A water level gauge reading dual-path visual measurement method, system, device and storage medium in an industrial scene

Abstract

The application discloses a water level gauge reading double-path visual measurement method, system and device in an industrial scene and a storage medium, relates to the technical fields of artificial intelligence, image recognition and water level detection, and comprises the following steps: collecting image samples to train a model in reliance on an industrial scene camera; determining current illumination conditions based on image average brightness, switching daytime mode and nighttime mode, and constructing a structured pixel coordinate set; and outputting a water level value by using a double-path output. The method has the advantages that the delay of manual reading is reduced, and the safety and reliability of work production are improved through automatic reading and real-time monitoring. The double-path recognition mechanism is suitable for various illumination conditions, and the robustness and accuracy are improved. The small-scale digital continuity verification and correction algorithm is adopted, the recognition error caused by dirt on the surface of the water level gauge is eliminated, and the detection precision is greatly improved.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence, image recognition and water level detection technology, specifically to a dual-path visual measurement method, system, device and storage medium for water level gauge readings in industrial scenarios. Background Technology

[0002] Water level data is a crucial basis for decision-making in flood season safety, power plant scheduling, and port operations. Especially in the Yangtze River basin, where thermal power plant wharves are built along the river, changes in water level directly affect coal transportation operations and personnel safety. Therefore, real-time monitoring of water levels has significant engineering value.

[0003] Currently, thermal power plants have widely deployed AI-powered detection systems for personnel identification, equipment inspection, and behavior monitoring, possessing excellent image analysis capabilities. Some plant areas already have cameras covering the riverbank water level gauge area, enabling image acquisition. Compared to building new, costly automated monitoring equipment such as radar or ultrasonic water level gauges, using existing cameras and AI to read water level data offers advantages such as lower cost and easier deployment. It also solves problems associated with traditional manual inspections, such as low efficiency, large data delays, and inconvenience during inclement weather or at night.

[0004] In recent years, some research and patent solutions have used artificial intelligence technology to identify the scale on water level gauges from images to automate water level readings. However, these solutions generally suffer from the following drawbacks: ignoring structural information: many water level gauges contain the character "E" as a reference identifier, but existing algorithms mostly fail to recognize this feature, lacking auxiliary positioning and methods to improve reading accuracy; weak adaptability: at night or when small numbers are blurry, OCR recognition accuracy drops significantly, leading to unstable readings; lack of redundancy mechanisms: most methods rely on only a single recognition method, resulting in low fault tolerance when faced with missing or occluded characters. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is that the existing manual inspection and reading methods have problems such as low efficiency, large data delay, and inconvenience in bad weather or at night.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a dual-path visual measurement method for water level gauge readings in industrial scenarios, comprising: acquiring image samples and training a model based on an industrial scene camera; determining the current lighting conditions based on the average brightness of the image, switching between daytime and nighttime modes, and constructing a structured pixel coordinate set; outputting water level values ​​using dual paths; the training model includes training a target recognition model and an OCR detection model based on labeled materials; the dual paths include calculating using the pixel difference of the scale and the unit scale height, and estimating based on the E character count and the proportion of missing parts.

[0008] As a preferred embodiment of the dual-path visual measurement method for water level gauge readings in industrial scenarios described in this invention, the image samples include the entire water level gauge area, the character "E", the large scale number "100" and the small scale number "200", which are used to train the target recognition model and the OCR detection model, respectively.

[0009] As a preferred embodiment of the dual-path visual measurement method for water level gauge readings in industrial scenarios described in this invention, the target recognition model includes a YOLOv8 architecture and a multi-task loss function that is used to jointly optimize the E-character detection task and the overall water level gauge recognition task.

[0010] The target recognition model and ORC detection model are trained using pre-labeled image materials. The target recognition model is trained using YOLOv8 to recognize water level gauges and the character "E". A multi-task loss function is introduced, with the character "E" being used as the detection task. The loss function is expressed as follows:

[0011] ;

[0012] in, Loss for detecting the E character. For weight parameters, YOLO main detection loss, This represents the total loss.

[0013] As a preferred embodiment of the dual-path visual measurement method for water level gauge readings in industrial scenarios described in this invention, the OCR detection model includes introducing sequence consistency loss during training to impose linear constraints on the recognition results.

[0014] The OCR detection model is trained to recognize the large tick mark 100 and the small tick mark 200. Sequence consistency constraints are introduced to regularize the recognition results, as shown below:

[0015] ;

[0016] in, The total number of digits identified. For the first The numerical value of each number. This represents the sequence consistency loss value.

[0017] The image to be identified is acquired by monitoring the video stream. The recognition mode is switched by using a light intensity judgment function, which is expressed as follows:

[0018] ;

[0019] in, The average brightness value of the image. This represents the total number of pixels. For pixel index, For the first The red channel value of each pixel For the first The green channel value of each pixel. For the first The blue channel value of each pixel. To determine the threshold;

[0020] like Switch to night recognition mode;

[0021] Using a pre-trained YOLOv8 model, detect the overall bounding box of the water level gauge and the E-character bounding box from the image;

[0022] Detect the water level gauge in the image and obtain the bounding box position of the water level gauge;

[0023] Within the detected water level gauge area, target recognition is performed on the character "E", and the pixel position and RGB color value are recorded respectively;

[0024] Based on the color value of the E character, the position of the missing E character at the bottom of the water level gauge is deduced, and the pixel height is recorded.

[0025] As a preferred embodiment of the dual-path visual measurement method for water level gauge readings in industrial scenarios described in this invention, the step of determining the current illumination conditions based on the average brightness of the image includes automatically determining the illumination conditions and switching the recognition mode based on the weighted average of the RGB channels and comparing it with a preset threshold.

[0026] As a preferred embodiment of the dual-path visual measurement method for water level gauge readings in industrial scenarios described in this invention, the method of using dual-path output of water level values ​​includes calculating the water level value using the difference between scale pixels and the unit scale height.

[0027] The current water level is calculated based on the pixel height difference between the large scale number 100 and the small scale number 200, combined with the actual scale height corresponding to a single pixel.

[0028] The first in the image The vertical pixel coordinates of each small tick mark are The pixel height corresponding to each unit of water level is , No. The actual water level readings at each large scale are: ;

[0029] The actual water level value calculated using the pixel difference of the scale and the unit scale height is expressed as follows:

[0030] ;

[0031] in, For the last large scale number, This refers to the vertical pixel position of the last small tick mark below the last large tick mark. This refers to the vertical pixel position of the first small tick mark below the last large tick mark. This is to calculate the actual water level value using the difference in pixel values ​​and the unit scale height.

[0032] As a preferred embodiment of the dual-path visual measurement method for water level gauge readings in industrial scenarios described in this invention, the dual-path output of water level values ​​includes estimation based on E character count and incompleteness ratio.

[0033] Locate the last large scale number and record the reading. ;

[0034] Calculate the number of complete E characters contained below the last major tick mark and record the value. ;

[0035] Calculate the pixel height of the bottom missing E character and output the water level range of the missing E character;

[0036] The output, based on the E character count and the percentage of missing characters, represents the actual water level value estimated as follows:

[0037] ;

[0038] in, The actual range for a single complete E character. This represents the top pixel position of the missing "E" character. This represents the pixel position of the bottom edge of the missing "E" character. The actual water level value is estimated based on the E character count and the proportion of missing characters;

[0039] In daytime mode, the final water level reading identified through dual-path identification is the minimum of the two values:

[0040] ;

[0041] in, This indicates the final water level reading;

[0042] Final water level reading in night mode:

[0043] .

[0044] Another objective of this invention is to provide a dual-path visual measurement system for water level gauge readings in industrial scenarios, comprising: a model training module, a mode switching and feature extraction module, and a dual-path reading calculation module; the model training module is used to build recognition capabilities by training a YOLOv8 target detection model and an OCR detection model based on labeled industrial camera image data; the mode switching and feature extraction module is used to automatically determine whether it is day or night based on real-time image brightness, select the recognition path, and extract feature information from the water level gauge to form structured data input; the dual-path reading calculation module is used to perform water level estimation based on the extracted structured data.

[0045] Another objective of this invention is to provide a dual-path visual measurement device for water level gauge readings in industrial settings, comprising a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a dual-path visual measurement method for water level gauge readings in industrial settings.

[0046] Another object of the present invention is to provide a dual-path visual measurement storage medium for water level gauge readings in industrial scenarios, wherein a computer program is stored thereon, and when the computer program is executed by a processor, the steps of a dual-path visual measurement method for water level gauge readings in industrial scenarios are implemented.

[0047] The beneficial effects of this invention are as follows: The method of this invention reduces the delay of manual reading through automated reading and real-time monitoring, thereby improving the safety and reliability of work and production. The dual-path recognition mechanism adapts to various lighting conditions, improving robustness and accuracy. The small-scale digital continuity verification and correction algorithm eliminates recognition errors caused by dirt on the surface of the water level gauge, significantly improving detection accuracy. Without increasing hardware costs, it can automatically read water level data from traditional water level gauges using existing AI vision systems, taking into account adaptability and robustness in both daytime and nighttime scenarios. This invention is particularly suitable for scenarios such as power plants and docks where there is existing camera coverage but no automatic water level monitoring equipment has been deployed, offering advantages such as simple deployment, high recognition accuracy, strong adaptability, and low maintenance costs. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 The flowchart of the water level gauge reading recognition process is provided in Embodiment 1 of the present invention for a dual-path visual measurement method for water level gauge reading in an industrial setting.

[0050] Figure 2 This is a daytime image of a water level using a dual-path visual measurement method for water level gauge readings in an industrial setting, as provided in Embodiment 1 of the present invention.

[0051] Figure 3 This is a nighttime image of a water level measured using a dual-path visual measurement method for water level gauge readings in an industrial setting, as provided in Embodiment 1 of the present invention.

[0052] Figure 4 This is a schematic diagram illustrating the correction of small-scale digital jumps in a dual-path visual measurement method for water level gauge readings in an industrial setting, as provided in Embodiment 1 of the present invention. Detailed Implementation

[0053] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0054] Example 1, referring to Figures 1-4 As an embodiment of the present invention, a dual-path visual measurement method for water level gauge readings in industrial scenarios is provided, comprising:

[0055] S1: Relying on industrial scene cameras, collect image samples to train the model.

[0056] Furthermore, relying on the power plant's existing monitoring system to collect water level gauge images, the image data covers different scene conditions such as day / night and sunny / rainy days. Real-time video frame images are acquired through surveillance cameras deployed in the power plant's dock area, such as... Figure 2 and Figure 3 As shown, where Figure 2 Images taken during the day, Figure 3 Images taken at night; both daytime and nighttime images include a large scale number 100 and a small scale number 200.

[0057] The image is labeled with the following targets: the entire area of ​​the water level gauge, the character "E", the large scale number 100, and the small scale number 200.

[0058] The target recognition model and ORC detection model are trained using pre-labeled image materials. The target recognition model is trained using YOLOv8 to identify water level gauges and the character "E". A multi-task loss function is introduced, treating the "E" character as an independent detection task. The loss function is expressed as follows:

[0059] ;

[0060] in, Loss detection for the character "E" For weight parameters, YOLO main detection loss, This represents the total loss.

[0061] An OCR detection model is trained to recognize the large-scale digit 100 and the small-scale digit 200. Sequence consistency constraints are introduced to regularize the recognition results and enhance the constraint of the linear relationship in the mathematical space. The formula is expressed as follows:

[0062] ;

[0063] in, The total number of digits identified. For the first The numerical value of each number. This represents the sequence consistency loss value.

[0064] It should be noted that the image to be identified is acquired by monitoring the video stream, and the day / night recognition mode is dynamically switched using a light intensity judgment function. The judgment function formula is as follows:

[0065] ;

[0066] in, The average brightness value of the image. This represents the total number of pixels. For pixel index, For the first The red channel value of each pixel For the first The green channel value of each pixel. For the first The blue channel value of each pixel. To determine the threshold, the system defaults to a value of 200.

[0067] like If so, switch to night recognition mode.

[0068] It should also be noted that the pre-trained YOLOv8 model was used to detect the overall bounding box of the water level gauge and the “E” character bounding box from the image.

[0069] First, the water level gauge in the image is detected, and its bounding box position is obtained. The corresponding region is extracted as input for subsequent processing. Then, within the detected water level gauge region, the "E" character is identified, and all complete "E" characters are detected. The pixel position and RGB color value are recorded respectively. Finally, based on the color value of the "E" character, the position of the missing "E" character at the bottom of the water level gauge is inferred, and its pixel height is recorded as the basis for subsequent calculations.

[0070] S2: Determine the current lighting conditions based on the average brightness of the image, switch between day mode and night mode, and construct a structured set of pixel coordinates.

[0071] Furthermore, the trained OCR detection model is used to extract the large scale number 100 and the small scale number 200 in the detection area, and the pixel position of the detection box is recorded. In the night mode, only the large scale number 100 is extracted.

[0072] In daytime mode, a digit continuity verification algorithm is used to complete missing digits. It checks for jumps between adjacent small tick marks; if a jump exists, the missing digits are filled in based on the jump distance. For example... Figure 4 As shown on the left, the missing number 301 indicates a missing digit 2. To fill in the missing digit, a digit continuity verification algorithm is executed. The specific process is as follows:

[0073] Let the sequence of small tick numbers identified consecutively between two adjacent large tick numbers be:

[0074] ;

[0075] The coordinates of its pixels are:

[0076] ;

[0077] Define the residual sequence as:

[0078] ;

[0079] Calculate the average pixel spacing ,like:

[0080] ;

[0081] This indicates that there are missing scale numbers, which need to be filled in. Intermediate numbers can be estimated by interpolation as follows:

[0082] ;

[0083] like Figure 4 As shown in the image on the right, the number 301' can be automatically completed after the digit continuity verification algorithm is executed. Figure 4The missing number 2 is between the small scale numbers 1 and 3 on the left side of the middle scale.

[0084] S3: Uses dual-path output for water level values.

[0085] Furthermore, based on the acquired data, the actual reading of the water level gauge is calculated, which involves two identification paths:

[0086] Path 1 is used only in daytime mode. It calculates the current water level based on the pixel height difference between the large scale number 100 and the small scale number 200, combined with the actual scale height corresponding to a single pixel. The mathematical formula is as follows:

[0087] Let the image contain the first... The vertical pixel coordinates of each small tick mark are: ;

[0088] The pixel height corresponding to each unit of water level is: ;

[0089] No. The actual water level readings for each large scale are: ;

[0090] The water level reading can then be calculated using the following formula:

[0091] ;

[0092] in, For the last large scale number, This refers to the vertical pixel position of the last small tick mark below the last large tick mark. This refers to the vertical pixel position of the first small tick mark below the last large tick mark. This is the actual water level value.

[0093] Path 2 is recognized and used in both daytime and nighttime modes.

[0094] First, locate the last large scale number and record its reading. .

[0095] Furthermore, calculate the number of complete "E" characters contained below the last large tick mark and record its value. .

[0096] Furthermore, the pixel height of the bottommost missing "E" character is calculated to obtain the water level range represented by the missing "E" character.

[0097] Finally, the actual water level reading is calculated using the values ​​obtained above, using the following formula:

[0098] ;

[0099] in, The actual range for a single complete "E" character. This represents the top pixel position of the missing "E" character. This represents the pixel position of the bottom edge of the missing "E" character. This is the actual water level value;

[0100] Ultimately, we can conclude that:

[0101] In daytime mode, the final water level reading identified through dual-path identification is the minimum of the two values:

[0102] ;

[0103] In night mode, the final water level reading is the reading obtained from path 2:

[0104] ;

[0105] Example 2, one embodiment of the present invention, provides a dual-path visual measurement method for water level gauge readings in an industrial setting. To verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculations and simulation experiments.

[0106] First, in the same scenario, covering daytime, nighttime, and situations where the water level gauge surface is obscured by dirt, different detection methods were used, and the average value of 1000 measurement results was taken to obtain the error comparison data, as shown in the table below:

[0107] Table 1. Comparison of Recognition Errors

[0108] Serial Number Scene Description Error of traditional method (cm) Error in this plan (cm) Accuracy improvement (%) 1 During the day, without obstruction 2.8 1.1 60.7% 2 During the day, there is shelter 4.6 1.9 58.7% 3 At night, without any cover 5.1 1.6 68.6% 4 At night, there is shelter. 6.8 2.4 64.7%

[0109] In the same scenario, different detection schemes were used to compare detection capabilities and stability. The comparison data is shown in the table below:

[0110] Table 2 Comparison of Detection Capability and Stability

[0111] Project Indicators Traditional solution This plan Improvement status Success rate of continuous frame recognition 81.2% 96.5% An increase of 15.3% "E" character detection support support support none Digital scale detection support Partial support support Add features OCR scale false detection rate 22.4% 7.1% Reduced by 68.3% Nighttime recognition capability Difference excellent Significant improvement

[0112] As can be seen from the experimental results in Tables 1 and 2, the present invention can effectively reduce the error of the final identification result of the water level gauge reading, thereby improving the accuracy of the detection result.

[0113] This solution employs a dual-path recognition method, ensuring accuracy both daytime and nighttime. Based on a scale digit completion mechanism, this solution effectively addresses the issue of missing small scale digits caused by dirt or obstruction on the water level gauge, further improving recognition accuracy.

[0114] By comparing the data in Tables 1 and 2, it can be concluded that the present invention can provide higher reading accuracy and more stable detection capability under various test conditions, and is an advanced technology that is superior to traditional technical solutions.

[0115] Example 3, an embodiment of the present invention, provides a dual-path visual measurement system for water level gauge readings in industrial scenarios, including a model training module, a mode switching and feature extraction module, and a dual-path reading calculation module.

[0116] The model training module is used to build recognition capabilities. It trains the YOLOv8 target detection model and OCR detection model based on labeled industrial camera image data. The mode switching and feature extraction module is used to automatically determine whether it is day or night based on the real-time image brightness, select the recognition path, and extract feature information from the water level gauge to form structured data input. The dual-path reading calculation module is used to perform water level estimation based on the extracted structured data.

[0117] This embodiment also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a dual-path visual measurement method for water level gauge readings in an industrial setting as proposed in the above embodiment.

[0118] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements a dual-path visual measurement method for water level gauge readings in an industrial setting as proposed in the above embodiment.

[0119] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0120] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0121] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0122] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dual-path visual measurement method for water level gauge readings in industrial settings, characterized in that, include: Using cameras in industrial settings, image samples are collected to train the model; The current lighting conditions are determined based on the average brightness of the image, and the daytime mode and nighttime mode are switched to construct a structured set of pixel coordinates. The water level value is output via a dual-path system. The training models include a target recognition model and an OCR detection model trained based on labeled materials; The target recognition model includes a YOLOv8 architecture and the introduction of a multi-task loss function to jointly optimize the E character detection task and the overall water level gauge recognition task. The target recognition model and ORC detection model are trained using pre-labeled image materials. The target recognition model is trained using YOLOv8 to recognize water level gauges and the character "E". A multi-task loss function is introduced, with the character "E" being used as the detection task. The loss function is expressed as follows: ， in, Loss for detecting the E character. For weight parameters, YOLO main detection loss, Total loss; The OCR detection model includes introducing sequence consistency loss during training to impose linear constraints on the recognition results; The OCR detection model is trained to recognize large-scale digits (100) and small-scale digits (200). Sequence consistency constraints are introduced to regularize the recognition results, as shown below: ， in, The total number of digits identified. For the first The numerical value of each number. This represents the sequence consistency loss value. The image to be identified is acquired by monitoring the video stream. The recognition mode is switched by using a light intensity judgment function, which is expressed as follows: ， in, The average brightness value of the image. This represents the total number of pixels. For pixel index, For the first The red channel value of each pixel For the first The green channel value of each pixel. For the first The blue channel value of each pixel. To determine the threshold; like Switch to night recognition mode; Using a pre-trained YOLOv8 model, detect the overall bounding box of the water level gauge and the E-character bounding box from the image; Detect the water level gauge in the image and obtain the bounding box position of the water level gauge; Within the detected water level gauge area, target recognition is performed on the character "E", and the pixel position and RGB color value are recorded respectively; Based on the color value of the E character, the position of the missing E character at the bottom of the water level gauge is inferred, and the pixel height is recorded. The dual-path approach includes calculation using the difference between scale pixels and the unit scale height, and estimation based on the E character count and the proportion of missing parts. The method of using dual-path output water level values ​​includes calculating the water level value using the difference between scale pixels and the unit scale height. The current water level is calculated based on the pixel height difference between the large scale number (100) and the small scale number (200), combined with the actual scale height corresponding to a single pixel. The first in the image The vertical pixel coordinates of each small tick mark are The pixel height corresponding to each unit of water level is , No. The actual water level readings at each large scale are: ; The actual water level value calculated using the pixel difference of the scale and the unit scale height is expressed as follows: ， in, For the last large scale number, This refers to the vertical pixel position of the last small tick mark below the last large tick mark. This refers to the vertical pixel position of the first small tick mark below the last large tick mark. The actual water level value is calculated using the difference between the pixels on the scale and the unit scale height; The dual-path output water level value includes estimation based on the E character count and the incompleteness ratio; Locate the last large scale number and record the reading. ; Calculate the number of complete E characters contained below the last major tick mark and record the value. ; Calculate the pixel height of the bottom missing E character and output the water level range of the missing E character; The output, based on the E character count and the percentage of missing characters, represents the actual water level value estimated as follows: ， in, The actual range for a single complete E character. The top pixel position of the missing E character. This represents the pixel position of the bottom edge of the missing "E" character. The actual water level value is estimated based on the E character count and the proportion of missing characters; In daytime mode, the final water level reading identified through dual-path identification is the minimum of the two values: ， in, This indicates the final water level reading; Final water level reading in night mode: 。 2. The dual-path visual measurement method for water level gauge readings in industrial scenarios as described in claim 1, characterized in that: The image samples include, The entire water level gauge area, the character "E", the large scale number (100) and the small scale number (200) are used to train the target recognition model and the OCR detection model, respectively.

3. The dual-path visual measurement method for water level gauge readings in industrial scenarios as described in claim 1, characterized in that: The method of determining the current lighting conditions based on the average brightness of the image includes... Based on the weighted average of the RGB channels, the system automatically determines the lighting conditions and switches the recognition mode by comparing the results with a preset threshold.

4. A dual-path visual measurement system for water level gauge readings in industrial settings, employing the dual-path visual measurement method for water level gauge readings in industrial settings as described in any one of claims 1 to 3, characterized in that: Includes a model training module, a mode switching and feature extraction module, and a dual-path reading calculation module; The model training module is used to build recognition capabilities, and trains the YOLOv8 target detection model and OCR detection model based on labeled industrial camera image data; The mode switching and feature extraction module is used to automatically determine whether it is day or night based on the real-time image brightness, select the recognition path, and extract feature information from the water level gauge to form structured data input. The dual-path reading calculation module is used to perform water level estimation based on the extracted structural data.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the dual-path visual measurement method for water level gauge readings in industrial scenarios as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the dual-path visual measurement method for water level gauge readings in industrial scenarios as described in any one of claims 1 to 3.

Images