of the water meter is retrieved, and the dual constraints composed of the dial topology prior rules and the three-layer light field propagation law are loaded again as the benchmark for closed-loop verification and residual correction. Based on the three-layer physical structure of the water meter, consisting of the outer glass, the internal sealed water vapor cavity, and the dial base, the optical field data that has completed semantic reconstruction is decomposed into layers. The optical field data of each layer is compared with the standard optical field data of the corresponding layer in the standard optical field model of the water meter pixel by pixel and metering feature by feature. The global optical field residuals in terms of optical field amplitude, shape, and position are calculated. By combining the residual distribution location and feature deviation type, the residuals of the global optical field are classified and traced back to their source, distinguishing the hierarchical operation residuals, feature fusion residuals, and semantic reconstruction residuals, and accurately locating the physical layer and metrological region where each type of residual is located. Based on the residual type and its location, directional residual correction is performed under dual constraints: for optical residuals generated by hierarchical operations and feature fusion, the amplitude and propagation shape of the light field are adjusted according to the propagation law of the three-layer light field; for geometric residuals generated by semantic reconstruction, the outline and position of the dial wheel and pointer are corrected according to the dial topology prior rules. During the correction process, the geometric shape and optical properties of the measurement features are always kept within the standard range. After the initial correction, the overall optical field data is compared with the standard optical field model of the water meter again to perform a closed-loop secondary verification to determine whether the current remaining optical field residual is lower than the preset residual allowable threshold. If the residual exceeds the threshold, the third and fourth steps are repeated for iterative correction until the residual meets the threshold requirement. Once the secondary verification confirms that the optical field residual meets the standard, a standardized dial image with accurate physical location, pure optical field signal, and complete metering characteristics is generated based on the corrected overall optical field data. The standardized dial image is then connected to the mechanical transmission rule verification stage of the water meter, and the final verification is completed in conjunction with the mechanical transmission rules to determine the final reading of the mechanical water meter.

[0013] To achieve the above objectives, this application adopts the following technical solution: A mechanical water meter data acquisition device for performing a mechanical water meter data acquisition method is characterized by comprising: a data acquisition part, a data processing part, and a data generation part.

[0014] An electronic device, comprising: The memory, processor, and computer program stored in the memory, wherein the processor is configured to run the computer program to perform a method for acquiring data from a mechanical water meter.

[0015] The aforementioned data acquisition method for mechanical water meters addresses several pain points caused by adverse conditions such as humidity, dirt accumulation, and strong light, including blurred imaging, missing metering features, and inaccurate readings due to reflective glass, internal water mist, and dial stains. It achieves several technological breakthroughs. Based on the three-layer physical structure of the water meter, a light field propagation model is constructed. By binding pixels to physical space through bidirectional tracing, the misalignment problem between pixels and physical space in traditional two-dimensional image processing is solved. Three types of interfering light fields—reflection, water mist, and stains—are directionally separated, and dual constraints are used to protect core metering features such as the dial, pointer, and scale. Cross-spatial layer feature fusion compensates for light field loss and image distortion, and three-dimensional semantic reconstruction repairs contour and line defects caused by stains, fully restoring the water meter's metering form. Finally, residual light field reverse closed-loop verification corrects computational deviations in each stage, further reducing recognition errors. This method can be adapted to scenarios with both single and multiple interferences, making up for the shortcomings of traditional image recognition such as poor anti-interference, feature repair distortion, and gradual accumulation of errors. It significantly improves the accuracy and stability of mechanical water meter reading recognition, while avoiding the problems of large subjective errors and low efficiency in manual meter reading. Attached Figure Description

[0016] Figure 1 This is a hardware structure block diagram of an electronic device according to an embodiment of this application.

[0017] Figure 2 This is a flowchart of a mechanical water meter data acquisition method according to an embodiment of this application.

[0018] Reference numerals in the attached figures: 100, electronic device; 11, for processor; 12, memory. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in specific embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0020] It should be noted that the terms "first," "second," and similar terms used in this application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. "A plurality of" indicates at least two. "Comprising" and similar terms mean that the elements or objects preceding "comprising" cover the elements or objects listed after "comprising" and their equivalents, and do not exclude other elements or objects. "Connection" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0021] The singular forms “a,” “the,” and “the” used in this application specification and appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0022] The mechanical water meter data acquisition method provided in this embodiment can be executed in electronic device 100 or similar device. Figure 1 This is a hardware structure block diagram of an electronic device 100 that implements an embodiment of this application. For example... Figure 1 As shown, the electronic device 100 may include one or more ( Figure 1 (Only one is shown) memory 12 and processor 11. The electronic device 100 is a control terminal for mechanical water meter data acquisition. It is used to control the operation of the mechanical water meter data acquisition device to improve the accuracy of mechanical water meter data acquisition.

[0023] The memory 12 stores program instructions, such as application software programs and modules, like a computer program for acquiring data from a mechanical water meter in this embodiment. The processor 11 executes the program instructions stored in the memory 12. By running the computer program stored in the memory 12, it can perform various functional applications and data processing, thereby improving the acquisition of data from the mechanical water meter.

[0024] The processor 11 may include, but is not limited to, a microprocessor 11 (Microcontroller Unit, abbreviated as MCU) or a programmable gate array (FPGA).

[0025] Those skilled in the art will understand that Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device 100 described above. For example, the electronic device 100 may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0026] In some embodiments, a portion of the mechanical water meter data acquisition method described below is executed in electronic device 100 to acquire mechanical water meter data.

[0027] This embodiment also provides a method for acquiring data from a mechanical water meter. Figure 2 This is a flowchart of a method for acquiring data from a mechanical water meter according to an embodiment of this application. This method can improve the accuracy of data acquisition from mechanical water meters.

[0028] like Figure 2 As shown in the embodiments provided in this application, a method for acquiring data from a mechanical water meter is characterized by comprising: The original water meter reading image is obtained as the first water meter image. Based on the three-layer physical structure of the water meter's outer glass, internal sealed water vapor cavity, and dial base, a three-layer light field propagation model adapted to the optical characteristics of the water meter cavity is constructed. Based on the first water meter image and the three-layer light field propagation model, the pixels of the first water meter image are bidirectionally traced and bound to the three-layer physical spatial position of the water meter to output the second water meter image. Any pixel of the second water meter image is accurately mapped to the three-layer physical spatial position of the water meter and its position is bound. Based on the three-layer optical field propagation model, directional optical field component separation calculation is performed on the optical signal superimposed in the second water meter image. According to the optical characteristics of the specular reflection light field corresponding to reflection, the cavity scattering light field corresponding to water mist, and the occlusion attenuation light field corresponding to dirt, the independent optical field components corresponding to the three types of interference are synchronously decoupled within the three-layer physical structure. During the calculation of the independent optical field components, the dial topology prior rule and the three-layer optical field propagation law are loaded to form a dual constraint, so as to lock the metering characteristic optical field boundary of the water meter's dial, pointer, and scale, and output the effective metering optical field signal of the three-layer physical structure after stripping the interference optical field components. Based on the effective measurement optical field signal of the three-layer physical structure, the second water meter image is subjected to cross-spatial layer dual-constraint feature fusion processing to compensate for the light field energy attenuation and feature distortion caused by reflection and water mist, so as to output a third water meter image containing the measurement optical field signal of feature fusion. Based on the associated light field distribution and topological information of the third water meter image and the three-layer physical structure, a three-dimensional spatial light field-driven global semantic reconstruction is performed on the metering feature defect area caused by dirt in the water meter to restore the complete digit wheel outline and pointer lines in the water meter, and output the light field data after semantic reconstruction. By combining the residual optical field reverse closed-loop verification mechanism, optical field data, and the third water meter image, the optical field data is compared with the standard optical field model of the water meter to correct the optical field residual generated in the third water meter image during the layered operation and feature fusion process, thereby obtaining the water meter reading in the third water meter image.

[0029] In the above steps, the mechanical water meter acquisition method of this application is mainly applied to scenarios where mechanical water meters are centrally deployed, such as underground integrated meter wells, outdoor open-air water supply networks, and corridors in old residential communities. In these scenarios, mechanical water meters operate in humid, complex lighting, and dusty environments. Strong light reflection easily forms on the surface of the meter mirror, water vapor easily condenses in the sealed cavity between the mirror and the dial, and scale and oil adhere to the dial area, resulting in complex operating conditions with single or multiple types of interference, including mirror reflection, internal water vapor, and dial stains. Traditional water meter image recognition technology generally uses two-dimensional pixel processing algorithms such as grayscale threshold segmentation, filtering and noise reduction, and texture clustering, combined with pre-interference detection, adaptive parameter tuning, and serial class-by-class interference cancellation. This not only significantly reduces processing effectiveness in scenarios with multiple interference coupling but also easily leads to large errors in water meter reading recognition due to problems such as spatial misalignment between image pixels and the water meter entity, and feature restoration distortion.

[0030] This application utilizes a dedicated embedded intelligent recognition device to automate the entire method. This device is an integrated unit adapted to the on-site installation environment of water meters. Its hardware comprises five core modules: an industrial image acquisition unit, a main processing processor, non-volatile memory, a high-speed data cache unit, and a reading output unit. These modules interact in real-time via an internal bus. Specifically, the industrial image acquisition unit is fixedly installed in front of the mechanical water meter and is responsible for continuously acquiring the original image of the meter face. The main processing processor is the core computing platform, executing all light field calculations, image processing, and logical judgment processes. The non-volatile memory is used for offline pre-storing of fixed reference data such as the three-layer light field propagation model calibrated based on the three-layer physical structure of the water meter, the meter face topology prior rules, and the standard light field model of the water meter. The high-speed data cache unit is used to temporarily store dynamic data such as intermediate images and light field signals generated in each step. The reading output unit is used to finally output the identified water meter value. The entire system abandons the traditional two-dimensional image processing logic and uses the three inherent physical structures of the mechanical water meter's outer glass mirror, internal sealed water vapor cavity, and dial base to perform three-dimensional light field calculations, solving the recognition problem caused by multiple interferences from the perspective of physical optics and spatial structure.

[0031] The initial step of the entire method involves acquiring the original image, constructing a three-layer light field propagation model, completing bidirectional pixel tracing and binding, and generating the second water meter image. This step is collaboratively completed by the industrial image acquisition unit, the main processing processor, and non-volatile memory. First, the industrial image acquisition unit captures images of the mechanical water meter dial at a preset sampling frequency, defining the captured image as the first water meter image. This first water meter image fully includes the core metering features such as the water meter dial, pointer, and scale, while also carrying any one or more interferences generated on-site, such as reflections, water mist, and stains. The image data is transmitted in real-time via an internal bus to a high-speed data cache unit for temporary storage.

[0032] Subsequently, the main processing processor retrieves the pre-calibrated optical parameters of the water meter cavity from the non-volatile memory. Combining this with the unique physical structure of the three layers of the mechanical water meter—the outer glass, the internal sealed water vapor cavity, and the dial base—it constructs a three-layer light field propagation model adapted to the optical propagation characteristics of the water meter cavity. This model is the core benchmark for all light field calculations in this scheme, quantifying the optical laws of reflection, scattering, and transmission of light within the three-layer physical structure of the water meter. After the model is constructed, the main processing processor uses the first water meter image in the cache and the newly built three-layer light field propagation model as common inputs to perform a bidirectional source binding operation between the image pixels and the three-layer physical space of the water meter.

[0033] The binding operation consists of two dimensions: forward mapping and reverse tracing. Forward mapping maps each pixel of the first water meter's two-dimensional image to its corresponding three-dimensional physical location on the outer glass, internal water vapor cavity, and dial base. Reverse tracing, based on a three-layer light field propagation model, infers the theoretical pixel location from the water meter's three-dimensional physical location. The processor performs pixel-by-pixel comparison and position calibration on the forward and reverse tracing results, ultimately establishing a unique binding relationship between pixels and physical space, and locking the pixel position to prevent offset in all subsequent computational processes. After pixel binding is complete, the main processing processor generates the second water meter image and stores it in a high-speed data cache unit as the sole input data for the next stage. This step establishes a deep correlation between image pixels and the physical space and optical laws of the water meter, fundamentally avoiding the pixel misalignment defects of traditional two-dimensional image algorithms.

[0034] Step Two: Directional Optical Field Component Separation and Decoupling. Based on the dual-constraint protection metering characteristics, the effective metering optical field signal of the three-layer physical structure is output. This step is executed independently by the main processing processor. The input is the second water meter image and the three-layer optical field propagation model stored in the high-speed data cache unit. The main processing processor performs directional optical field component separation on the overall optical signal superimposed in the second water meter image according to the three-layer optical field propagation model. Combining the optical properties of different interferences, the optical signal corresponding to the mirror reflection is divided into a mirror reflection optical field, the optical signal corresponding to the internal water mist is divided into a cavity scattering optical field, and the optical signal corresponding to the dial stains is divided into an obstruction attenuation optical field. These three types of interference optical fields are then simultaneously decoupled within the three-layer physical space of the water meter. This synchronous decoupling mode differs from the traditional serial, class-by-class interference cancellation method, as it can simultaneously remove multiple coupled interferences, significantly improving processing efficiency.

[0035] Throughout the entire process of optical field component decoupling, the main processing processor continuously loads a dual constraint consisting of dial topology prior rules and three-layer optical field propagation laws. The dial topology prior rules, pre-stored in non-volatile memory, record the standard contours of the meter's dial arrangement, pointer shape, scale distribution, and various metrological features. The three-layer optical field propagation laws define the propagation path and energy distribution of light in each layer of the meter's space. This dual constraint continuously locks the boundaries of the metrological feature optical fields corresponding to the dial, pointer, and scale, strictly preventing interference light field separation operations from damaging effective metrological features. Once all three types of interference light fields have been stripped away, the main processing processor extracts the optical signal containing only the metrological features, generates the effective metrological optical field signal of the three-layer physical structure, and caches it as input for subsequent feature fusion steps.

[0036] Step 3: Cross-spatial-layer dual-constraint feature fusion to compensate for optical field defects and generate a third water meter image. The input for this step is the effective metering optical field signal of the three-layer physical structure output from the previous step. The main processing unit performs cross-spatial-layer dual-constraint feature fusion processing in conjunction with dual constraints. Since the three physical layers of the water meter are independent of each other, the effective optical field signal after removing interference is dispersed in different spatial layers. At the same time, reflection will cause optical field energy attenuation, and water mist will cause optical field morphology distortion. Therefore, it is necessary to integrate and repair the layered optical fields.

[0037] Under the dual control of the dial topology prior rules and the propagation law of the three-layer light field, the main processing processor first integrates and splices the effective light field signals of the three-layer physical structure, and then performs targeted defect compensation: directional gain compensation is performed for the light field energy loss caused by reflection in the outer glass area, and optical morphology correction is performed for the characteristic distortion caused by water mist in the internal water vapor cavity. Throughout the compensation and fusion process, the dual constraints ensure that the geometry of the dial, pointer, and scale remains unchanged. After completing the cross-layer fusion and defect repair, the main processing processor generates a third water meter image with the feature-fused metering light field signal, and caches the image and the corresponding light field data together, providing complete data support for the stain and defect repair process.

[0038] Step Four: 3D Spatial Light Field-Driven Global Semantic Reconstruction, Repairing Feature Defects and Outputting Semantic Reconstruction Light Field Data. This step uses the cached third water meter image, the light field distribution associated with the three physical spaces, and the dial topology information as the basis for computation. The main processing processor executes the 3D spatial light field-driven global semantic reconstruction operation. Stains on the water meter surface can obscure the dial and pointers, causing incomplete metering feature contours and broken lines. Traditional solutions use two-dimensional local pixel interpolation for repair, which easily leads to morphological distortion. This embodiment abandons this traditional method and performs 3D reconstruction based on the global light field distribution and topology information of the three physical spaces of the water meter.

[0039] The main processing processor first performs a full-domain scan of the third water meter image to accurately locate areas with missing metering features caused by dirt. Then, it extracts the global light field direction, energy gradient, and inter-layer correlation features of the three physical layers, and constructs a three-dimensional semantic reconstruction baseline model based on geometric and optical constraints. Next, based on this model, it performs semantic completion by region, restoring the wheel shape according to the standard topological contour and completing the pointer lines according to the continuity of the inter-layer light field. After reconstruction, the processor performs joint calibration of the features of the three physical layers to ensure complete uniformity of the light field and geometric shapes of different physical layers. Finally, it generates and caches the semantically reconstructed light field data, which is then sent to the final verification stage.

[0040] Step 5: Residual Optical Field Reverse Loop Verification, Correcting Calculation Residuals and Obtaining Water Meter Readings. This step is the final stage of the entire method. The inputs are the optical field data after semantic reconstruction and the image of the third water meter. The main processing processor activates the residual optical field reverse loop verification mechanism. First, it retrieves the pre-stored standard optical field model of the water meter in non-volatile memory. This model serves as the standard optical reference when the water meter is free from any interference. The processor performs a full-domain point-by-point comparison between the measured optical field data, the image of the third water meter, and the standard optical field model to identify the minute optical field residuals accumulated during the hierarchical calculations, feature fusion, and semantic reconstruction processes.

[0041] For the identified residuals, the processor performs directional correction using dual constraints: optical residuals adjust the amplitude and propagation shape of the light field according to the propagation laws of the three-layer light field, while geometric residuals correct the contour and position according to the dial topological prior rules. After the residual correction is completed, the processor compares and verifies the residuals with the standard light field model again until the residuals meet the preset accuracy requirements. Finally, the main processing processor analyzes and identifies the reading of the mechanical water meter from the third water meter image after residual correction and outputs it to the outside through the reading output unit.

[0042] In summary, this embodiment is based on the unique three-layer physical structure of the water meter and the three-dimensional light field theory. It operates according to a complete serial process of "image acquisition and spatial binding, interference light field decoupling, light field fusion compensation, three-dimensional feature reconstruction, and residual closed-loop verification." The data flow of each step is coherent, and the functions are mutually coordinated. The entire solution can stably adapt to single and coupled interference conditions such as reflection, water mist, and dirt. It completely solves the shortcomings of traditional two-dimensional image processing technology from the perspective of physical optics, and significantly improves the accuracy and stability of mechanical water meter reading recognition in harsh environments.

[0043] As one implementation method, the execution steps of the three-layer optical field propagation model include: Offline sample collection and layered optical parameter extraction: For the three independent physical layers of mechanical water meters—outer glass, internal sealed water vapor cavity, and dial base—interference-free standard light field samples, single-interference light field samples, and multi-coupled interference light field samples were collected respectively. Basic optical parameters such as light field propagation path, light field energy attenuation coefficient, and optical reflection / scattering response characteristics were extracted for each layer. Based on the optical parameters extracted from each layer, a mathematical model of the mirror reflection light field is constructed for the outer glass, a mathematical model of the cavity scattering light field is constructed for the internal sealed water vapor cavity, and a mathematical model of the substrate imaging light field is constructed for the dial substrate. The calculation rules for the light field of each layer as a function of spatial position and interference type are clarified. Combining the optical crosstalk characteristics of the sealed cavity of the water meter, rules for light field transmission, energy transfer, and morphological linkage between the three physical spaces are established. The three single-layer light field mathematical models are coupled and fused to form a complete three-layer light field propagation mathematical model, and the overall operation logic of the model is solidified. The three-layer optical field propagation model that has been coupled is pre-stored in the computing unit. When the mechanical water meter data acquisition method is executed, the pre-stored model is retrieved to complete online initialization and establish a mapping relationship between the model and the first water meter image and the three-layer physical space of the water meter. In each computational stage, including pixel bidirectional source binding, directional light field component separation, cross-spatial layer feature fusion, three-dimensional semantic reconstruction, and residual light field closed-loop verification, the three-layer light field propagation model is continuously invoked. Based on the optical propagation laws and inter-layer linkage rules built into the model, it participates in light field calculation, feature constraints, and deviation judgment until the entire water meter data acquisition process is completed.

[0044] The above steps involve offline sample acquisition and layered optical parameter extraction, which is the initial step in constructing the entire three-layer optical field propagation model. This is performed offline during the equipment factory calibration or on-site batch debugging of water instruments. First, let's explain the core terms involved in this step: A light field sample refers to the set of optical signals presented by the three-layer physical structure of a mechanical water meter under different operating conditions. Based on the operating state, it is divided into three categories: interference-free standard light field samples, single-interference light field samples, and multi-coupled interference light field samples. Interference-free standard samples correspond to ideal operating conditions where the water meter is clean, without water mist, and without reflection. Single-interference samples correspond to operating conditions where only reflection, water mist, and dirt exist, respectively. Multi-coupled interference samples simulate the most common operating conditions where two or more types of interference coexist. Basic optical parameters include the light field propagation path, the light field energy attenuation coefficient, and optical reflection / scattering response characteristics. The light field propagation path describes the trajectory of light rays in each physical layer of the water meter. The light field energy attenuation coefficient characterizes the degree of energy loss of the optical signal during propagation. The optical reflection / scattering response characteristics distinguish the specular reflection characteristics of the outer glass, the scattering characteristics of the internal water vapor cavity, and the imaging characteristics of the dial substrate. In terms of data acquisition, technicians used specialized optical acquisition equipment to collect three types of light field samples from three independent physical layers of the mechanical water meter: the outer glass, the internal sealed water vapor cavity, and the dial base. During the acquisition process, the interference types of each physical layer were strictly distinguished to ensure accurate matching between the samples and the corresponding physical structures and operating conditions. After sample acquisition, the computing unit analyzed and calculated the light field data for each layer, extracting the corresponding basic optical parameters layer by layer. From a step-by-step perspective, this step is the data source for all subsequent modeling work; the extracted layered optical parameters will directly serve as input data for the next single-layer model construction step. From a technical perspective, multi-condition, layered sample acquisition and parameter extraction can comprehensively capture the true optical laws of different physical layers of the water meter, avoiding the disconnect between the subsequent model and actual operating conditions caused by single samples or incomplete parameters, thus building a solid data foundation for high-precision light field models.

[0045] After completing the extraction of layered optical parameters, the second step is to construct single-layer optical field mathematical models based on the optical parameters of each layer. This step builds upon the layered basic optical parameters output in the previous step. The core terminology for single-layer optical field mathematical models is divided into three categories: the specular reflection optical field mathematical model corresponding to the outer glass layer, the cavity scattering optical field mathematical model corresponding to the internal sealed water vapor cavity, and the substrate imaging optical field mathematical model corresponding to the dial substrate. Each single-layer model is used to independently define the operational rules for light rays changing with spatial position and interference type within the corresponding physical layer. The computational unit is divided according to physical layers. The outer glass optical parameters extracted in the first step are substituted into the computational formula to build the specular reflection optical field mathematical model, specifically characterizing the optical changes corresponding to the watch mirror reflection; the optical parameters of the internal sealed water vapor cavity are used as input to construct the cavity scattering optical field mathematical model, simulating the light scattering phenomenon caused by water vapor; and finally, the substrate imaging optical field mathematical model is established by combining the optical parameters of the dial substrate to restore the imaging optical characteristics of the dial itself. The three single-layer models are independent of each other, each corresponding to the specific optical behavior of the three-layer structure of the water meter. The output of this step is three independent single-layer optical field mathematical models, which are the core inputs for subsequent inter-layer coupling and fusion. The technical effect of this step is to realize the mathematical expression of the layered optical laws of the water meter, transforming abstract optical phenomena into mathematical operation rules that can be recognized and invoked by the computing unit. The decomposed modeling approach also reduces the difficulty of constructing the overall model, while ensuring the characterization accuracy of the optical characteristics of each layer.

[0046] The third step involves coupling and fusing the single-layer models by combining the characteristics of the water meter's sealed cavity, forming a complete three-layer optical field propagation mathematical model. Here, optical crosstalk refers to the fact that mechanical water meters have a sealed cavity structure, meaning light is not confined to a single physical layer but undergoes transmission, energy transfer, and morphological linkage between the three layers. This is the core reason why ordinary single-layer optical models cannot adapt to the working conditions of water meters. Based on the three single-layer optical field mathematical models generated in the previous step, the computational unit, considering the structural characteristics of the water meter's sealed cavity, further defines the rules for light field transmission, energy transfer, and morphological linkage between the three physical spaces. This organically couples the three independent single-layer models, integrating them into a unified three-layer optical field propagation mathematical model, and solidifies the overall computational logic of the entire model, eliminating the need for temporary modifications to the internal rules. This step connects to the single-layer modeling stage and outputs the final customized light field model. Its technical effect is to break through the limitations of the single-layer model, restore the real state of light field interaction inside the sealed cavity of the water meter, make the model more consistent with the physical optical environment of the water meter entity, and avoid subsequent calculation deviations caused by ignoring the crosstalk between interlayer light fields.

[0047] The fourth step is model storage and online initialization, which facilitates the transition from offline modeling to online computation. The complete three-layer optical field propagation mathematical model, after coupling and fusion, is permanently pre-stored in non-volatile memory by the computation unit. When the mechanical water meter data acquisition method and equipment begin on-site operation, the computation unit actively retrieves the pre-stored model from memory, completes the online initialization of the model, and establishes a mapping relationship between the model and the first water meter image and the three-layer physical space of the water meter. This binds the abstract optical field model to the currently processed on-site image and the physical water meter. From the perspective of step association, this step is the key connection point between the offline modeling stage and the online operation stage. After storage and initialization, the complete model generated offline directly serves the pixel processing stage of the first water meter image. Technically, the pre-storage design allows for one-time model calibration and repeated retrieval, eliminating the need for remodeling for each operation and significantly improving equipment operating efficiency. The mapping relationship established through online initialization ensures the consistency of model computation, image processing, and the water meter physical space.

[0048] The final step involves the continuous invocation of the model throughout the entire methodology. In each computational stage—pixel bidirectional source binding, directional light field component separation, cross-spatial layer feature fusion, 3D semantic reconstruction, and residual light field closed-loop verification—the computational unit continuously invokes the initialized three-layer light field propagation model. Based on the model's built-in optical propagation laws and inter-layer linkage rules, it performs light field calculations, feature constraints, and deviation judgments until the entire single water meter data acquisition process is completed. This step is the ultimate manifestation of the model's value. The initialized model serves as a unified underlying benchmark, ensuring consistent computational rules and logical coherence throughout the entire process, thereby improving the overall accuracy and stability of image preprocessing and reading recognition from the ground up.

[0049] As one implementation method, the step of bidirectionally binding the pixels of the first water meter image with the three-layer physical spatial location of the water meter includes: Perform a full-domain pixel traversal on the first water meter image, and combine the optical imaging response characteristics of the three physical structures of the water meter's outer glass, internal sealed water vapor cavity, and dial base to determine the actual target physical layer attached to each pixel. Establish a positive mapping relationship between the two-dimensional coordinates of image pixels and the three-dimensional coordinates of the three-layer physical space of the water surface, and complete the positive tracing from image pixels to physical spatial location; Based on the three-layer light field propagation model, the theoretical imaging position of the corresponding pixel is derived from the three-dimensional coordinates of the three-layer physical space of the water surface, thus completing the reverse tracing from the physical space position to the image pixel. By comparing the forward and reverse tracing results and completing position calibration, the unique binding relationship between the pixel and the three-layer physical spatial position of the water meter is locked. After binding, the pixel's position remains constant and does not shift in all subsequent calculation processes to generate the second water meter image.

[0050] In the above steps, the operation of performing a full-domain pixel traversal on the first water meter image and determining the target physical layer to which the pixels belong is performed. In this step, the first water meter image is the original dial image of a mechanical water meter captured on-site by the device's image acquisition module. The image contains metering features such as dials, pointers, and scales, and is superimposed with interference signals such as reflections, water mist, and stains generated on-site. The full-domain pixel traversal refers to the main processing unit performing a comprehensive search of all pixels in the first water meter image row by row and column by column according to preset scanning rules. The three-layer physical structure of the water meter specifically refers to the inherent outer glass, the internal sealed water vapor cavity, and the dial base of the mechanical water meter. The optical imaging response characteristics are the inherent differences in the reflection, transmission, and imaging performance of incident light formed by the three-layer physical structure due to differences in materials and spatial positions, and are also the core basis for distinguishing the pixel attachment position. In terms of execution logic, the image The acquisition module transmits the captured first water meter image to the high-speed data cache unit for temporary storage. The main processing unit retrieves the image data from the cache and initiates a full-domain pixel traversal. Simultaneously, it loads the three-layer physical structure optical imaging response features pre-stored in the device and analyzes and judges each pixel one by one to determine whether the pixel is actually attached to the outer glass, the internal sealed water vapor cavity, or the dial substrate. This step serves as the first basic step in the two-way traceability binding. The output pixel layer discrimination result is directly used as the input data for the next forward traceability operation. From a technical perspective, by combining full-domain traversal with optical features to complete the physical classification of pixels, it is possible to accurately distinguish the pixel groups corresponding to different spatial layers, avoid the problem of cross-layer pixel aliasing, and lay a solid foundation for the subsequent mapping work between two-dimensional pixels and three-dimensional physical space, ensuring the basic accuracy of location association from the source. After completing pixel-level discrimination, the next step is to establish a forward mapping between the two-dimensional coordinates of image pixels and the three-dimensional coordinates of the three physical spaces of the water meter, thus achieving forward tracing of pixels to physical spaces. The two-dimensional coordinates of image pixels involved in this step are the horizontal and vertical position parameters defined in the planar image coordinate system of the first water meter image. The three-dimensional coordinates of the three physical spaces of the water meter are the three-dimensional spatial point parameters established based on the physical structure of the water meter. Forward tracing is the one-way mapping process from planar image pixels to the three-dimensional physical position of the water meter entity. The main computing unit uses the pixel-level discrimination results obtained in the previous step as the basis for calculation. For each pixel that has been determined to belong, it matches its two-dimensional coordinates with the three-dimensional spatial coordinates of the corresponding physical layer, establishes a one-to-one forward mapping relationship, and completely records the physical position of the entity corresponding to each pixel. This step follows the pixel-level discrimination results, and the generated forward mapping data table is the core comparison benchmark for subsequent reverse tracing. Its technical effect is to completely break the inherent mode of traditional image processing that is limited to the two-dimensional plane, endowing the planar image with real three-dimensional physical space attributes, so that image pixels are no longer isolated planar points, but optical carriers deeply associated with the physical structure of the water meter.Next, the operation of reverse tracing is performed, which involves deriving the pixel position from the three-dimensional coordinates of the water meter based on the three-layer light field propagation model. The three-layer light field propagation model is a water meter-specific optical model that has been modeled, stored, and initialized. This model fully quantifies the propagation, reflection, and imaging laws of light within the three-layer structure of the water meter. Reverse tracing is the reverse derivation process compared to forward tracing; that is, starting from the three-dimensional physical coordinates of the water meter entity and combining optical laws, it deduces the theoretical pixel position corresponding to it in the image. The main computing unit retrieves the already running three-layer light field propagation model to perform the forward tracing step. Starting with the determined three-dimensional coordinates of each layer of the water meter, the theoretical pixel position of each three-dimensional physical point in the first water meter image is calculated by reverse calculation based on the light field propagation rules built into the model. This step relies on the three-dimensional coordinate system of the previous forward tracing and the underlying light field model. The output reverse theoretical pixel position data will be used for subsequent result comparison. Technically, unlike simple geometric coordinate conversion, this step combines the actual optical propagation law of the water meter to complete the reverse derivation, making the tracing process fit the actual imaging principle of the water meter, and further improving the rationality and accuracy of pixel and physical position matching. After obtaining the forward tracing mapping relationship and the reverse tracing theoretical position, the following operations are performed: comparing the two types of tracing results, calibrating the position, and locking the unique binding relationship between pixels and physical space. The tracing result comparison refers to verifying the difference between the actual pixel position obtained from the forward method and the theoretical pixel position derived from the reverse method point by point. The position calibration is to correct and optimize the slight positional deviation between the two. The unique binding relationship means that after calibration, each pixel corresponds to a unique three-dimensional physical point of the water meter, and this binding relationship remains constant and does not shift in all subsequent operation processes of the method. The main operation unit performs a global operation on the two sets of tracing data. The process involves comparing and calibrating the positions of each deviated pixel. After calibration, the binding rules between pixels and their three-layer physical spatial positions are officially locked, ensuring that the pixel positions remain unchanged throughout the entire process. This step integrates all data from both the forward and reverse traceability stages and is the core calibration step connecting the initial traceability calculations with the final image generation. Its technical effect lies in eliminating the inherent errors of a single traceability mode through bidirectional comparison, forming a stable and reliable pixel-physical position binding system. This effectively prevents pixel drift and spatial misalignment during subsequent light field calculations and feature processing, ensuring the positional consistency of the entire process. Finally, a second water meter image is generated based on the locked binding relationship. The second water meter image is a standardized intermediate image generated after completing all bidirectional traceability binding operations and is also the designated input carrier for the directional light field component separation step. The main computing unit reconstructs the image based on the calibrated and locked pixel binding relationship, generates the second water meter image, and stores it in the high-speed data cache unit. This step is the final output of the entire process. The generated second water meter image is directly transferred to the next calculation step, and the pixel binding rules established in this step will be used in all subsequent stages of the mechanical water meter data acquisition method.From the perspective of overall logic and step coordination, the execution process forms a complete and closed-loop operation chain of "full-domain pixel traversal layering → forward coordinate mapping tracing → optical model reverse tracing → bidirectional comparison and calibration → generation of target image". Each sub-step is interconnected, and the output of the preceding step always serves as the input for the subsequent step, resulting in a coherent data flow and rigorous logic. In the entire mechanical water meter data acquisition method system, this step plays a crucial role in connecting the preceding and following steps. It receives the original first water meter image above and provides a positional reference for all light field-related calculations below. Its core innovation lies in abandoning the traditional two-dimensional image processing method of directly calculating pixels. Instead, it combines the unique three-layer physical structure of the water meter with a dedicated three-layer light field propagation model to achieve bidirectional tracing binding, effectively solving the pain point of image pixels being disconnected from the physical space of the water meter in traditional technologies. In complex operating conditions such as reflection, water mist, and stains, whether alone or coupled interference, a stable pixel-physical position binding relationship can avoid amplifying recognition errors due to pixel misalignment. This not only provides a unified and standardized image carrier for subsequent tasks such as light field decoupling, feature fusion, semantic reconstruction, and residual verification, but also strengthens the computational foundation of the entire method from a physical space perspective. Combined with the full-process light field processing system, it systematically improves the accuracy and operational stability of mechanical water meter reading acquisition under harsh conditions, fully demonstrating the creativity and practicality of this technical solution compared to existing conventional image processing technologies.

[0051] As one implementation method, the steps of directional optical field component separation calculation and decoupling of three types of interfering independent optical field components include: Using the second water meter image and the three-layer optical field propagation model as common inputs, and based on the three-layer physical structure of the water meter—outer glass, internal sealed water vapor cavity, and dial substrate—and considering the adhesion characteristics of reflection, water mist, and dirt, the optical field action areas corresponding to the specular reflection light field, cavity scattering light field, and occlusion attenuation light field are divided. Based on the three-layer optical field propagation model, optical feature parameters corresponding to the three types of optical fields are predefined, while the dial topology prior rules and the three-layer optical field propagation law are applied throughout the process to form a dual constraint. In the three-layer physical space, directional optical field component separation operations are performed in parallel, simultaneously extracting and separating the three types of independent interfering optical field components: specular reflection light field, cavity scattering light field, and occlusion attenuation light field. During the separation process, the measurement feature light field boundaries of the dial wheel, pointer, and scale are locked in real time according to the dual constraints to avoid damage to the effective measurement features. Finally, the calculation results of the three-layer physical space are integrated, all interfering optical field components are stripped away, and the effective measurement optical field signal of the three-layer physical structure is output.

[0052] In the above steps, the second water meter image and the three-layer light field propagation model are used as common computational inputs. Combining the physical structure of the three layers of the water meter with the adhesion characteristics of reflection, water mist, and stains, the effective areas corresponding to the three types of light fields are divided. First, the core terms are explained: the second water meter image is an intermediate image generated after completing the bidirectional pixel tracing and binding. Each pixel in the image has a unique and fixed physical position mapping relationship with the outer glass layer of the water meter, the internal sealed water vapor cavity, and the dial base. This image is temporarily stored in the device's high-speed cache and is the core image input for this step. The three-layer light field propagation model is a dedicated optical model that has been calibrated offline, pre-stored, and initialized online. It fully quantifies the propagation, transmission, and energy change laws of the light field within the three-layer structure of the water meter and serves as the computational rule input for this step. The adhesion characteristics of reflection, water mist, and stains are summarized based on the interference distribution patterns derived from the physical structure of the water meter. Reflection from the mirror adheres only to the outer glass surface, water mist diffuses within the sealed water vapor cavity between the mirror and the dial, and stains mostly adhere to the dial base, with some extending across layers. The mirror reflection light field, cavity scattering light field, and obstruction attenuation light field are the optical signal carriers corresponding to these three types of interference. The mirror reflection light field is formed by reflection from the glass surface, the cavity scattering light field is formed by the scattering of light by water mist inside the cavity, and the obstruction attenuation light field is formed by the attenuation of the light field caused by stains blocking light. In terms of data acquisition and execution logic, the main processing unit first retrieves the second water meter image from the cache and simultaneously calls the initialized three-layer light field propagation model in the device rule base. Combining the three-layer physical structure of the water meter and the actual adhesion locations of the three types of interference, it divides the independent light field action areas of the mirror reflection light field, cavity scattering light field, and obstruction attenuation light field into three-dimensional spaces, achieving physical partitioning of different interference light fields. From the perspective of the steps involved, this step is a preliminary preparation for optical field decoupling. Based on the results of the previous pixel binding, the physical ownership of each pixel is clarified, and then the region is divided in combination with the interference distribution, which defines the computational boundary for subsequent directional separation. From the perspective of technical effect, dividing the optical field action area according to the actual attachment position of the interference can physically isolate different types of interference optical fields, avoid the problem of mutual mixing and difficulty in separation of optical fields when multiple interferences are coupled, and make subsequent optical field calculations more targeted, thereby improving the accuracy of interference decoupling from the source.

[0053] After the light field's effective area is defined, the second execution phase begins: based on a three-layer light field propagation model, optical characteristic parameters corresponding to three types of light fields are predefined, while the dual constraints formed by the dial topology prior rules and the three-layer light field propagation law are applied throughout the process. The optical characteristic parameters are specific optical parameters calculated based on the three-layer light field propagation model to distinguish different interfering light fields, including the light field energy range, propagation angle, and signal attenuation rate. Different interferences correspond to light fields with differentiated parameter characteristics. The dial topology prior rules are the water meter's inherent structural parameters pre-stored in the device, recording the layout, outline, and position information of the dial wheel, pointer, and scale. The three-layer light field propagation law comes from the optical operation rules built into the specific light field model. The combination of these two constitutes a dual constraint to protect the water meter's metering characteristics during the calculation process. During execution, the main computing unit, based on the three-layer optical field propagation model, matches and predefines the optical characteristic parameters of the corresponding optical fields for the three types of optical field action areas that have been divided. This serves as the basis for identifying and separating interfering optical fields. Simultaneously, the device retrieves the pre-stored dial topology prior rules and the three-layer optical field propagation laws, activates dual constraints, and ensures their continuous effectiveness throughout the entire optical field decoupling process. This step takes over the results of the optical field area division and configures parameters and control rules for parallel separation operations. Its technical advantage lies in using a water meter-specific optical field model to define optical parameters, which differs from the fixed parameters of general image algorithms. This allows for precise adaptation to the complex optical environment of water meters. Furthermore, the pre-loading and continuous effectiveness of dual constraints ensures real-time protection of core metering features during subsequent operations, effectively preventing interference separation operations from damaging valid signals.

[0054] Next, directional light field component separation operations are performed in parallel within the three-layer physical space, simultaneously extracting and separating the specular reflection light field, cavity scattering light field, and occlusion attenuation light field—the core decoupling operation. Parallel execution and synchronous decoupling are the core features that distinguish this solution from traditional technologies. Traditional solutions generally employ a serial, class-by-class interference cancellation approach, processing reflections, water mist, and stains sequentially. In contrast, this solution simultaneously performs separation operations on the three types of interfering light fields within the three-layer physical space, completing multi-interference decoupling in one go. Based on the preset light field action area and optical characteristic parameters, the main computing unit initiates directional light field component separation operations in parallel within the three-dimensional space of the outer glass of the water surface, the internal sealed water vapor cavity, and the dial base. According to the optical characteristics of each type of light field, it simultaneously extracts and separates the three types of interfering light fields from the mixed optical signal one by one. From the perspective of step association, this step is the core operation subject of this claim, and the work is carried out based on all the results of the preceding partitioning, parameter definition and constraint loading. In terms of technical effect, the parallel synchronous decoupling mode breaks the limitations of serial processing, which not only greatly improves the processing efficiency in multi-interference scenarios, but also avoids the problem of secondary interference and error accumulation caused by the preceding interference processing to the subsequent signal in the serial operation, especially suitable for the complex working conditions of multi-interference coupling on site.

[0055] Throughout the entire process of optical field component separation, dual constraints continue to play a role, locking the boundaries of the metering feature optical fields of the dial, pointer, and scale in real time to prevent the effective metering features from being destroyed. This step is carried out synchronously with the parallel separation operation and is a real-time protection mechanism in the operation process. The technical effect is to protect the boundaries of the effective metering features while stripping away interference, ensuring that the decoupled signal completely retains the core optical information required for water meter reading identification and that features are not lost due to interference with the separation operation.

[0056] The final step is to integrate the computation results from the three physical spaces, stripping away all interfering light field components and outputting the effective metering light field signal for the three physical structures. The effective metering light field signal refers to the pure metering optical signal corresponding only to the dial, pointer, and scale after removing the three types of interference components: specular reflection, cavity scattering, and occlusion attenuation. This is also the final output data of this step. The main computation unit summarizes the light field separation results from each of the three physical spaces, uniformly removes all interfering light field components, integrates the purified effective metering light field signal, and outputs it. This output effective metering light field signal will directly serve as the input for the next cross-spatial-layer dual-constraint feature fusion step, achieving seamless data flow integration for the entire method. The technical effect is to completely remove all coupling interference, obtaining a pure, interference-free metering light field signal, providing high-quality basic data for subsequent feature fusion, defect repair, and other processes, fundamentally reducing the negative impact of interference on subsequent water meter reading identification.

[0057] As one implementation method, the dual constraints formed by the dial topology prior rules and the three-layer light field propagation law, in order to lock the metering feature light field boundaries corresponding to the dial wheel, hands, and scale in real time during the entire process of light field component decoupling, include: Retrieve the pre-stored prior rules of dial topology and the propagation law of the three-layer light field, and combine them with the three-layer physical structure of the water meter corresponding to the second water meter image to build a dual constraint control system that links dial topology and optical field properties. Based on the preset parameters of the character wheel arrangement, pointer shape, scale distribution and contour range in the dial topology prior rules, the theoretical geometric boundaries corresponding to the character wheel, pointer and scale are calibrated in the three-layer physical space; Based on the propagation law of the three-layer light field, and combined with the light field propagation path and energy distribution characteristics of the outer glass, the internal sealed water vapor cavity, and the dial base, the aforementioned theoretical geometric boundary is transformed into a metrological characteristic light field boundary threshold that can be used for calculation and judgment. Throughout the process of separating directional light field components and synchronously decoupling three types of interference independent light field components, real-time light field operation data in the three-layer physical space is collected, and the real-time light field position is continuously compared with the preset metrological characteristic light field boundary threshold. If interfering light fields such as mirror reflection light field, cavity scattering light field, and occlusion attenuation light field are detected to intrude into the boundary of the metrological characteristic light field, the calculation range of light field separation is immediately corrected through dual constraints to achieve forced isolation between the interfering light field and the effective metrological light field. Throughout the process, the boundaries of the metrological characteristic light fields corresponding to the dial, pointer, and scale remain stable until the decoupling calculation of the light field components is completed, ultimately outputting an effective metrological light field signal with an intact three-layer physical structure. In the above steps, the pre-stored dial topology prior rules and three-layer light field propagation laws are retrieved and combined with the three-layer physical structure of the water meter corresponding to the second water meter image to build a dual-constraint control system that links dial topology and optical field characteristics. First, the core terms are explained: the dial topology prior rules are a set of static rules pre-calibrated and stored based on the factory structure of mechanical water meters and industry-standard specifications, which fully record the inherent geometric information such as the arrangement of the water meter's dials, pointer shape, scale distribution, and the outline of the metering area; the three-layer light field propagation laws originate from the constructed three-layer light field propagation model, which quantifies the propagation path, energy change, and optical linkage of light in the three-layer physical structure of the water meter's outer glass, internal sealed water vapor cavity, and dial base; the dual-constraint control system is an integrated real-time control framework that combines geometric topology rules and optical propagation rules. In terms of data acquisition and execution logic, the main computing unit first retrieves the long-term stored dial topology prior rules and three-layer light field propagation laws from the device's non-volatile memory. Simultaneously, it reads the output second water meter image from the high-speed data cache and combines this image with the corresponding three-layer physical structure of the water meter to complete the fusion and linkage of the two types of rules, formally establishing a dual-constraint control system. From a step-by-step perspective, this step is the starting point of the entire boundary locking process. The retrieved rules, images, and physical structures are all existing output results. The established control system is also the foundation for all subsequent boundary calibration, threshold conversion, and real-time monitoring work. From a technical effect perspective, by integrating geometric and optical core rules to establish a unified control system, the limitations of single-constraint control are avoided. This provides a complete framework for the full-process protection of metering feature boundaries, ensuring the uniformity and effectiveness of subsequent constraint actions.

[0058] The second step involves calibrating the theoretical geometric boundaries of the dial wheel, pointer shape, scale distribution, and contour range parameters within the three-layer physical space, based on the pre-defined parameters of the dial topology prior rules. These theoretical geometric boundaries, defined by the inherent structural parameters of the water meter, represent the standard contour range of the metering features (digital wheels, pointers, scales, etc.) within the three-layer physical space, and also serve as the fundamental geometric boundary distinguishing the metering area from the interference area. During execution, the main computing unit uses the dual-constraint control system established in the previous step as a carrier, calling upon the structural parameters within the dial topology prior rules and combining them with the actual scale of the three-layer physical space of the water meter to calibrate the theoretical geometric boundaries of all metering features one by one within the three-dimensional space. This step builds upon the achievements of the dual-constraint system, transforming abstract topological rules into concrete spatial geometric ranges. Its technical effect lies in accurately delineating the core metering area of ​​the water meter from a physical morphological perspective, clarifying the target area requiring key protection, providing a geometric benchmark for subsequent optical threshold conversion, and preventing subsequent optical field calculations from blurring the boundary between the metering area and the interference area.

[0059] The third step involves transforming the aforementioned theoretical geometric boundaries into metrological characteristic light field boundary thresholds that can be used for computational judgment, based on the propagation laws of the three layers of light fields and the propagation paths and energy distribution characteristics of the outer glass, the internal sealed water vapor cavity, and the dial substrate. In this step, the light field boundary thresholds are quantified computational parameters derived from the static geometric boundaries combined with optical characteristics. These thresholds are the core criterion for the device's computational unit to identify whether interfering light fields have intruded into the metering area. During execution, the main computational unit retrieves the propagation laws of the three layers of light fields, matches the propagation paths and energy distribution characteristics of the light fields in each layer of the water meter's physical structure, and converts the pure geometric boundaries obtained in the previous step into light field boundary thresholds with amplitude and distribution range that can be recognized by the program. This step connects to the geometric boundary calibration stage, realizing the conversion of geometric constraints into optical computational parameters. The technical effect lies in adapting to the underlying logic of the entire light field computation, transforming the originally static geometric contours into dynamically identifiable judgment standards, and enabling the dual constraints to be deeply integrated into the computational process of light field decoupling, rather than simply being external rule restrictions.

[0060] The fourth step involves real-time acquisition of optical field operation data within the three physical spaces during the entire process of directional optical field component separation and synchronous decoupling of the three types of independent interfering optical fields. This data is continuously compared with the real-time optical field position and the preset metrological characteristic optical field boundary thresholds. The real-time optical field operation data consists of dynamic data such as the optical field distribution, energy, and position generated in real-time within the three physical spaces during the parallel optical field decoupling process. This step is performed synchronously with the core decoupling operation. The main computing unit continuously acquires global optical field data in each operation cycle of optical field separation and compares the measured data point-by-point with the optical field boundary thresholds generated in the previous step. In terms of step correlation, this step is a dynamic monitoring stage. It operates based on the constraint system, geometric boundaries, and optical field thresholds generated in the first three steps, while simultaneously monitoring the optical field decoupling process in real-time. The technical effect is to achieve all-weather real-time monitoring of the movement of interfering optical fields, enabling the immediate detection of the trend of interfering optical fields approaching the metrological area, providing early warning for subsequent isolation actions, and preventing problems before they occur.

[0061] The fifth step involves immediately correcting the operational range of the optical field separation through dual constraints if interfering optical fields such as mirror reflection, cavity scattering, or occlusion attenuation are detected intruding into the boundary of the metrological characteristic optical field. This achieves forced isolation between the interfering optical field and the effective metrological optical field. When the real-time comparison results determine that the interfering optical field exceeds the preset threshold and intrudes into the metrological area, the dual constraint control system immediately triggers a dynamic adjustment mechanism to actively correct the operational area of ​​the optical field component separation, physically isolating the interfering optical field from the effective metrological optical field. This step is the core execution action of the dual constraints, taking into account the results of real-time monitoring. Its technical effect lies in actively intervening in the optical field decoupling operation, blocking the erosion of the interfering optical field on the effective features such as the dial, pointer, and scale from the operational level, and completely preventing the interference separation operation from damaging the core metrological optical field.

[0062] The sixth step involves maintaining the stable boundaries of the metering feature light field corresponding to the dial, pointer, and scale throughout the entire process, until the light field component decoupling calculation is completed. The final output is an effective metering light field signal representing the intact three-layer physical structure. Before the entire light field decoupling process ends, dual constraints continuously lock the light field boundaries of the metering features, ensuring that the boundary range does not shift or deform. After decoupling, a complete and undamaged effective metering light field signal representing the three-layer physical structure is output. By finalizing the constraint effect throughout the entire process, the technical effect ensures that the metering feature boundaries remain stable throughout the entire light field decoupling cycle, and the final output effective light field signal completely retains the water meter's metering information, providing high-quality basic data for subsequent processes.

[0063] As one implementation method, cross-spatial-layer dual-constraint feature fusion processing uses the effective metering optical field signal of the three-layer physical structure as the fusion object, specifically compensating for the optical field energy loss caused by reflection and the image feature distortion caused by water mist, so that the fused metering optical field signal completely retains the original metering information of the water meter; the specific steps of cross-spatial-layer dual-constraint feature fusion processing include: The effective metering optical field signal of the three-layer physical structure is received as input. At the same time, the dial topology prior rule and the propagation law of the three-layer optical field are loaded again to form a dual constraint. Based on the three-layer physical structure of the water meter, the outer glass, the inner sealed water vapor cavity, and the dial base, the effective metering optical field signal of the three-layer physical structure is split into independent optical field subsets corresponding to each physical layer. Under dual constraints and control, state detection is performed on each light field subset to identify the light field energy loss area caused by reflection in the outer glass layer, the characteristic distortion area caused by water mist in the inner sealed water vapor cavity layer, and the normal metering light field area of ​​the dial base layer. Based on the propagation law of the three-layer light field, directional energy gain compensation is performed on the light field region with energy loss in the outer glass layer, and optical morphology correction is performed on the light field region with characteristic distortion in the inner sealed water vapor cavity layer. During the compensation and correction process, the dial topology prior rules are continuously referenced to ensure that the geometric shape of the dial wheel, pointer and scale corresponding to each layer of light field does not change. Based on the global coordinate system defined by the dial topology prior rules, cross-spatial layer coordinate registration is performed on the three independent light field subsets after compensation and correction to eliminate positional and temporal deviations between light field signals of different physical layers. With dual constraints throughout the process, the three-layer optical field subsets that have been registered are spliced ​​and fused across the entire domain to form a complete metrological optical field signal. The integrity of the fused overall metering optical field signal is verified. After confirming that there is no missing optical field energy or metering features, a third water meter image containing the fused metering optical field signal is output.

[0064] The above steps involve receiving the effective metering optical field signal of the three-layer physical structure as input, while simultaneously loading the dial topology prior rules and the three-layer optical field propagation law to form a dual constraint, and splitting the overall signal into independent optical field subsets according to the three-layer physical structure of the water meter. First, the core words and terms are explained. The effective metering optical field signal of the three-layer physical structure is the core data output after completing the decoupling of the interference optical field, and it is also the only pre-input in this embodiment. This signal has been stripped of three types of interference components: the specular reflection optical field, the cavity scattering optical field, and the occlusion attenuation optical field, retaining only the effective optical signals corresponding to the water meter dial, pointer, and scale. The dual constraint is composed of the dial topology prior rules and the three-layer optical field propagation law. The dial topology prior rules record the inherent layout and contour parameters of the water meter's metering features, while the three-layer optical field propagation law originates from the dedicated optical field model built in claim 2. Both types of rules are pre-stored in the device rule library and can be called at any time. The independent optical field subset is the layered data obtained after splitting the overall optical field signal according to the three-layer physical structure of the outer glass, the internal sealed water vapor cavity, and the dial base. From the perspective of data acquisition and execution logic, the main computing unit first reads the effective metering optical field signal of the three-layer physical structure from the high-speed data cache. Then, it retrieves two types of pre-stored rules to complete the loading of dual constraints. Next, combined with the division criteria of the three-layer physical structure of the water meter entity, it decomposes the integrated effective optical field signal into optical field subsets corresponding to each physical layer. From the perspective of step correlation, this step is a preliminary preparation for cross-spatial feature fusion. The computational results directly connected to the upstream, and the decomposed optical field subsets, are also the basic data for subsequent defect detection. From a technical perspective, combining the optical field signal with the actual physical structure of the water meter breaks the limitations of traditional globally unified processing, allowing the processing method of each optical field layer to match the optical characteristics of the corresponding physical layer, laying a physical foundation for subsequent layered targeted defect repair.

[0065] The second execution step involves conducting state detection on each subset of the light field under dual constraint control, identifying and marking different types of defect areas. Light field energy loss specifically refers to the optical defect where the overall energy of the light field decreases due to strong light reflection on the outer glass of the water meter. Feature distortion refers to the problem of light field shape distortion and contour distortion caused by water mist within the internal sealed water vapor cavity. The normal measurement light field area represents the standard light field range on the dial substrate where no defects have appeared. Logically, under the full control of dual constraints, the main computing unit scans and detects each of the three subsets of the split light field. For the light field subset corresponding to the outer glass, it identifies energy loss areas; for the light field subset corresponding to the internal sealed water vapor cavity, it identifies feature distortion areas; and simultaneously marks the normal area of ​​the dial substrate layer, completely recording the location and type of all defects. This step uses the light field subset obtained in the previous step as input to accurately locate defects, providing a direct basis for subsequent compensation and correction operations. Technically, layered detection can accurately distinguish the differentiated defects of different physical layers, avoiding mutual interference between different defects and ensuring the targeted nature of subsequent repair operations.

[0066] The third execution step involves performing directional energy gain and optical morphology correction based on the propagation laws of the three-layer light field, while simultaneously ensuring the geometric shape of the metering features remains unchanged by referring to the dial topology prior rules. Directional energy gain replenishes the light field energy lost due to reflection, optical morphology correction repairs light field distortion caused by water vapor, and the dial topology prior rules act as geometric constraints in this step, preventing changes in the contours of the dial wheel, pointers, and scales during the repair process. The processing unit targets the defect area marked in the second step, performs energy gain processing on the outer glass loss area according to the propagation laws of the three-layer light field, and performs morphology correction on the water vapor cavity distortion area. Throughout the process, the dial topology prior rules are continuously referenced to lock the shape of the metering features. This step, based on the defect detection results, performs directional repair, connecting with the defect marking results of the previous step. Its technical effect lies in specifically addressing two typical optical defects caused by reflection and water vapor, while simultaneously relying on dual constraints to maintain the geometric boundaries of the metering features, achieving the dual goals of "repairing defects + protecting features," without damaging the original metering information of the water meter due to light field correction.

[0067] The fourth execution step uses the global coordinate system defined by the dial topology prior rules as a benchmark to perform cross-spatial layer coordinate registration on the three independent light field subsets that have undergone compensation and correction, eliminating positional and temporal deviations. In technical terms, the global coordinate system is a unified spatial coordinate system established based on the dial topology prior rules, serving as the benchmark for multi-layer light field registration. Positional and temporal deviations are spatial misalignments and signal asynchrony issues caused by different propagation paths of the light field signals in the three physical spaces. During execution, the computation unit uses the global coordinates as a unified standard to align the coordinates of the repaired three-layer light field subsets, correcting positional and temporal deviations between layers. This step, following the layered repair of the light field subsets, is a necessary pre-process for multi-layer light field fusion. Its technical effect is to unify the spatial position and signal timing of the three light fields, avoiding image misalignment and signal discontinuity problems that occur after direct stitching, and ensuring spatial consistency after multi-layer signal fusion.

[0068] The fifth execution step involves globally stitching and fusing the registered light field subsets under the full support of dual constraints, forming a unified and complete metering light field signal. Global stitching and fusion refers to integrating the three independent light field subsets that have completed coordinate registration into a single, continuous overall light field. Under the continuous control of dual constraints, the computing unit completes global stitching according to the distribution logic of the physical layers, integrating the layered signals into a unified light field. This step uses the registered light field subsets as input to integrate multiple layers of signals. The technical effect is to reintegrate the effective light fields dispersed in the three physical spaces into a complete dial optical signal, restoring the overall imaging characteristics of the water meter dial.

[0069] The sixth execution step is to perform an integrity check on the fused overall metering light field signal. After confirming that there are no missing parts, a third water meter image containing the feature-fused metering light field signal is output. The integrity check is a comprehensive verification of the energy and metering features of the fused light field to determine whether there are any signal missing or feature incompleteness issues. The third water meter image is the final output of this embodiment and also the designated input for the next three-dimensional spatial light field-driven global semantic reconstruction step. After the computing unit completes the verification work and confirms that the metering features such as light field energy, dial, pointer, and scale are complete, it generates and caches the third water meter image based on the fused light field signal. This process ensures the quality of the fusion results from the previous step and transmits data to subsequent steps. The technical effect is to control the overall quality of feature fusion and ensure that the output third water meter image and corresponding light field signal fully meet the operational requirements of the subsequent stain and damage repair step.

[0070] As one implementation method, three-dimensional spatial light field-driven global semantic reconstruction abandons the traditional two-dimensional local pixel interpolation image restoration method. Based on the associated light field distribution and topological information of three-layer physical space, it performs global reconstruction on the damaged areas of the character wheel and pointer caused by stains, completely restoring the outline of the character wheel and the lines of the pointer. The specific steps of three-dimensional spatial light field-driven global semantic reconstruction include: Based on the third water meter image and the corresponding feature fusion metering light field signal, the associated light field distribution data and dial topology information of the three physical spaces of the water meter are loaded simultaneously as the basic data source for semantic reconstruction. By combining the dial topology information and the light field distribution characteristics of the three-layer physical structure, the third water meter image is scanned across the entire area to accurately identify and mark the incomplete areas of the wheel outline and the broken areas of the pointer lines caused by dirt obscuring them, and to determine the location, range and type of each defect area. Centered on the defective area, the global correlation light field features of the three physical spaces of the water meter's outer glass, internal sealed water vapor cavity, and dial base are extracted, including light field direction, energy gradient, and inter-layer correlation rules. Combined with the geometric constraints of the dial topology information, the optical constraints of the three-layer light field propagation rules, and the extracted global correlation light field features, a three-dimensional semantic reconstruction benchmark model adapted to the water meter's metering features is constructed. Based on the three-dimensional semantic reconstruction benchmark model, light field-driven semantic completion is performed in different regions. For the missing area of ​​the character wheel, the complete character outline is restored according to the topological contour and the direction of the three-dimensional light field. For the broken area of ​​the pointer, the pointer lines are completed according to the inherent shape of the pointer and the continuity of the light field between layers. The completed local areas are subjected to cross-three-layer physical space feature joint adjustment to ensure that the light field morphology of the reconstructed word wheel and pointer is continuous and the geometric morphology is unified across different physical layers. All reconstructed areas are integrated with the original normal measurement light field areas to form a complete overall light field, and finally the semantically reconstructed light field data is output.

[0071] The above steps, based on the third water meter image and the corresponding feature-fused metering optical field signal, simultaneously load the associated optical field distribution data and dial topology information of the three physical spaces of the water meter as the basic data source for semantic reconstruction. From a lexical and terminological perspective, the third water meter image is an intermediate image generated after completing cross-spatial layer dual-constraint feature fusion. The image's optical field energy and basic morphology have already compensated for defects caused by reflection and water mist, making it the core image input for this step. The feature-fused metering optical field signal is an integrated optical signal accompanying the third water meter image, recording the complete optical field information after the integration of the three physical spaces. The associated optical field distribution data of the three physical spaces is a set of optical parameters pre-stored in the device's non-volatile memory, describing the propagation and interaction of light between the three layers of the water meter structure. The dial topology information is also a set of static rules pre-stored in the device, recording the inherent arrangement, standard shape, and contour range of the mechanical water meter's dial, pointer, and scale, serving as the geometric basis for determining whether metering features are missing. From the perspective of execution logic and data sources, the main computing unit first reads the third water meter image and feature-fused metering light field signal from the high-speed data cache, and then retrieves the pre-stored associated light field distribution data and dial topology information in the memory, integrating the four types of data into the basic data source for semantic reconstruction. In terms of step correlation, this step is the initial preparatory work for the entire 3D semantic reconstruction process. The upstream computation results are entirely based on this, and all the integrated data sources are the sole basis for subsequent scanning, modeling, and repair work. From a technical effect analysis perspective, this step integrates data from three core dimensions: image, light field, and geometric topology. It breaks through the limitations of traditional repair relying solely on local pixel information, allowing subsequent reconstruction operations to simultaneously consider the water meter's physical structure, optical propagation characteristics, and inherent shape features, ensuring the comprehensiveness and rationality of the repair work from the data source level.

[0072] The second execution step involves combining the dial topology information with the light field distribution characteristics of the three-layer physical structure to perform a full-area scan of the third water meter image. This accurately identifies and marks areas with incomplete digit wheel outlines and broken pointer lines caused by dirt obscuring the dial, determining the location, extent, and type of each defect. The full-area scan refers to the processing unit performing point-by-point and region-by-region traversal detection on all pixels and corresponding light field areas of the third water meter image. Metering feature defects specifically refer to two typical defects: incomplete digit wheel outlines and broken pointer lines caused by scale, oil, or other dirt obscuring or eroding the dial. Logically, the processing unit, based on the data source integrated in the first step, compares the actual image features with the standard shape and normal light field distribution of the dial topology to quickly distinguish between complete and defective areas, and marks and records the location, coverage, and type of the defects. This step uses the complete data source from the previous step as input to accurately locate the target area for repair. Its technical effect is to achieve precise positioning of the defective area, avoid performing invalid calculations on the normal measurement area, and make the subsequent semantic reconstruction have clear targeting. At the same time, it distinguishes different defect types, laying the foundation for differentiated repair by region.

[0073] The third execution step involves extracting the global correlated light field features of the three physical spaces of the water meter—the outer glass, the internal sealed water vapor cavity, and the dial base—centered on the damaged area. This, combined with the geometric constraints of the dial topology and the optical constraints of the three-layer light field propagation laws, constructs a three-dimensional semantic reconstruction benchmark model adapted to the water meter's metering characteristics. This step is the core difference between this solution and traditional repair techniques. The global correlated light field features include light field direction, energy gradient, and interlayer correlation laws. Unlike traditional two-dimensional local pixel interpolation, which only extracts pixels in a small area around the damaged area, this solution extracts the global light field parameters of the three physical spaces of the water meter. Geometric constraints define the standard shape of the dial wheel and pointer based on the dial topology information, while optical constraints define the light field operation rules based on the three-layer light field propagation laws established according to claim 2. The three-dimensional semantic reconstruction benchmark model is an integrated computational model that combines geometric shape and optical characteristics. Logically, the computational unit collects the global light field features of the three-layer structure centered on the damaged area marked in the second step, and then combines these with dual constraints to build a dedicated reconstruction model. From a step-by-step perspective, this stage, which completes modeling based on information about the damaged area, serves as the core bridge connecting damage identification and feature repair. In terms of technical effectiveness, 3D global modeling, combined with geometric and optical constraints, completely eliminates the limitations of 2D local interpolation. It can highly reproduce the metering characteristics of the water meter in its true 3D space, effectively avoiding problems such as contour distortion, line misalignment, and local distortion that are prone to occur in traditional repair methods.

[0074] The fourth execution step involves performing light field-driven semantic completion based on the 3D semantic reconstruction benchmark model, divided into regions. For the missing areas of the digit wheel, the complete character outline is restored according to the topological contour and the direction of the 3D light field. For the broken areas of the pointer, the pointer lines are completed based on the inherent shape of the pointer and the continuity of the interlayer light field. Light field-driven semantic completion refers to using 3D light field data as the core computational power to complete feature restoration according to the rules of the benchmark model. Logically, the computation unit performs repair operations for the two different measurement features—digit wheel and pointer—referring to the topological contour, light field direction, and interlayer continuity within the model. This step operates based on the benchmark model constructed in the previous step and is the core repair action of semantic reconstruction. The technical effect is to directionally restore the missing features according to the 3D light field and topological rules, ensuring that the shape and lines of the digit wheel and pointer are consistent with the original factory standard of the water meter, achieving precise repair of stained and damaged areas.

[0075] The fifth execution step involves cross-layer physical space feature integration and calibration of the completed local areas. This ensures the continuity of the optical morphology and the uniformity of the geometry of the reconstructed dial and pointer across different physical layers. It also integrates all reconstructed areas with the original normal metering optical field areas to form a complete overall optical field, ultimately outputting semantically reconstructed optical field data. Cross-layer physical space feature integration and calibration is an adaptation operation for the multi-layered structure of water meters, eliminating subtle deviations in the optical field and shape between layers. Logically, the computation unit performs cross-layer calibration on the repaired local areas, then integrates the repaired areas with the original normal areas, ultimately generating complete optical field data and outputting it. The output semantically reconstructed optical field data will serve as input data for subsequent steps. The technical effect is to unify the characteristic morphology of the three physical layers of the water meter, avoiding inter-layer discontinuities and inconsistent morphologies after layered repair, ultimately generating optical field data with complete signals and unified features, providing high-quality basic data for subsequent residual verification.

[0076] As one implementation method, after correcting the optical field residual through a residual optical field reverse closed-loop verification mechanism, a standardized dial image with accurate physical location, pure optical field signal, and complete metering characteristics is generated. This standardized dial image is then connected to the subsequent water meter mechanical transmission rule verification process. The final steps for determining the water meter's final reading based on the verification results include: Based on the semantically reconstructed light field data and the third water meter image, the pre-stored standard light field model of the water meter is retrieved, and the dual constraints composed of the dial topology prior rules and the three-layer light field propagation law are loaded again as the benchmark for closed-loop verification and residual correction. Based on the three-layer physical structure of the water meter, consisting of the outer glass, the internal sealed water vapor cavity, and the dial base, the optical field data that has completed semantic reconstruction is decomposed into layers. The optical field data of each layer is compared with the standard optical field data of the corresponding layer in the standard optical field model of the water meter pixel by pixel and metering feature by feature. The global optical field residuals in terms of optical field amplitude, shape, and position are calculated. By combining the residual distribution location and feature deviation type, the residuals of the global optical field are classified and traced back to their source, distinguishing the hierarchical operation residuals, feature fusion residuals, and semantic reconstruction residuals, and accurately locating the physical layer and metrological region where each type of residual is located. Based on the residual type and its location, directional residual correction is performed under dual constraints: for optical residuals generated by hierarchical operations and feature fusion, the amplitude and propagation shape of the light field are adjusted according to the propagation law of the three-layer light field; for geometric residuals generated by semantic reconstruction, the outline and position of the dial wheel and pointer are corrected according to the dial topology prior rules. During the correction process, the geometric shape and optical properties of the measurement features are always kept within the standard range. After the initial correction, the overall optical field data is compared with the standard optical field model of the water meter again to perform a closed-loop secondary verification to determine whether the current remaining optical field residual is lower than the preset residual allowable threshold. If the residual exceeds the threshold, the third and fourth steps are repeated for iterative correction until the residual meets the threshold requirement. Once the secondary verification confirms that the optical field residual meets the standard, a standardized dial image with accurate physical location, pure optical field signal, and complete metering characteristics is generated based on the corrected overall optical field data. The standardized dial image is then connected to the mechanical transmission rule verification stage of the water meter, and the final verification is completed in conjunction with the mechanical transmission rules to determine the final reading of the mechanical water meter.

[0077] The above steps retrieve the benchmark, which is based on the semantically reconstructed light field data and the third water meter image. The standard light field model of the water meter is retrieved, and simultaneously, the dial topology prior rules and the three-layer light field propagation law form a dual constraint, establishing a closed-loop verification benchmark system. From a lexical and terminological perspective, the semantically reconstructed light field data and the third water meter image are the output results after completing the 3D semantic reconstruction and are also the core input data for this step. The third water meter image carries the fused metering light field signal, while the light field data records the complete optical features in three-dimensional space. The standard light field model of the water meter is the benchmark model for offline equipment calibration, corresponding to the standard light field distribution under ideal working conditions of no reflection, no water mist, and no stains on the water meter, and is the core reference for judging deviations. The dual constraint is a control system used throughout the entire process, consisting of the dial topology prior rules and the three-layer light field propagation law. The former constrains the geometry of the dial wheel, pointer, and scale, while the latter constrains the optical propagation characteristics of the light field. In terms of data acquisition and execution logic, the main computing unit first reads the light field data and the third water meter image from the high-speed data cache according to claim 7, and then retrieves the pre-stored standard light field model of the water meter, the prior rules of the dial topology, and the propagation law of the three-layer light field from the non-volatile memory. The above content is integrated as a unified benchmark for this residual verification and correction. From the perspective of step correlation, this step is the initial preparation work for closed-loop verification, which is connected to the three-dimensional semantic reconstruction step upstream. The benchmark data retrieved is also the core rules that are repeatedly used to ensure the uniformity of the calculation standard throughout the process. In terms of technical effect, the simultaneous introduction of the standard light field model and dual constraints establishes a verification benchmark from two dimensions: optical signal and geometric shape. Compared with a single reference standard, it can comprehensively identify different types of calculation deviations, laying a complete evaluation foundation for subsequent residual detection.

[0078] The second execution step involves layer-by-layer decomposition of the light field and global comparison to calculate the global light field residual. Layer-by-layer decomposition involves dividing the overall light field data into three independent layers based on the three physical structures of the water meter: the outer glass layer, the internal sealed water vapor cavity, and the dial base, thus conforming to the actual physical structure of the water meter. The global light field residual refers to the deviation between the measured light field and the standard light field in three dimensions: amplitude, shape, and position. It represents the small errors accumulated from the layered calculations, feature fusion, and semantic reconstruction processes. Logically, the computation unit performs layered processing on the semantically reconstructed light field data according to the three-layer physical structure of the water meter. Then, it performs a pixel-by-pixel, metering feature-by-meter global comparison of the corresponding layer data of each layer of the measured light field with the corresponding layer data of the standard light field model of the water meter, quantifying and calculating the global light field residual. This step builds upon the verification benchmark established in the previous step and is the core of residual identification. By comparing the three-layer structure of the water meter layer by layer, it avoids the problem of ignoring the differences between layers in the global comparison. Its technical effect is that it can accurately capture the tiny deviations of each physical layer and each metering feature, realize the full-domain and layered detection of residuals, and ensure that all calculation errors can be identified without any missed detections.

[0079] The third execution step involves tracing and classifying the residuals across the entire optical field and locating their distribution areas. In this step, the residuals from the hierarchical computation originate from the optical field component decoupling process, the feature fusion residuals arise from the cross-spatial feature fusion stage, and the semantic reconstruction residuals are deviations formed during the three-dimensional semantic reconstruction stage. These three types of residuals correspond to different computational nodes within the entire method. Logically, the computational unit, combining the distribution location and deviation morphology of the residuals, traces their origin, classifies the entire residuals into the aforementioned three categories, and accurately locates the physical layer and water meter measurement area where the residuals are located. From a step-by-step correlation analysis, this step, based on the residual data calculated in the second step, is a prerequisite for targeted correction. Technically, tracing and classifying the residuals allows for differentiation of error types generated at different stages, abandoning the traditional, uniform correction approach, and providing a clear basis for subsequent differentiated and targeted correction, avoiding the introduction of new deviations through blind correction.

[0080] The fourth execution step involves targeted residual correction based on the residual type under dual constraint control. Terminologically, optical residuals include deviations in light field amplitude and propagation morphology, which are adjusted according to the three-layer light field propagation law; geometric residuals are deviations in the position of the dial wheel, pointer, and scale, corrected using prior rules of dial topology. Logically, the computation unit, based on the classification results of the third step, adjusts the light field parameters according to the three-layer light field propagation law for optical residuals, and corrects the feature contours for geometric residuals by referring to prior rules of dial topology. The entire correction process is consistently under dual constraint control, ensuring that the water meter's measurement characteristics do not deviate from the standard range. This step is the core action of residual correction, directly corresponding to the residual classification results. Its technical effect is to achieve targeted processing where "one type of residual corresponds to one correction method," balancing the integrity of optical signals and geometric shapes, eliminating errors while protecting core measurement features such as the dial wheel, pointer, and scale from secondary damage.

[0081] The fifth execution step is reverse closed-loop secondary verification and iterative correction. The residual tolerance threshold is a preset error acceptance standard, representing the maximum acceptable deviation range for water meter reading recognition; closed-loop iteration is the core feature that distinguishes this scheme from traditional open-loop image processing. Logically, after correction, the computing unit compares the light field data with the standard light field model again to determine whether the remaining residual is lower than the preset threshold. If it does not meet the standard, the residual classification and directional correction steps are executed repeatedly until the residual meets the accuracy requirements. This step forms a closed-loop iterative mechanism to verify the initial correction result. The technical effect is that by gradually compressing the error through multiple iterations, the light field residual is controlled within the acceptable range, further improving data accuracy and compensating for the limitations of single correction.

[0082] The sixth execution step is to generate a standardized dial image, integrate it with mechanical transmission rule verification, and determine the final water meter reading. The standardized dial image is the final image after residual correction meets the standards, ensuring that the optical field signal and geometric features conform to the standards. The water meter mechanical transmission rule verification is the final verification step of the entire method, combining the mechanical structural characteristics of the mechanical water meter for final confirmation. Logically, after the residual meets the standards, the computing unit generates a standardized dial image based on the corrected optical field data, integrates the image with the subsequent mechanical transmission verification step, and determines and outputs the water meter reading by combining all verification results. Upstream, it inherits the results of iterative correction. The technical effect is to complete the entire process from optical field correction to final reading output. The reading, after multi-layered verification, can adapt to the dual mechanical and optical characteristics of the water meter, ensuring the authenticity and reliability of the reading under harsh operating conditions.

[0083] To achieve the above objectives, this application adopts the following technical solution: A mechanical water meter data acquisition device for performing a mechanical water meter data acquisition method is characterized by comprising: a data acquisition part, a data processing part, and a data generation part.

[0084] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for acquiring data from a mechanical water meter, characterized in that, The method includes: The original water meter reading image is obtained as the first water meter image. Based on the three-layer physical structure of the water meter's outer glass, internal sealed water vapor cavity, and dial base, a three-layer light field propagation model adapted to the optical characteristics of the water meter cavity is constructed. Based on the first water meter image and the three-layer light field propagation model, the pixels of the first water meter image are bidirectionally traced and bound to the three-layer physical spatial position of the water meter to output the second water meter image. Any pixel of the second water meter image is accurately mapped to and bound to the three-layer physical spatial position of the water meter. Based on the three-layer optical field propagation model, directional optical field component separation calculation is performed on the optical signal superimposed in the second water meter image. According to the optical characteristics of the specular reflection light field corresponding to reflection, the cavity scattering light field corresponding to water mist, and the occlusion attenuation light field corresponding to dirt, the independent optical field components corresponding to the three types of interference are synchronously decoupled within the three-layer physical structure. During the calculation of the independent optical field components, the dial topology prior rule and the three-layer optical field propagation law are loaded to form a dual constraint, so as to lock the metering characteristic optical field boundary of the water meter's dial, pointer, and scale, and output the effective metering optical field signal of the three-layer physical structure after stripping the interference optical field components. Based on the effective metering optical field signal of the three-layer physical structure, the second water meter image is subjected to cross-spatial layer dual-constraint feature fusion processing to compensate for the light field energy attenuation and feature distortion caused by reflection and water mist, so as to output a third water meter image containing the metering optical field signal of feature fusion. Based on the associated light field distribution and topological information of the third water meter image and the three-layer physical structure, a three-dimensional spatial light field-driven global semantic reconstruction is performed on the metering feature defect area caused by dirt in the water meter to restore the complete digit wheel outline and pointer lines in the water meter, and output the light field data of the completed semantic reconstruction. By combining the residual optical field reverse closed-loop verification mechanism, the optical field data, and the third water meter image, the optical field data is compared with the standard optical field model of the water meter to correct the optical field residual generated in the third water meter image during the layering operation and feature fusion process, thereby obtaining the water meter reading in the third water meter image.

2. The method for acquiring data from a mechanical water meter according to claim 1, characterized in that, The execution steps of the three-layer optical field propagation model include: Offline sample collection and layered optical parameter extraction: For the three independent physical layers of mechanical water meters—outer glass, internal sealed water vapor cavity, and dial base—interference-free standard light field samples, single-interference light field samples, and multi-coupled interference light field samples were collected respectively. Basic optical parameters such as light field propagation path, light field energy attenuation coefficient, and optical reflection / scattering response characteristics were extracted for each layer. Based on the optical parameters extracted from each layer, a mathematical model of the mirror reflection light field is constructed for the outer glass, a mathematical model of the cavity scattering light field is constructed for the internal sealed water vapor cavity, and a mathematical model of the substrate imaging light field is constructed for the dial substrate. The calculation rules for the light field of each layer as a function of spatial position and interference type are clarified. Combining the optical crosstalk characteristics of the sealed cavity of the water meter, rules for light field transmission, energy transfer, and morphological linkage between the three physical spaces are established. The three single-layer light field mathematical models are coupled and fused to form a complete three-layer light field propagation mathematical model, and the overall operation logic of the model is solidified. The three-layer optical field propagation model that has been coupled is pre-stored in the computing unit. When the mechanical water meter data acquisition method is executed, the pre-stored model is retrieved to complete online initialization and establish a mapping relationship between the model and the first water meter image and the three-layer physical space of the water meter. In each computational stage, including pixel bidirectional source binding, directional light field component separation, cross-spatial layer feature fusion, three-dimensional semantic reconstruction, and residual light field closed-loop verification, the three-layer light field propagation model is continuously invoked. Based on the optical propagation laws and inter-layer linkage rules built into the model, it participates in light field calculation, feature constraints, and deviation judgment until the entire water meter data acquisition process is completed.

3. The method for acquiring data from a mechanical water meter according to claim 2, characterized in that, The step of bidirectionally binding the pixels of the first water meter image with the three-layer physical spatial location of the water meter includes: Perform a full-domain pixel traversal on the first water meter image, and combine the optical imaging response characteristics of the three physical structures of the water meter outer glass, the internal sealed water vapor cavity, and the dial base to determine the actual target physical layer attached to each pixel. Establish a positive mapping relationship between the two-dimensional coordinates of image pixels and the three-dimensional coordinates of the three-layer physical space of the water surface, and complete the positive tracing from image pixels to physical spatial location; Based on the three-layer light field propagation model, the theoretical imaging position of the corresponding pixel is derived from the three-dimensional coordinates of the three-layer physical space of the water surface, thus completing the reverse tracing from the physical space position to the image pixel. By comparing the forward and reverse tracing results and completing position calibration, the unique binding relationship between the pixel and the three-layer physical spatial position of the water meter is locked. The bound pixel maintains a constant position and does not shift in all subsequent operation processes to generate the second water meter image.

4. The method for acquiring data from a mechanical water meter according to claim 3, characterized in that, The steps of directional optical field component separation calculation and decoupling of the three types of interfering independent optical field components include: Using the second water meter image and the three-layer light field propagation model as common inputs, and based on the three-layer physical structure of the water meter's outer glass, internal sealed water vapor cavity, and dial base, combined with the adhesion characteristics of reflection, water mist, and stains, the light field action areas corresponding to the specular reflection light field, cavity scattering light field, and occlusion attenuation light field are divided. Based on the three-layer light field propagation model, optical feature parameters corresponding to the three types of light fields are predefined, and the dial topology prior rules and the three-layer light field propagation law are applied throughout the process to form a dual constraint. In the three-layer physical space, directional light field component separation operations are performed in parallel, simultaneously extracting and separating the three types of independent interfering light field components: specular reflection light field, cavity scattering light field, and occlusion attenuation light field. During the separation process, the metering feature light field boundaries of the dial wheel, pointer, and scale are locked in real time according to the dual constraints to avoid damage to the effective metering features. Finally, the operation results of the three-layer physical space are integrated, all interfering light field components are stripped away, and the effective metering light field signal of the three-layer physical structure is output.

5. The method for acquiring data from a mechanical water meter according to claim 4, characterized in that, The dual constraints formed by the dial topology prior rule and the three-layer light field propagation law, in the process of real-time locking of the metering feature light field boundaries corresponding to the dial wheel, hands, and scale during the entire process of light field component decoupling, include: Retrieve the pre-stored prior rules of dial topology and the propagation law of the three-layer light field, and combine them with the three-layer physical structure of the water meter corresponding to the second water meter image to build a dual constraint control system that links dial topology and optical field properties. Based on the preset parameters of the character wheel arrangement, pointer shape, scale distribution and contour range in the dial topology prior rules, the theoretical geometric boundaries corresponding to the character wheel, pointer and scale are calibrated in the three-layer physical space; Based on the propagation law of the three-layer light field, and combined with the light field propagation path and energy distribution characteristics of the outer glass, the internal sealed water vapor cavity, and the dial base, the aforementioned theoretical geometric boundary is transformed into a metrological characteristic light field boundary threshold that can be used for calculation and judgment. Throughout the process of separating directional light field components and synchronously decoupling three types of interference independent light field components, real-time light field operation data in the three-layer physical space is collected, and the real-time light field position is continuously compared with the preset metrological characteristic light field boundary threshold. If interfering light fields such as mirror reflection light field, cavity scattering light field, and occlusion attenuation light field are detected to intrude into the boundary of the metrological characteristic light field, the calculation range of light field separation is immediately corrected through dual constraints to achieve forced isolation between the interfering light field and the effective metrological light field. Throughout the process, the boundary of the measurement feature light field corresponding to the dial wheel, pointer, and scale remains stable until the decoupling operation of the light field components is completed, and finally the effective measurement light field signal of the three-layer physical structure without damage is output.

6. The method for acquiring data from a mechanical water meter according to claim 5, characterized in that, The cross-spatial layer dual-constraint feature fusion processing takes the effective measurement optical field signal of the three-layer physical structure as the fusion object, and specifically compensates for the light field energy loss caused by reflection and the image feature distortion caused by water mist, so that the fused measurement optical field signal completely retains the original measurement information of the water meter. The steps of the cross-spatial-layer dual-constraint feature fusion processing include: The effective metering optical field signal of the three-layer physical structure is received as input. At the same time, the dial topology prior rule and the propagation law of the three-layer optical field are loaded again to form a dual constraint. Based on the three-layer physical structure of the water meter, the outer glass, the inner sealed water vapor cavity, and the dial base, the effective metering optical field signal of the three-layer physical structure is split into independent optical field subsets corresponding to each physical layer. Under dual constraints and control, state detection is performed on each light field subset to identify the light field energy loss area caused by reflection in the outer glass layer, the characteristic distortion area caused by water mist in the inner sealed water vapor cavity layer, and the normal metering light field area of ​​the dial base layer. Based on the propagation law of the three-layer light field, directional energy gain compensation is performed on the light field region with energy loss in the outer glass layer, and optical morphology correction is performed on the light field region with characteristic distortion in the inner sealed water vapor cavity layer. During the compensation and correction process, the dial topology prior rules are continuously referenced to ensure that the geometric shape of the dial wheel, pointer and scale corresponding to each layer of light field does not change. Based on the global coordinate system defined by the dial topology prior rules, cross-spatial layer coordinate registration is performed on the three independent light field subsets after compensation and correction to eliminate positional and temporal deviations between light field signals of different physical layers. With dual constraints throughout the process, the three-layer optical field subsets that have been registered are spliced ​​and fused across the entire domain to form a complete metrological optical field signal. The integrity of the fused overall metering optical field signal is verified. After confirming that there is no missing optical field energy or metering features, a third water meter image containing the fused metering optical field signal is output.

7. The method for acquiring data from a mechanical water meter according to claim 6, characterized in that, The three-dimensional spatial light field-driven global semantic reconstruction abandons the traditional two-dimensional local pixel interpolation image repair method. Based on the associated light field distribution and topological information of the three-layer physical space, it performs global reconstruction on the damaged areas of the character wheel and pointer caused by stains, and completely restores the outline of the character wheel and the lines of the pointer. The specific steps of the three-dimensional spatial light field-driven global semantic reconstruction include: Based on the third water meter image and the corresponding feature fusion metering light field signal, the associated light field distribution data and dial topology information of the three physical spaces of the water meter are loaded synchronously as the basic data source for semantic reconstruction. By combining the dial topology information and the light field distribution characteristics of the three-layer physical structure, the third water meter image is scanned across the entire area to accurately identify and mark the incomplete areas of the wheel outline and the broken areas of the pointer lines caused by dirt obscuring them, and to determine the location, range and type of each defect area. Centered on the defective area, the global correlation light field features of the three physical spaces of the water meter's outer glass, internal sealed water vapor cavity, and dial base are extracted, including light field direction, energy gradient, and inter-layer correlation rules. Combined with the geometric constraints of the dial topology information, the optical constraints of the three-layer light field propagation rules, and the extracted global correlation light field features, a three-dimensional semantic reconstruction benchmark model adapted to the water meter's metering features is constructed. Based on the three-dimensional semantic reconstruction benchmark model, light field-driven semantic completion is performed in different regions. For the missing area of ​​the character wheel, the complete character outline is restored according to the topological contour and the direction of the three-dimensional light field. For the broken area of ​​the pointer, the pointer lines are completed according to the inherent shape of the pointer and the continuity of the light field between layers. The completed local areas are subjected to cross-three-layer physical space feature joint adjustment to ensure that the light field morphology of the reconstructed word wheel and pointer is continuous and the geometric morphology is unified across different physical layers. All reconstructed areas are integrated with the original normal measurement light field areas to form a complete overall light field, and finally the semantically reconstructed light field data is output.

8. The method for acquiring data from a mechanical water meter according to claim 7, characterized in that, The steps of correcting the optical field residual through the residual optical field reverse closed-loop verification mechanism to generate a standardized dial image with accurate physical location, pure optical field signal, and complete metering characteristics, and then incorporating this standardized dial image into the subsequent water meter mechanical transmission rule verification process, and finally determining the final water meter reading based on the verification results, include: Based on the semantically reconstructed light field data and the third water meter image, the pre-stored standard light field model of the water meter is retrieved, and the dual constraints composed of the dial topology prior rules and the three-layer light field propagation law are loaded again as the benchmark for closed-loop verification and residual correction. Based on the three-layer physical structure of the water meter, consisting of the outer glass, the internal sealed water vapor cavity, and the dial base, the optical field data that has completed semantic reconstruction is decomposed into layers. The optical field data of each layer is compared with the standard optical field data of the corresponding layer in the standard optical field model of the water meter pixel by pixel and metering feature by feature. The global optical field residuals in terms of optical field amplitude, shape, and position are calculated. By combining the residual distribution location and feature deviation type, the residuals of the global optical field are classified and traced back to their source, distinguishing the hierarchical operation residuals, feature fusion residuals, and semantic reconstruction residuals, and accurately locating the physical layer and metrological region where each type of residual is located. Based on the residual type and its location, directional residual correction is performed under dual constraints: for optical residuals generated by hierarchical operations and feature fusion, the amplitude and propagation shape of the light field are adjusted according to the propagation law of the three-layer light field; for geometric residuals generated by semantic reconstruction, the outline and position of the dial wheel and pointer are corrected according to the dial topology prior rules. During the correction process, the geometric shape and optical properties of the measurement features are always kept within the standard range. After the initial correction, the overall optical field data is compared with the standard optical field model of the water meter again to perform a closed-loop secondary verification to determine whether the current remaining optical field residual is lower than the preset residual allowable threshold. If the residual exceeds the threshold, the third and fourth steps are repeated for iterative correction until the residual meets the threshold requirement. Once the secondary verification confirms that the optical field residual meets the standard, a standardized dial image with accurate physical location, pure optical field signal, and complete metering characteristics is generated based on the corrected overall optical field data. The standardized dial image is then connected to the water meter mechanical transmission rule verification stage. The final verification is completed in conjunction with the mechanical transmission rules to determine the final reading of the mechanical water meter.

9. A mechanical water meter data acquisition device for implementing the mechanical water meter data acquisition method according to any one of claims 1 to 8, characterized in that, include: The data acquisition section, the data processing section, and the data generation section.

10. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored on the memory, the processor being configured to run the computer program to perform the mechanical water meter data acquisition method according to any one of claims 1 to 8.

Images