A camera water meter reading accuracy self-checking method

Abstract

The application discloses a kind of camera water meter reading accuracy self-checking method, belongs to camera water meter reading accuracy self-checking method technical field, and its technical scheme main point is: including the following steps: including the following steps:S1.camera water meter reading accuracy self-checking method image acquisition and processing: control camera continuous acquisition multiple frame dial image, and each frame image is preprocessed, S2.camera water meter reading accuracy self-checking method single frame evaluation and identification: to each frame preprocessed image, image quality score and character recognition are executed in parallel, respectively obtain the image quality score and preliminary reading of each frame, S3.camera water meter reading accuracy self-checking method multiple frame consistency check: compare the preliminary reading of all frames, if consistent or based on voting decision can determine candidate reading, then enter step S4, effect is the combination of two constitutes rigid test to identification reliability, significantly reduces the upload probability of false alarm and data anomaly from source.

Description

Technical Field

[0001] This invention belongs to the technical field of self-testing methods for the accuracy of camera water meter readings, and in particular, a self-testing method for the accuracy of camera water meter readings. Background Technology

[0002] Camera water meters capture images of the meter dial using a built-in camera and automatically read the readings, enabling remote meter reading. They offer advantages such as not altering the mechanical structure and low power consumption. However, they face significant challenges in practical applications, the core issue being the lack of an effective mechanism to determine and guarantee the reliability of single-read results.

[0003] Specifically, the challenges mainly come from two aspects: First, at the data input level, dark and humid installation environments can easily lead to lens contamination, moisture blockage, and uneven lighting, resulting in poor image quality; second, at the data processing level, existing recognition algorithms are prone to misjudgment when faced with low-quality images or critical situations where the character wheel is about to carry over. Currently, the system typically uploads the results of a single recognition directly, with only delayed manual review in the background, making it impossible to detect and correct errors in real time.

[0004] More importantly, existing technologies fail to systematically utilize the inherent physical logic of water meters (such as the non-negativity of water consumption and the continuity of digit carry) to perform real-time logical verification of readings, nor do they effectively cross-validate with heterogeneous data sources such as pulse sensors (if present).

[0005] Therefore, there is an urgent need for a method that can proactively assess the reliability of readings, integrate multi-dimensional information for self-checking, and trigger corresponding maintenance, so as to fundamentally improve the accuracy and reliability of the output of camera water meters.

[0006] The purpose of this invention is to provide a self-testing method for the accuracy of camera water meter readings, so as to solve the problems mentioned in the background art. Summary of the Invention

[0007] The purpose of this invention is to provide a self-testing method for the accuracy of camera water meter readings, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a self-testing method for the accuracy of camera water meter readings, comprising the following steps: S1. Self-test method for the accuracy of video water meter readings: Image acquisition and processing: Control the camera to continuously acquire multiple frames of dial images, and preprocess each frame; S2. Self-testing method for the accuracy of video water meter readings: Single-frame evaluation and recognition: For each pre-processed image frame, image quality scoring and character recognition are performed in parallel to obtain the image quality score for each frame. and preliminary readings ; S3. Self-test method for the accuracy of video water meter readings: Multi-frame consistency verification: Compare the preliminary readings of all frames. If the candidate readings can be determined by consensus or through voting decisions If the condition is met, proceed to step S4; otherwise, trigger the exception handling process. S4. Self-testing method for the accuracy of video-recorded water meter readings: Mechanical logic verification based on the physical laws of water meter measurement, for the candidate readings... Perform a rationality check, including verifying the non-negativity of the water volume increment, the rationality of the range, and the continuity of the last digit; if the verification passes, proceed to step S5; otherwise, trigger the exception handling process. S5. Confidence Synthesis and Output of Self-Test Method for Accuracy of Camera Water Meter Reading: Based on the results of the multi-frame consistency check, the image quality score, and the mechanical logic verification, calculate the comprehensive confidence level C of this reading and output the final reading. And its corresponding confidence level C.

[0009] Furthermore, the image quality scoring in step S2 includes calculating at least one of the following indicators: image sharpness, contrast, illumination uniformity, and contamination occlusion rate, and then comprehensively obtaining the image quality score. .

[0010] Furthermore, in step S5, the formula for calculating the overall confidence level C is: ,in, The percentage of frames that passed the consistency check. The normalized average of the image quality scores for all frames. The values ​​are the quantified values ​​of the logical verification results, and w1, w2, and w3 are preset weight coefficients.

[0011] Furthermore, the anomaly handling process includes at least one of the following measures: instructing the camera to adjust parameters and re-acquire images; triggering the lens cleaning mechanism to perform cleaning; and uploading the anomaly data and logs to the management platform.

[0012] Furthermore, after step S4, the method further includes: S4a. Self-test method for the accuracy of video water meter readings: Pulse data cross-validation: Synchronously read the cumulative values ​​of independent water volume pulse sensors. Calculate pulse increment The pulse increment ΔS is compared with the water volume increment identified by the image. Compare; If the deviation between the two is within the allowable error range, then the overall confidence level C is increased; If the deviation between the two exceeds the allowable error range and step S4 fails verification, then the pulse data is adopted as the current water volume benchmark, and calibration or alarm is triggered.

[0013] Furthermore, the trigger calibration includes: utilizing the accumulated value of the pulse sensor. The parameters of the image recognition model are corrected in reverse.

[0014] Furthermore, when the camera water meter is a pointer-type water meter, the character recognition in step S2 is replaced by pointer angle recognition to obtain the reading; the multi-frame consistency verification in step S3 is replaced by multi-frame consistency verification of the same pointer recognition angle; and the mechanical logic verification in step S4 is replaced by verification of the linkage carry relationship between pointers.

[0015] Furthermore, it also includes a pointer motion trend analysis step: based on the continuous multi-frame images, the movement trajectory of the pointer tip is tracked to determine whether its continuity conforms to the water flow driving law.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention systematically improves reading accuracy by integrating a dual mechanism of "multi-frame image consistency verification" and "mechanical state logic verification". Multi-frame verification can effectively resist instantaneous interference and recognition fluctuations, while logic verification uses the physical laws of water volume being "non-negative, continuous, and bounded" to intercept abnormal results. The combination of the two constitutes a rigid test for recognition reliability, significantly reducing the probability of false alarms and data anomalies from the source. 2. This invention innovatively introduces and calculates a comprehensive confidence level, enabling a quantitative assessment of the reliability of readings. This confidence level not only serves as a data quality label but also drives automated system decision-making. When the confidence level is insufficient, it automatically triggers proactive maintenance measures such as retaking or cleaning. Furthermore, it can monitor the health status of equipment through its long-term trend. This transforms the system from a passive and lagging manual verification mode into an intelligent agent with a closed-loop capability of "perception-assessment-action," significantly improving operational efficiency. 3. This invention expands the verification dimensions through data fusion, constructing a more robust reliability barrier. Building upon the inherent verification dimensions of image sequences and mechanical logic, it further introduces pulse sensors for heterogeneous data cross-verification. This design forms a fault-tolerant mechanism of "image as the primary criterion, pulse as the secondary verification," which can strengthen corroboration when consistent and intelligently arbitrate and calibrate when deviations exceed limits, thus providing a more robust guarantee for accurate measurement in complex environments. 4. The self-testing method proposed in this invention is a highly versatile technical framework with good scalability and adaptability. Its core processes (quality assessment, consistency verification, logic verification, confidence synthesis, and anomaly handling) do not depend on specific phenotypes. By adapting the recognition targets and rules (such as adapting character recognition to angle recognition), it can be seamlessly migrated to different types such as pointer-type water meters. This framework can also be easily compatible with other sensors in the future, providing a scalable path for the reliability design of smart meters. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention and the technical solutions in the prior art, the drawings used in the description of the specific embodiments and the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system flow in this invention; Figure 2 This is a schematic diagram of the system module connections in this invention. Detailed Implementation

[0019] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without any of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0020] Unless otherwise defined, the directions mentioned herein, such as up, down, left, right, front, back, inside, and outside, are based on the directions shown in the figures of this invention, and are explained here together.

[0021] The connection method can be any existing method, such as bonding, welding, or bolting, depending on the actual needs.

[0022] Please see Figures 1 to 2 As shown, this invention provides a self-testing method for the accuracy of camera water meter readings, including Example 1: System Architecture and Data Link This embodiment provides a data fusion and integration system for primary and secondary equipment in a power distribution network based on a power grid resource business platform, including: Example 1: A basic self-testing method based on multi-frame recognition and logical consistency.

[0023] This embodiment provides a basic self-test process applicable to most video water meters that use a single camera to capture images of the dial.

[0024] The method includes the following steps: Step S101: Image acquisition and preprocessing.

[0025] The control module instructs the camera to continuously capture images under specific lighting conditions (such as when a supplementary light is on). Frames (e.g.) ); The water meter dial image is preprocessed for each frame, including grayscale conversion, noise reduction, perspective correction (if tilting exists), and region of interest (ROI). (i.e., digital area) cropping.

[0026] Step S102: Single-frame image quality assessment and reading recognition.

[0027] For each preprocessed image frame, the following two sub-steps are performed in parallel: S1021: Image quality score, calculated for this frame. Image quality metrics for the region, including but not limited to: ①: Sharpness: Calculated using the Laplace variance method, the higher the value, the sharper the image.

[0028] ②: Contrast: Calculates the grayscale contrast between the numbers and the background.

[0029] ③: Illumination uniformity: assessment Internal luminance standard deviation.

[0030] ④: Pollution Coverage Rate: Estimate the percentage of area covered by suspected water stains and dirt through edge detection and morphological operations.

[0031] A threshold is set for each indicator, and the results are weighted and combined to obtain an image quality score from 0 to 100. ( (Frame number).

[0032] S1022: Character recognition, utilizing pre-trained deep learning models (such as CRNN networks) or traditional template matching algorithms to... The region is segmented and recognized to obtain the initial reading of the frame. (Usually a multi-digit numeric string).

[0033] Step S103: Multi-frame reading consistency verification.

[0034] contrast Readings from frame recognition .

[0035] ①: If all If they are exactly the same, proceed to step S105.

[0036] ②: If there is disagreement, a voting decision will be made: the reading with the highest frequency will be selected as the candidate reading. If the percentage of frames with the highest frequency readings exceeds a preset threshold (e.g., 70%), then Proceed to the next step; otherwise, it is determined as "consistency check failed", and directly jumps to step S106 to trigger advanced self-check or alarm.

[0037] Step S104: Logical verification based on mechanical state.

[0038] The water meter's digits use a mechanical increment mechanism, and its readings must follow physical laws. This step involves processing candidate readings. Perform a reasonableness check: ①: Read historical readings (as in the last successfully identified reading).

[0039] ②: Calculate the increase in water volume: Theoretically (The water meter does not reverse.)

[0040] ③: Verify the increment range: Based on the water meter's maximum flow rate and the time interval between two readings, calculate the possible reasonable increment range. ,like If the value exceeds this range (e.g., a negative value or an abnormally large positive value appears), it is considered to be illogical.

[0041] ④: Verify the continuity of the last digit carry: Analyze the historical reading sequence. The last digit wheel should increase cyclically from 0 to 9. If the change of the last digit in this identification does not conform to the continuous increase pattern of the last digit in the previous identification (e.g., if the last digit was 8, the current digit should be 9, 0, or 1, and should not jump to 4), and there are no large-scale water use events during the period, then it is suggested that there may be an identification error.

[0042] If this step verifies the accuracy, proceed to S105; otherwise, proceed to S106. Self-test method for the accuracy of video water meter readings. Step S105: Output the final reading and confidence level.

[0043] Calculate the overall confidence level C of this reading based on the following factors: ①: The percentage of frames that pass the self-test method for ensuring the accuracy of water meter readings via video camera.

[0044] ②: The normalized average of the image quality scores for all frames.

[0045] ③: If the logic verification passes, the value is 1.0; otherwise, it is 0.7 (indicating doubt).

[0046] ④: Final confidence level , among which For weighting coefficients (e.g., 0.4, 0.3 for the self-test method of video water meter reading accuracy, 0.3 for the self-test method of video water meter reading accuracy).

[0047] Output final reading The reading and its confidence level C are recorded. If C is higher than the high confidence threshold (e.g., 0.85), it is marked as "highly reliable". If C is between the medium and low thresholds, it is marked as "reference, recommended to pay attention". At the same time, the reading and status are stored in the historical database.

[0048] Step S106: Exception handling and proactive self-check.

[0049] If a verification fails or the confidence level is too low during the above steps, the exception handling process is triggered: 1. Retake Command: Adjust parameters (such as fill light intensity and shooting angle) to re-acquire the image.

[0050] 2. Trigger Cleaning: If the "Contamination Occlusion Rate" in the image quality score remains too high, the solenoid valve will be activated to start the water jet or the scraping mechanism (if equipped) to clean the lens.

[0051] 3. Report anomalies: Upload low-confidence readings, original images, and self-test failure logs to the management platform and request manual intervention for verification.

[0052] 4. Trend Compensation: In extreme cases, if reliable image recognition cannot be obtained continuously, a reference usage can be temporarily estimated based on the historical water use trend model and marked as "estimated value".

[0053] Example 2: Enhanced self-testing method using a fusion water pulse sensor This embodiment, based on embodiment 1, adds a high-precision water volume pulse sensor (such as a Hall sensor to detect impeller rotation) to the water meter. This sensor generates one pulse for every unit volume (e.g., 0.01 liters) of water flowing through it, forming a cumulative water volume independent of the image reading. .

[0054] The enhanced self-test method, following step S104 in Example 1, adds: Step S104': Pulse data cross-validation.

[0055] ①: At each image recognition moment, the cumulative value of the pulse sensor is read synchronously. .

[0056] ②: Calculate the pulse increment since the last image recognition: .

[0057] ③: Recognize the water volume increment from the image With pulse increment In comparison, due to the different measurement principles of the two instruments, minor errors are permissible (within the overall instrument error range, such as ±1%). The deviation ratio is then calculated. .

[0058] ④: If If the error is less than the allowable error threshold, it strongly supports the image reading. Accuracy, significantly improving its confidence level (e.g., by adding an item to the final confidence level calculation). ).

[0059] ⑤: If If the threshold is exceeded and the image self-check (Example 1) also fails, a recognition error or sensor malfunction is highly likely. The system will: (a): Prioritize pulse data As a more reliable benchmark for current cumulative water volume.

[0060] (b): Using pulse data Reverse calibration of image recognition models: for example, known The corresponding true number combination should be close to "ABC.D", which can correct the model's misjudgment of fuzzy numbers and realize the model's self-learning optimization.

[0061] (c): Issue a specific alarm indicating that "the deviation between the image reading and the pulse measurement exceeds the limit".

[0062] This embodiment introduces heterogeneous data sources to construct a robust system of "image-based primary judgment, pulse-based secondary verification, and mutual correction," greatly improving the ability to ensure reading accuracy in complex environments. This is a self-testing method for the accuracy of camera-based water meter readings. Example 3: Self-test adaptation scheme for dial pointer water meters For pointer-type camera water meters, the core of their self-test lies in verifying the accuracy of angle recognition. The method in this embodiment is adjusted as follows: ①: Step S1022: Change to identifying the center of each pointer and the angle of the pointer tip, and convert it into a reading based on the dial scale.

[0063] ②: Step S103: Multi-frame consistency verification becomes a comparison of the angles identified by the same pointer, and the standard deviation of the angle difference is calculated. If the standard deviation is too large, it is unreliable.

[0064] ③: Step S104: The logic verification focuses on ensuring that the relative positional relationship between the pointers conforms to the range carry rule (e.g., when the tens pointer rotates one full circle, the units pointer moves one increment). Any identification result that violates this mechanical relationship will be judged as incorrect.

[0065] ④: Add a step: analyze the pointer movement trend. Combine multiple frames of images captured continuously within a short period of time to track the movement trajectory of the pointer tip. This trajectory should match the uniform or accelerated movement driven by the water flow. If the identified pointer position jumps discontinuously, it may be a misidentification.

[0066] Industrial Applicability: The self-inspection method provided by this invention can be integrated into the edge computing module of a camera water meter through software algorithms or deployed on a cloud analysis platform. It can effectively distinguish between reliable readings and suspicious readings, reduce the workload of manual verification, and improve the automation and reliability of remote meter reading systems. At the same time, the confidence index and abnormal data it provides provide valuable information for water affairs big data analysis and predictive maintenance of water meter status.

[0067] It should be noted that, in this document, relational terms such as "one" and "two" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, and also elements inherent to such process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0068] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A self-testing method for the accuracy of camera water meter readings, characterized in that... This includes the following steps: S1. Self-test method for the accuracy of video water meter readings: Image acquisition and processing: Control the camera to continuously acquire multiple frames of dial images, and preprocess each frame; S2. Self-testing method for the accuracy of video water meter readings: Single-frame evaluation and recognition: For each pre-processed image frame, image quality scoring and character recognition are performed in parallel to obtain the image quality score for each frame. and preliminary readings ; S3. Self-test method for the accuracy of video water meter readings: Multi-frame consistency verification: Compare the preliminary readings of all frames. If the candidate readings can be determined by consensus or through voting decisions If the condition is met, proceed to step S4; otherwise, trigger the exception handling process. S4. Self-testing method for the accuracy of video-recorded water meter readings: Mechanical logic verification based on the physical laws of water meter measurement, for the candidate readings... Perform a rationality check, including verifying the non-negativity of the water volume increment, the rationality of the range, and the continuity of the last digit; if the verification passes, proceed to step S5; otherwise, trigger the exception handling process. S5. Confidence Synthesis and Output of Self-Test Method for Accuracy of Camera Water Meter Reading: Based on the results of the multi-frame consistency check, the image quality score, and the mechanical logic verification, calculate the comprehensive confidence level C of this reading and output the final reading. And its corresponding confidence level C.

2. The self-testing method for the accuracy of camera water meter readings according to claim 1, characterized in that... The image quality scoring in step S2 includes calculating at least one of the following indicators: image sharpness, contrast, illumination uniformity, and contamination occlusion rate, and then comprehensively obtaining the image quality score. .

3. The self-testing method for the accuracy of camera water meter readings according to claim 1, characterized in that... In step S5, the formula for calculating the overall confidence level C is: ,in, The percentage of frames that passed the consistency check. The normalized average of the image quality scores for all frames. The values ​​are the quantified values ​​of the logical verification results, and w1, w2, and w3 are preset weight coefficients.

4. The self-testing method for the accuracy of camera water meter readings according to claim 1, characterized in that... The anomaly handling process includes at least one of the following measures: instructing the camera to adjust parameters and re-acquire images; triggering the lens cleaning mechanism to perform cleaning; and uploading the anomaly data and logs to the management platform.

5. The self-testing method for the accuracy of camera water meter readings according to claim 1, characterized in that... After step S4, the following is also included: S4a. Self-test method for the accuracy of video water meter readings: Pulse data cross-validation: Synchronously read the cumulative values ​​of independent water volume pulse sensors. Calculate pulse increment The pulse increment ΔS is compared with the water volume increment identified by the image. Compare; If the deviation between the two is within the allowable error range, then the overall confidence level C is increased; If the deviation between the two exceeds the allowable error range and step S4 fails verification, then the pulse data is adopted as the current water volume benchmark, and calibration or alarm is triggered.

6. The self-testing method for the accuracy of camera water meter readings according to claim 5, characterized in that... The trigger calibration includes: utilizing the accumulated value of the pulse sensor. The parameters of the image recognition model are corrected in reverse.

7. The self-testing method for the accuracy of camera water meter readings according to claim 1, characterized in that... When the camera water meter is a pointer-type water meter, the character recognition in step S2 is replaced by pointer angle recognition to obtain the reading; the multi-frame consistency verification in step S3 is replaced by multi-frame consistency verification of the same pointer recognition angle; and the mechanical logic verification in step S4 is replaced by verification of the linkage carry relationship between pointers.

8. The self-testing method for the accuracy of camera water meter readings according to claim 7, characterized in that... It also includes a pointer motion trend analysis step: based on continuous multi-frame images, the movement trajectory of the pointer tip is tracked to determine whether its continuity conforms to the water flow driving law.

Images