Field surface water depth measuring method and system based on image intelligent identification

Abstract

The invention relates to the technical field of image processing, and discloses a field surface water depth measuring method and system based on image intelligent identification. The method comprises the following steps: performing multi-period image acquisition and processing on a field to obtain preprocessed image data; inputting the preprocessed image data into a two-stage network for matching to obtain three-dimensional coordinates of the water gauge; segmenting based on the water gauge coordinates to obtain scale readings and water surface intersection point data; performing interpolation processing on the reading data and the field point cloud to obtain a water depth model, and performing decision analysis to obtain a heterogeneity distribution diagram and a diagnosis matrix; and inputting the distribution diagram and the matrix into a learning framework for optimization processing to obtain an irrigation control parameter set. The water gauge scale can be accurately identified by using an image intelligent identification technology under a complex field environment condition, high-precision automatic measurement of the field surface water depth is realized, and a measurement result and an irrigation control decision are organically combined to form an intelligent closed-loop management system, so that the utilization efficiency of irrigation water resources is improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and particularly to a method and system for measuring surface water depth based on image intelligent recognition. Background Art

[0002] In the field of agricultural irrigation, traditional methods for measuring surface water depth mainly rely on manual observation of water gauge readings. Staff need to regularly visit the field site, record the water level values on the water gauge, and calculate the water depth data through simple calculations. With the development of technology, some semi-automated methods have been applied in practice, such as installing water level sensors, buoy-type water level gauges and other devices to achieve automatic collection of water level data. However, these methods are still limited to point measurements and cannot comprehensively reflect the spatial distribution of water depth within the field. In recent years, remote sensing technology has been applied to a certain extent in large-scale water area monitoring, but due to resolution limitations, it is difficult to meet the needs of refined management of small-scale farmland irrigation. At the same time, computer vision technology has made remarkable progress in the field of industrial inspection, but its application research in the field of agricultural water conservancy lags behind relatively, especially the specialized algorithms for water gauge recognition and water depth measurement are still not perfect.

[0003] There are many deficiencies in traditional technical methods: First, the manual observation method has a large labor intensity and low efficiency, and it is difficult to achieve high-frequency monitoring, resulting in a lack of timely data support for irrigation management decisions; Second, point measurements cannot reflect the spatial distribution characteristics of water depth within the field, resulting in a lack of scientific basis for evaluating irrigation uniformity; Third, the existing automated equipment has high costs and complex maintenance, and faces economic obstacles in large-scale popularization and application; Fourth, there is a lack of a closed-loop management mechanism that organically combines the water depth measurement results with the optimization of irrigation decisions, resulting in limited room for improving irrigation efficiency; Fifth, traditional methods are difficult to cope with interference factors in complex irrigation environments, such as light changes, water surface fluctuations, weed occlusion, etc., and it is difficult to ensure the accuracy and stability of measurement results. These deficiencies seriously restrict the scientific and precise level of farmland irrigation, and affect the utilization efficiency of water resources and the improvement of crop yields. Summary of the Invention

[0004] This application provides a method and system for measuring surface water depth based on image intelligent recognition, which is used to accurately identify the water gauge scale using image intelligent recognition technology under complex field environmental conditions, achieve high-precision automatic measurement of surface water depth, and organically combine the measurement results with irrigation control decisions to form an intelligent closed-loop management system, thereby improving the utilization efficiency of irrigation water resources.

[0005] In a first aspect, the present application provides a method for measuring the water depth of a paddy field surface based on image intelligent recognition. The method for measuring the water depth of a paddy field surface based on image intelligent recognition includes: performing multi-period image acquisition on a paddy field with a water gauge and performing time-series processing to obtain time-series preprocessed image data; inputting the time-series preprocessed image data into a two-stage feature pyramid network for multi-scale anchor box matching to obtain water gauge three-dimensional coordinate and attitude angle data; performing instance segmentation based on the water gauge three-dimensional coordinate and attitude angle data to obtain water gauge scale reading and water surface boundary intersection point data; combining the water gauge scale reading and water surface boundary intersection point data with the high-density discrete point cloud of the actually measured sampled paddy field, and performing variational adaptive thin plate spline interpolation and hydraulic simulation correction processing to obtain a dynamic water depth distribution grid model; performing Markov decision analysis based on the dynamic water depth distribution grid model to obtain an irrigation heterogeneity distribution map and an irrigation system fault diagnosis matrix; inputting the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into a deep reinforcement learning framework for multi-step decision optimization processing under constraint conditions to obtain a regional irrigation control parameter set.

[0006] In a second aspect, the present application provides a system for measuring the water depth of a paddy field surface based on image intelligent recognition. The system for measuring the water depth of a paddy field surface based on image intelligent recognition includes:

[0007] A processing module, configured to perform multi-period image acquisition on a paddy field with a water gauge and perform time-series processing to obtain time-series preprocessed image data;

[0008] A matching module, configured to input the time-series preprocessed image data into a two-stage feature pyramid network for multi-scale anchor box matching to obtain water gauge three-dimensional coordinate and attitude angle data;

[0009] A segmentation module, configured to perform instance segmentation based on the water gauge three-dimensional coordinate and attitude angle data to obtain water gauge scale reading and water surface boundary intersection point data;

[0010] A correction module, configured to combine the water gauge scale reading and water surface boundary intersection point data with the high-density discrete point cloud of the actually measured sampled paddy field, and perform variational adaptive thin plate spline interpolation and hydraulic simulation correction processing to obtain a dynamic water depth distribution grid model;

[0011] An analysis module, configured to perform Markov decision analysis based on the dynamic water depth distribution grid model to obtain an irrigation heterogeneity distribution map and an irrigation system fault diagnosis matrix;

[0012] An input module, configured to input the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into a deep reinforcement learning framework for multi-step decision optimization processing under constraint conditions to obtain a regional irrigation control parameter set.

[0013] In a third aspect, a device for measuring the water depth of a paddy field surface based on image intelligent recognition is provided, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory to enable the device for measuring the water depth of a paddy field surface based on image intelligent recognition to execute the above-mentioned method for measuring the water depth of a paddy field surface based on image intelligent recognition.

[0014] In a fourth aspect, a computer-readable storage medium is provided, in which instructions are stored. When it runs on a computer, it enables the computer to execute the above-mentioned method for measuring the water depth of a paddy field surface based on image intelligent recognition.

[0015] In the technical solution provided by this application, by applying the image intelligent recognition technology to the field of measuring the water depth of a paddy field surface, the full-process automation from image acquisition to irrigation control is realized, effectively solving the technical problems of low efficiency and poor accuracy of traditional water depth measurement methods. At the same time, the combination of the deep reinforcement learning framework and Markov decision analysis provides intelligent decision-making support for irrigation optimization. Specifically, the multi-period image acquisition and time-series processing technology ensures the continuity and reliability of the original data, providing a high-quality data basis for subsequent analysis; the introduction of the two-stage feature pyramid network greatly improves the accuracy and robustness of water gauge positioning. Especially in complex lighting conditions and partial occlusion situations, the multi-scale anchor box matching mechanism effectively extracts the water gauge features, accurately obtaining its three-dimensional coordinates and attitude angle data; the instance segmentation method based on the water gauge coordinates solves the problem of water gauge scale recognition, and can accurately read the scale value even under the interference of water surface fluctuations; the method combining the variational adaptive thin plate spline interpolation algorithm and hydraulic simulation correction realizes the accurate conversion from discrete points to continuous water depth distribution, overcoming the limitation that traditional point measurements cannot express the spatial distribution; Markov decision analysis introduces the ability to process time-series data, and through the comprehensive analysis of historical states and current observations, accurately diagnoses irrigation anomalies; the deep reinforcement learning framework adaptively generates the optimal irrigation control strategy through multi-step decision optimization under constraints. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of an embodiment of the method for measuring the water depth of a paddy field surface based on image intelligent recognition in the embodiments of this application;

[0018] Figure 2Schematic diagram of an embodiment of the paddy field water depth measurement system based on image intelligent recognition in the embodiment of the present application;

[0019] Figure 3 It is a structural schematic block diagram of the paddy field water depth measurement device based on image intelligent recognition in the embodiment of the present invention. Detailed implementation manners

[0020] The embodiment of the present application provides a paddy field water depth measurement method and system based on image intelligent recognition. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that shown or described here. In addition, the terms "include" or "have" and any deformation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] For the convenience of understanding, the specific process of the embodiment of the present application will be described below. Please refer to Figure 1 An embodiment of the paddy field water depth measurement method based on image intelligent recognition in the embodiment of the present application includes:

[0022] Step S101: Perform multi-period image acquisition on the paddy field with a water gauge and perform time series processing to obtain time series preprocessed image data;

[0023] Step S102: Input the time series preprocessed image data into a two-stage feature pyramid network for multi-scale anchor box matching to obtain the three-dimensional coordinates and attitude angle data of the water gauge;

[0024] Step S103: Perform instance segmentation based on the three-dimensional coordinates and attitude angle data of the water gauge to obtain the water gauge scale reading and water surface boundary intersection point data;

[0025] Step S104: Combine the water gauge scale reading and water surface boundary intersection point data with the high-density discrete point cloud of the paddy field sampled by actual measurement, and through variational adaptive thin plate spline interpolation and hydraulic simulation correction processing, obtain a dynamic water depth distribution grid model;

[0026] Step S105: Perform Markov decision analysis based on the dynamic water depth distribution grid model to obtain an irrigation heterogeneity distribution map and an irrigation system fault diagnosis matrix;

[0027] Step S106: Input the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into the deep reinforcement learning framework for multi-step decision optimization under constraints to obtain a regional irrigation control parameter set.

[0028] It can be understood that the execution subject of this application can be a field surface water depth measurement system based on image intelligent recognition, or a terminal or a server. Specifically, it is not limited here. In this embodiment of the application, the server is taken as the execution subject for illustration.

[0029] Specifically, multi-period image acquisition is performed on the fields equipped with water gauges. In this process, a camera device fixed at the edge of the field is used to control the shooting frequency to be once per hour to obtain the original image sequence. These images contain information about the water gauge and the water surface, but there are often interference factors such as water surface ripples and weed occlusion. These interferences are eliminated through noise filtering processing to obtain clear field water surface images. Then, histogram equalization processing is performed on these images to balance the brightness differences under different lighting conditions and standardize the image data. Then, perspective correction is performed based on the reference points at the four corners of the field to make the image conform to the actual proportional relationship. Detail enhancement processing is performed on the corrected image to sharpen the edge features of the water gauge scale for subsequent recognition. Finally, the enhanced image is combined with the acquisition time, field number, and geographical location information to form time-series preprocessed image data.

[0030] Input the time-series preprocessed image data into the two-stage feature pyramid network. This network first extracts feature maps of different scales through the first-stage network to generate a coordinate set of water gauge candidate regions. Non-maximum suppression operation is performed on these candidate regions, and a confidence threshold of 0.85 is set to screen out the high-confidence water gauge region bounding boxes. Then, the images cropped within these bounding boxes are input into the second-stage network for fine-grained feature extraction to obtain the fine contour data of the water gauge. Based on these contour data, Hough transform calculation is performed to detect the vertical edge line of the water gauge and obtain the direction angle value of the water gauge in the image plane. These angle values are subjected to mapping transformation with the pre-calibrated reference scale points to establish the correspondence between the image space and the physical space, obtaining the spatial transformation matrix. Finally, combined with the calculation of the image resolution scale factor, through the principle of triangulation, the three-dimensional coordinates and attitude angle data of the water gauge are determined.

[0031] Perform instance segmentation based on the three-dimensional coordinates and attitude angle data of the water gauge. First, perform image enhancement processing on the water gauge image area, adjust the contrast parameter, and obtain a high-contrast water gauge image. Then, perform adaptive binarization processing, set a dynamic threshold to separate the water gauge scale from the background, and generate a binary image of the scale features. Input this image into a convolutional neural network for feature extraction, identify the main scale lines and numbers on the water gauge, and establish the data of the water gauge scale reference system. Based on this parameter system data, perform edge detection operations to locate the boundary line between the water surface and the water gauge, and obtain the coordinate point set of the water surface boundary line. Perform robustness analysis on these coordinate points to eliminate the influence of fluctuations and reflections, and obtain the stable water surface line position data. Finally, perform linear interpolation calculation on the water surface line position data and the water gauge scale reference system data, and determine the water gauge scale reading and the water surface boundary intersection data through the calibration of the water gauge zero elevation. Combine the water gauge scale reading and the water surface boundary intersection data with the high-density discrete point cloud of the field. Preprocess the point cloud of the field obtained by RTK-GPS technology, detect and correct abnormal points through a sliding window, and obtain the terrain feature point set. Input these point sets into the variational adaptive thin plate spline interpolation algorithm, set a resolution parameter of 0.5 m × 0.5 m, and construct a continuous digital elevation model of the field surface. Perform regional analysis on this model to identify special terrain elements such as field ridges and drainage ditches, and form complete terrain structure data. Register the water gauge scale reading and the water surface boundary intersection data with the terrain structure data, establish a unified elevation reference, and obtain the water surface-terrain correspondence. Based on this relationship, perform raster calculation, compare the water level elevation with the field surface elevation, and generate the initial water depth distribution data. Finally, input it into a hydraulic model for flow field analysis, consider the water flow direction and velocity parameters to correct the water surface imbalance factor, and obtain the dynamic water depth distribution grid model.

[0032] Perform Markov decision analysis based on the dynamic water depth distribution grid model. First, compare and analyze the dynamic water depth distribution grid model with the crop irrigation water requirement database, extract the ideal irrigation water depth parameters of the current crop, and obtain the water depth deviation distribution data. Calculate the uniformity index for these data, quantitatively evaluate the consistency of the water depth distribution, and obtain the irrigation uniformity evaluation index. Combine this index with the irrigation deficiency rate and the irrigation excess rate, and obtain the comprehensive irrigation efficiency index through a comprehensive evaluation method. Based on this index, perform spatial clustering analysis to identify the distribution law of the water depth abnormal area, and form the characteristic map of the water depth abnormal area. Input this map into the Markov decision model, and obtain the classification result of the irrigation abnormal cause through pattern matching with the expert knowledge base. Finally, combine the classification result with the characteristic map of the abnormal area, establish the correspondence between the irrigation problem and the regional distribution, and obtain the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix.

[0033] The irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix are input into the deep reinforcement learning framework for optimization. First, the data of the irrigation heterogeneity distribution map and the fault diagnosis matrix are integrated to establish the dynamic relationship of soil moisture in the field, and the irrigation regulation demand data are obtained. These data are input into the multi-objective optimization algorithm, considering the irrigation uniformity, water resource utilization rate and operation cost simultaneously, and the irrigation control candidate schemes are obtained. Based on these candidate schemes, classification processing is carried out to distinguish different types of irrigation problems, and a differentiated irrigation optimization scheme is formed. The priority of these schemes is sorted, considering the implementation difficulty and the expected effect comprehensively, and the irrigation execution task list is obtained. This list is matched with the field characteristic parameters to calculate the optimal irrigation time and flow rate, and the sub-region irrigation parameters are obtained. Finally, the timing arrangement processing is carried out on these parameters to generate the operation instruction sequence and the execution schedule, and the sub-region irrigation control parameter set is obtained.

[0034] For example, for a specific field, first, images are collected by the camera device installed at the edge of the field, and clear water gauge images are obtained after image processing. The dual-stage feature pyramid network identifies the position of the water gauge and determines its three-dimensional coordinates as (X = 23.56 m, Y = 45.78 m, Z = 102.35 m) and the attitude angle of 15.7 degrees. The instance segmentation network identifies the water gauge scale reading as 54.3 cm, and the water surface boundary intersection point is located between the 5th and 6th scales. Combining the point cloud data of the field topography collected by RTK-GPS, a digital elevation model of the field surface with a resolution of 0.5 m × 0.5 m is constructed. After spatial registration and hydraulic simulation, the water depth values at each point in the field are obtained, and a dynamic water depth distribution grid model is formed. Markov decision analysis finds that the water depth in the northwest corner area of the field is significantly lower than the average value, which is diagnosed as a problem of ridge leakage, and an irrigation heterogeneity distribution map and a fault diagnosis matrix are generated. Based on these data, the deep reinforcement learning framework calculates that the irrigation time in the northwest corner area needs to be increased by 25%, and generates a sub-region irrigation control parameter set, including the irrigation flow rate, duration and order of each region, to form an irrigation execution plan.

[0035] In the embodiments of the present application, by applying image intelligent recognition technology to the field of paddy field water depth measurement, the full process automation from image acquisition to irrigation control is realized, effectively solving the technical problems of low efficiency and poor accuracy of traditional water depth measurement methods. At the same time, the combination of the deep reinforcement learning framework and Markov decision analysis provides intelligent decision-making support for irrigation optimization. Specifically, the multi-period image acquisition and time-series processing technology ensures the continuity and reliability of the original data, providing a high-quality data basis for subsequent analysis; the introduction of the two-stage feature pyramid network greatly improves the accuracy and robustness of water gauge positioning. Especially in complex lighting conditions and partial occlusion cases, the water gauge features are effectively extracted through the multi-scale anchor box matching mechanism, and the three-dimensional coordinates and attitude angle data are accurately obtained; the instance segmentation method based on the water gauge coordinates solves the problem of water gauge scale recognition, and can accurately read the scale value even under the interference of water surface fluctuations; the method combining the variational adaptive thin plate spline interpolation algorithm and hydraulic simulation correction realizes the accurate conversion from discrete points to continuous water depth distribution, overcoming the limitation that traditional point measurements cannot express spatial distribution; Markov decision analysis introduces the ability to process time-series data, and through the comprehensive analysis of historical states and current observations, accurately diagnoses irrigation anomalies; the deep reinforcement learning framework adaptively generates the optimal irrigation control strategy through multi-step decision optimization under constraints.

[0036] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0037] Adjust the lighting parameters of the camera device fixedly installed on the edge of the field, and obtain the original image sequence containing the water gauge and the water surface by controlling the shooting frequency to be once per hour;

[0038] Perform noise filtering processing on the original image sequence, and obtain the field water surface image by eliminating the interference of water surface ripples and weeds;

[0039] Input the field water surface image into the histogram equalization processing unit, and obtain the standardized image data by balancing the image brightness under different lighting conditions;

[0040] Perform perspective correction based on the standardized image data, and obtain the orthographic image by using the reference points at the four corners of the field for coordinate transformation;

[0041] Perform detail enhancement processing on the orthographic image, and obtain the enhanced image by sharpening the edge features of the water gauge scale;

[0042] Combine and process the enhanced image with the acquisition time, field number, and geographical location information to obtain the time-series preprocessed image data.

[0043] Specifically, adjust the lighting parameters of the camera device fixedly installed at the edge of the field. By controlling the shooting frequency to once per hour, obtain the original image sequence containing the water gauge and the water surface. Perform noise filtering on the original image sequence. By eliminating the interference of water surface ripples and weeds, obtain the field water surface image. Input the field water surface image into the histogram equalization processing unit. By balancing the image brightness under different lighting conditions, obtain the standardized image data. Based on the standardized image data, perform perspective correction. By using the reference points at the four corners of the field for coordinate transformation, obtain the orthographic image. Perform detail enhancement processing on the orthographic image. By sharpening the edge features of the water gauge scale, obtain the enhanced image. Combine the enhanced image with the acquisition time, field number, and geographical location information to obtain the time-series preprocessed image data.

[0044] Fix the camera device at the best observation position on the edge of the field. Adjust parameters such as the aperture, shutter speed, and ISO sensitivity of the camera device so that it can adapt to different weather and lighting conditions. The camera device uses a high-resolution camera module. Set the shooting frequency to once per hour, and the system automatically records the shooting time. When the lighting conditions change, the camera device will adjust according to the preset automatic exposure parameters to ensure image clarity. The obtained original image sequence is saved as an RGB three-channel color image with a resolution of 3840×2160 pixels.

[0045] For the noise filtering of the original image sequence, use an improved Gaussian filtering algorithm to specifically process interference factors such as the ripple reflection and waterweed occlusion unique to the farmland water surface. During the filtering process, set the kernel size to 5×5 pixels, and the filtering intensity parameter is dynamically adjusted according to the image noise level. This processing can effectively remove the high-frequency interference and random noise generated by the water surface ripples, while retaining the clear contour of the water gauge edge. The water gauge scale in the obtained field water surface image is more clearly distinguishable.

[0046] After inputting the field water surface image into the histogram equalization processing unit, the system analyzes the image brightness distribution, calculates the pixel brightness histogram, and then applies the adaptive histogram equalization algorithm for processing. This algorithm divides the image into multiple small regions for local equalization, and then combines the processing results through bilinear interpolation, effectively solving the problem of uneven image brightness caused by different lighting differences at different times. The processed standardized image data has a more balanced brightness distribution, and images taken on cloudy or sunny days can all present similar brightness characteristics.

[0047] When performing perspective correction based on standardized image data, the system first identifies the reference marker points pre-set at the four corners of the field block, and these marker points have precise geographical coordinate information. The perspective transformation matrix is calculated through these reference points to geometrically correct the image and eliminate the perspective distortion caused by the camera installation angle. In the orthoimage after correction, the actual physical size of the water gauge scale is proportional to the pixel distance in the image, providing an accurate spatial reference for subsequent water depth measurement.

[0048] When performing detail enhancement processing on the orthoimage, a non-linear sharpening algorithm is adopted to focus on enhancing the edge features of the water gauge scale lines. During the processing, selective enhancement is performed on the specific frequency range of the water gauge scale lines, while suppressing background textures and noises. In the enhanced image, the edges of the water gauge scale lines are sharper and the digital markings are clearer, facilitating subsequent scale identification and water level reading. The enhanced image is combined with relevant metadata, adding accurate acquisition timestamps, field block numbers, geographical location coordinates (latitude and longitude), and camera device numbers, etc. to each image to form time-series preprocessed image data with complete marking information. These metadata are stored in the form of image file header information for subsequent data analysis and traceability. The time-series preprocessed image data is input as the basic data of the system, providing a high-quality image source for subsequent water gauge identification and positioning.

[0049] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0050] Input the time-series preprocessed image data into the first-stage network including a backbone network and a feature pyramid structure, and obtain a set of water gauge candidate region coordinates through feature map extraction at different scales;

[0051] Perform non-maximum suppression operation on the set of water gauge candidate region coordinates, and obtain a high-confidence water gauge region bounding box by setting a confidence threshold of 0.85 for screening;

[0052] Crop the image within the high-confidence water gauge region bounding box and input it into the second-stage network, and obtain the fine contour data of the water gauge through fine-grained feature extraction;

[0053] Perform Hough transform calculation based on the fine contour data of the water gauge, and obtain the direction angle value of the water gauge in the image plane by detecting the vertical edge lines of the water gauge;

[0054] Perform mapping transformation on the direction angle value of the water gauge in the image plane and the pre-calibrated reference scale points, and obtain the spatial transformation matrix of the water gauge by establishing the correspondence between the image space and the physical space;

[0055] Combine the calculation of the spatial transformation matrix with the image resolution scale factor, and obtain the three-dimensional coordinates and attitude angle data of the water gauge through the principle of triangulation.

[0056] Specifically, the time-series preprocessed image data is input into the first-stage network including a backbone network and a feature pyramid structure. Through the extraction of feature maps at different scales, a set of coordinates of the water gauge candidate regions is obtained; a non-maximum suppression operation is performed on the set of coordinates of the water gauge candidate regions, and through setting a confidence threshold of 0.85 for screening, a bounding box of the water gauge region with high confidence is obtained; the image within the bounding box of the water gauge region with high confidence is cropped and input into the second-stage network. Through fine-grained feature extraction, the fine contour data of the water gauge is obtained; based on the fine contour data of the water gauge, a Hough transform calculation is performed. By detecting the vertical edge lines of the water gauge, the direction angle value of the water gauge in the image plane is obtained; a mapping transformation is performed on the direction angle value of the water gauge in the image plane and the pre-calibrated reference scale points. By establishing the correspondence between the image space and the physical space, the spatial transformation matrix of the water gauge is obtained; the spatial transformation matrix is combined with the image resolution scale factor and calculated. By the principle of triangulation, the three-dimensional coordinates and attitude angle data of the water gauge are obtained.

[0057] Input the time-series preprocessed image data into the first stage of a two-stage feature pyramid network. This network uses ResNet50 as the backbone network to extract multi-level features of the image, and then fuses feature maps with different resolutions through a feature pyramid structure. The feature pyramid structure includes five levels from P2 to P6, corresponding to the 1 / 4, 1 / 8, 1 / 16, 1 / 32, and 1 / 64 scales of the original image respectively. The network sets anchor boxes of different sizes on these feature maps, and optimizes the design for the slender features of the water gauge, with the aspect ratios set to 1:5, 1:8, and 1:10. Through the region proposal network, the confidence score and the bounding box coordinate offset containing the water gauge are calculated for each anchor box, generating a set of coordinates of the water gauge candidate regions.

[0058] When performing the non-maximum suppression operation on the set of coordinates of the water gauge candidate regions, first arrange all candidate regions in descending order of confidence scores, and then calculate the intersection over union (IoU) value between candidate regions. When the IoU value of two candidate regions exceeds 0.5, the region with higher confidence is retained, and the region with lower confidence is suppressed. This process is iterated until all regions are processed. Finally, a confidence threshold of 0.85 is set to screen out the bounding boxes of the water gauge regions with high confidence, ensuring the accuracy of the detection results.

[0059] After cropping the image within the bounding box of the water gauge region with high confidence, it is input into the second-stage network for fine-grained feature extraction. The second-stage network adopts a convolutional neural network structure with an attention mechanism, which is specifically optimized for the scale features of the water gauge. This network includes a spatial attention module and a channel attention module, which can adaptively focus on the scale lines and digital regions of the water gauge, effectively extracting the fine structure features of the water gauge. The network outputs a pixel-level segmentation mask of the water gauge, forming the fine contour data of the water gauge, and accurately positioning the water gauge edge and scale positions.

[0060] When performing the Hough transform calculation based on the fine contour data of the water gauge, first perform edge detection on the contour data to extract the edge point set. Then map these edge points to the Hough space, and use an accumulator to count the line parameters. Considering the vertical characteristics of the water gauge, set the angle constraint condition to preferentially detect the edge lines close to the vertical direction. Identify the main vertical edge lines through peak detection, and calculate their direction angle values in the image plane to accurately characterize the tilt state of the water gauge.

[0061] When performing the mapping transformation on the direction angle value of the water gauge in the image plane and the pre-calibrated reference scale points, use the pre-marked reference scale points on the water gauge, and these points have known physical space coordinates. By establishing the correspondence between the pixel coordinates and the physical space coordinates of the reference points in the image, calculate the perspective transformation parameters to form the spatial transformation matrix. This matrix describes the transformation relationship from the image space to the physical space. When combining the spatial transformation matrix with the image resolution scale factor, use the known correspondence between the image resolution and the actual physical size to determine the conversion ratio from pixels to physical units. Then, based on the principle of triangulation, combined with the position, direction angle, and spatial transformation relationship of the water gauge in the image, calculate the accurate coordinates and attitude angles of the water gauge in the three-dimensional space. The calculated three-dimensional coordinates of the water gauge are represented in the field coordinate system, and the attitude angles include the inclination angle of the water gauge relative to the vertical direction and the azimuth angle relative to the north direction.

[0062] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0063] Perform image enhancement processing on the water gauge image area according to the three-dimensional coordinates and attitude angle data of the water gauge to obtain a high-contrast water gauge image;

[0064] Perform adaptive binarization processing on the high-contrast water gauge image to obtain a binary image of the scale features;

[0065] Input the binary image of the scale features into a convolutional neural network for feature extraction, and obtain the water gauge scale reference system data by identifying the main scale lines and numbers on the water gauge;

[0066] Perform edge detection operation based on the water gauge scale reference system data, and obtain the coordinate point set of the water surface boundary line by locating the intersection line between the water surface and the water gauge;

[0067] Perform robustness analysis processing on the coordinate point set of the water surface boundary line to obtain the water surface line position data;

[0068] Perform linear interpolation calculation on the water surface line position data and the water gauge scale reference system data, and obtain the water gauge scale reading and the water surface boundary intersection point data through the calibration of the zero elevation of the water gauge.

[0069] Specifically, based on the three-dimensional coordinates and attitude angle data of the water gauge, image enhancement processing is performed on the water gauge image area to obtain a high-contrast water gauge image; adaptive binarization processing is performed on the high-contrast water gauge image to obtain a binary image of scale features; the binary image of scale features is input into a convolutional neural network for feature extraction, and by recognizing the main scale lines and numbers on the water gauge, water gauge scale reference system data is obtained; edge detection operations are performed based on the water gauge scale reference system data, and by locating the boundary line between the water surface and the water gauge, a set of coordinate points of the water surface boundary line is obtained; robust analysis processing is performed on the set of coordinate points of the water surface boundary line to obtain water surface line position data; linear interpolation calculation is performed on the water surface line position data and the water gauge scale reference system data, and through the calibration of the zero elevation of the water gauge, water gauge scale readings and water surface boundary intersection data are obtained. According to the obtained three-dimensional coordinates and attitude angle data of the water gauge, precise positioning and extraction are performed on the water gauge area in the preprocessed time-series image. The extracted water gauge image area often has the problem of insufficient contrast, especially in the case of poor lighting conditions or strong water surface reflection. To address this problem, an adaptive contrast enhancement algorithm is used to process the water gauge image area. This algorithm dynamically adjusts the contrast parameters according to the brightness distribution characteristics of the local area of the image, and focuses on enhancing the contrast between the water gauge scale and the background. During the processing, wavelet decomposition is performed on the water gauge area image, and different frequency components are enhanced separately. In particular, the middle and high frequency components corresponding to the scale lines are strengthened, while the low frequency components corresponding to the background noise are suppressed. Finally, a high-contrast water gauge image is synthesized.

[0070] When performing adaptive binarization processing on the high-contrast water gauge image, considering the uneven lighting conditions and local reflection interference of the water gauge in the image, an adaptive threshold algorithm is used instead of the global threshold method. This algorithm divides the water gauge image into multiple small windows, calculates the optimal local threshold for each window, and then forms a threshold map by smoothly connecting the thresholds of each window. This threshold map is used to perform binarization processing on the water gauge image, effectively separating the water gauge scale lines, digital markings from the background, and obtaining a clear binary image of scale features.

[0071] The binary image of the scale feature is input into a specially trained convolutional neural network for feature extraction. This network adopts a multi-branch structure, including a branch for detecting scale lines and a branch for recognizing digital markers. The scale line detection branch uses a densely connected convolutional network, and the reuse of inter-layer features improves the detection ability for slender scale lines; the digital recognition branch uses a convolutional structure with an attention mechanism to improve the recognition accuracy of small-sized digits. The network establishes a scale reference system by comprehensively analyzing the scale line spacing, arrangement rules, and digital marker positions, identifies the main scale points and their corresponding values, and forms the data of the water gauge scale reference system. Based on the data of the water gauge scale reference system, edge detection operations are performed, focusing on locating the intersection position between the water surface and the water gauge. In the intersection area, the water surface often shows obvious changes in brightness and texture. An improved Canny edge detection algorithm is used to set appropriate gradient thresholds and edge connection parameters to extract the edge contour of the water surface. Combining with the vertical characteristics of the water gauge, the search range is limited, and the edge lines close to the horizontal direction are mainly detected to exclude interfering edges, and finally the intersection position between the water surface and the water gauge is determined, generating a set of coordinate points of the water surface boundary line. Robustness analysis and processing are performed on the set of coordinate points of the water surface boundary line to eliminate false detection problems caused by water surface fluctuations and local occlusions. The RANSAC (Random Sample Consensus) algorithm is used to remove outliers from the coordinate point set and fit the optimal horizontal line. For the case where there are multiple candidate water surface lines, combined with the timing information and the water level data of adjacent frames, the most reasonable water surface line is selected through time continuity constraints. After processing, stable and reliable water surface line position data are obtained, accurately reflecting the water surface position at the current moment.

[0072] Perform linear interpolation calculations on the water surface line position data and the water gauge scale reference system data to determine the accurate water level reading. Based on the identified positions and corresponding values of the main scale points, combined with the relative position relationship between the water surface line and the scale, the accurate scale value corresponding to the water surface line is calculated through the linear interpolation algorithm. At the same time, according to the pre-calibrated zero elevation data of the water gauge, the relative reading is converted into an absolute elevation value to ensure the consistency and comparability of the measurement results within the entire irrigation area. The finally generated water gauge scale reading and water surface boundary intersection data contain complete information such as measurement time, water gauge number, relative water level reading, absolute elevation value, etc.

[0073] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0074] Perform data preprocessing on the high-density discrete point cloud of the field obtained by measurement. Through sliding window outlier detection and correction, a set of terrain feature points is obtained;

[0075] Input the set of terrain feature points into the variational adaptive thin plate spline interpolation algorithm. By setting a resolution parameter of 0.5 m × 0.5 m, a continuous digital elevation model of the field surface is obtained;

[0076] Perform regional analysis and processing on the continuous digital elevation model of the field surface to obtain terrain structure data;

[0077] Perform spatial registration on the water gauge scale readings and the water surface boundary intersection data with the terrain structure data to obtain the water surface-terrain correspondence;

[0078] Based on the water surface-terrain correspondence, perform raster calculations. By comparing the water level elevation with the field surface elevation, obtain the initial water depth distribution data;

[0079] Input the initial water depth distribution data into a hydraulic model for flow field analysis. By considering the water flow direction and velocity parameters to correct the water surface imbalance factors, obtain the dynamic water depth distribution grid model.

[0080] Specifically, perform data preprocessing on the high-density discrete point cloud of the field block obtained through measurement. Through sliding window outlier detection and correction, obtain the terrain feature point set; input the terrain feature point set into the variational adaptive thin plate spline interpolation algorithm. By setting the resolution parameter of 0.5 m × 0.5 m, obtain the continuous digital elevation model of the field surface; perform regional analysis and processing on the continuous digital elevation model of the field surface to obtain terrain structure data; perform spatial registration on the water gauge scale readings and the water surface boundary intersection data with the terrain structure data to obtain the water surface-terrain correspondence; based on the water surface-terrain correspondence, perform raster calculations. By comparing the water level elevation with the field surface elevation, obtain the initial water depth distribution data; input the initial water depth distribution data into a hydraulic model for flow field analysis. By considering the water flow direction and velocity parameters to correct the water surface imbalance factors, obtain the dynamic water depth distribution grid model. Use RTK-GPS technology to perform high-density measurement on the field block to obtain discrete point cloud data containing three-dimensional coordinate information. These data points are evenly distributed within the field block, and each point records accurate longitude, latitude, and elevation values, with the coordinate accuracy reaching the centimeter level. There may be outliers in the collected original point cloud data, such as measurement errors, temporary obstacles, or non-ground target points. To address these issues, a sliding window outlier detection method is used for data preprocessing. This method sets a fixed-size spatial window on the point cloud data, with the window size of 3 m × 3 m. Calculate the statistical characteristics of the elevation values within the window, including the mean, median, and standard deviation. When the elevation value of a certain point deviates from the median within the window by more than three standard deviations, it is determined as an outlier, and it is corrected or removed through the interpolation method of adjacent points. This process is iteratively performed on the entire point cloud data, and finally a clean and accurate terrain feature point set is obtained.

[0081] The topographic feature point set is input into the variational adaptive thin plate spline interpolation algorithm for processing to construct a continuous digital elevation model of the field surface. This algorithm is based on the principle of minimum curvature and uses the thin plate spline function to fit the discrete point cloud data to generate a smooth and continuous topographic surface. A variational adaptive mechanism is introduced in the algorithm to increase the control point density in areas with large topographic changes to improve the interpolation accuracy, and appropriately reduce the control points in flat terrain areas to improve the calculation efficiency. The interpolation process sets a fixed grid resolution of 0.5 m × 0.5 m, that is, each 25 square meter area contains an elevation sample point to ensure that the micro-topographic change characteristics of the field surface can be reflected. After the interpolation calculation is completed, a continuous digital elevation model covering the entire field block is formed, and each grid cell records the accurate elevation value. Regional analysis and processing are carried out on the continuous digital elevation model of the field surface to identify the key topographic structures within the field block. The slope value and aspect value of each grid cell are calculated through the slope analysis algorithm to identify areas with obvious topographic changes. Combining the gradient threshold segmentation method, linear features such as the field block boundary, ridge line, and drainage ditch are extracted. These features often show areas of abrupt elevation change. At the same time, the elevation model is partitioned based on the region growing algorithm, and the field block is divided into several sub-regions with similar elevation characteristics. Through these analysis and processing, a complete topographic structure data containing information such as the field block boundary, elevation partition, and topographic structure line is formed, providing a topographic basis for water depth calculation.

[0082] The water gauge scale readings and the water surface boundary intersection point data are spatially registered with the topographic structure data to establish a unified spatial reference system. First, based on the GPS coordinate system or the local coordinate system, the location of the water gauge is geographically registered with the digital elevation model of the field block to determine the exact location of the water gauge in the digital elevation model. Then, according to the absolute elevation reading of the water gauge and the water surface boundary intersection point data, the absolute elevation value of the current water surface is determined. Considering that the water surface may not be completely horizontal during the irrigation process, based on the readings of multiple water gauges within the field block, a water surface elevation model is constructed through spatial interpolation methods to accurately describe the water surface morphology within the entire field block range. Finally, the corresponding relationship between the water surface elevation and the topographic elevation in the same coordinate system is formed, that is, the water surface-topography correspondence.

[0083] Grid calculations are performed based on the water surface-topography correspondence to calculate the water depth value of each grid cell within the field block. For each grid cell in the digital elevation model, its topographic elevation value is obtained, and the water surface elevation value at the same location is extracted from the water surface elevation model. The water depth value at this point is calculated by taking the difference between the two, generating water depth distribution data covering the entire field block. The calculation formula is: water depth = water surface elevation - topographic elevation. When the calculation result is positive, it indicates that the point is covered by water, and the value is the water depth; when the calculation result is negative, it indicates that the point is above the water surface and is not covered by water, and the water depth is recorded as zero. In this way, the initial water depth distribution data is formed, intuitively reflecting the water distribution situation within the field block.

[0084] The initial water depth distribution data is input into the hydraulic model for flow field analysis, considering the influence of the hydrodynamic characteristics of the water flow on the water depth distribution. The shallow water equations are used to describe the movement law of the irrigation water flow, considering the influence of factors such as gravity, friction force, and inertia force on the water flow velocity and water surface morphology. The model input parameters include the field slope, soil roughness, inlet flow rate, and outlet conditions, etc. The water flow field distribution is calculated by the numerical solution method to obtain the velocity vector and water surface slope of each grid cell. Based on the calculation results, the initial water depth distribution data is corrected, considering the water surface imbalance factors caused by the hydrodynamic characteristics of the water flow, such as the water level elevation near the inlet and the water surface lowering in the area with a large flow velocity, etc. After correction, a dynamic water depth distribution grid model is formed.

[0085] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0086] The dynamic water depth distribution grid model is compared and analyzed with the crop irrigation water requirement database, and by extracting the ideal irrigation water depth parameters of the current crop, the water depth deviation distribution data is obtained;

[0087] The uniformity index of the water depth deviation distribution data is calculated, and by quantitatively evaluating the consistency of the water depth distribution, the irrigation uniformity evaluation index is obtained;

[0088] The irrigation uniformity evaluation index is combined with the irrigation deficiency rate and the irrigation excess rate, and through the comprehensive evaluation method, the comprehensive irrigation efficiency index is obtained;

[0089] Based on the comprehensive irrigation efficiency index, spatial clustering analysis is carried out, and by identifying the distribution law of the water depth abnormal area, the characteristic map of the water depth abnormal area is obtained;

[0090] The characteristic map of the water depth abnormal area is input into the Markov decision model, and by matching with the expert knowledge base mode, the classification result of the irrigation abnormal cause is obtained;

[0091] The classification result of the irrigation abnormal cause is combined with the characteristic map of the water depth abnormal area, and by establishing the corresponding relationship between the irrigation problem and the regional distribution, the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix are obtained.

[0092] Specifically, a comparative analysis is conducted between the dynamic water depth distribution grid model and the crop irrigation water demand database. By extracting the ideal irrigation water depth parameters of the current crop, the water depth deviation distribution data is obtained. The uniformity index is calculated for the water depth deviation distribution data, and by quantitatively evaluating the consistency of the water depth distribution, the irrigation uniformity evaluation index is obtained. The irrigation uniformity evaluation index is combined with the irrigation deficiency rate and the over-irrigation rate, and through a comprehensive evaluation method, the comprehensive irrigation efficiency index is obtained. Based on the comprehensive irrigation efficiency index, a spatial clustering analysis is carried out, and by identifying the distribution law of the water depth abnormal area, the characteristic map of the water depth abnormal area is obtained. The characteristic map of the water depth abnormal area is input into the Markov decision model, and by matching with the expert knowledge base model, the classification result of the irrigation abnormal cause is obtained. The classification result of the irrigation abnormal cause is combined with the characteristic map of the water depth abnormal area, and by establishing the corresponding relationship between the irrigation problem and the regional distribution, the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix are obtained. A comparative analysis is conducted between the dynamic water depth distribution grid model and the crop irrigation water demand database. The crop irrigation water demand database contains the ideal irrigation water depth parameters of different crops at each growth stage, and these parameters are comprehensively determined based on crop type, growth stage, soil type, and climate conditions. The system extracts the corresponding ideal irrigation water depth value from the database according to the current crop type and growth stage planted in the field. The ideal value is compared point by point with the actual water depth value in the dynamic water depth distribution grid model, and the water depth deviation value of each grid cell is calculated. A positive deviation value indicates over-irrigation, and a negative deviation value indicates irrigation deficiency. In this way, the water depth deviation distribution data covering the entire field is generated, intuitively reflecting the irrigation imbalance situation.

[0093] The uniformity index is calculated for the water depth deviation distribution data to quantitatively evaluate the irrigation uniformity. The standard Cristal uniformity coefficient (CU) evaluation index is used, which reflects the consistency degree of the irrigation water depth distribution. In the calculation process, first, the water depth values of all effective grid cells in the field are obtained, the average water depth value is calculated, then the absolute deviation of each grid cell's water depth value from the average value is calculated, and finally, the uniformity coefficient is calculated based on these deviation values. The closer the uniformity coefficient is to 100%, the better the irrigation uniformity. In addition, statistical indicators such as the water depth standard deviation and the coefficient of variation are also calculated to comprehensively evaluate the characteristics of the irrigation water depth distribution, forming an irrigation uniformity evaluation index set.

[0094] The evaluation index of irrigation uniformity is combined with the irrigation deficiency rate and the irrigation excess rate for comprehensive efficiency evaluation. The irrigation deficiency rate calculates the percentage of the area where the water depth in the field block is lower than the ideal value in the total area, reflecting the irrigation deficiency situation. The irrigation excess rate calculates the percentage of the area where the water depth exceeds the ideal value in the total area, reflecting the irrigation excess situation. The weighted average method is used to combine the uniformity coefficient, the irrigation deficiency rate, and the irrigation excess rate to calculate the comprehensive irrigation efficiency index. The weight parameters are determined according to the sensitivity of the crop to water stress and the local water resource situation. Usually, the weight of the irrigation deficiency rate is higher than that of the irrigation excess rate. The comprehensive irrigation efficiency index is expressed as a percentage from 0 to 100, and the higher the value, the better the irrigation efficiency. Based on the comprehensive irrigation efficiency index, spatial clustering analysis is carried out to identify irrigation anomaly areas. The density clustering algorithm is used to process the water depth deviation distribution data, and the grid cells with similar deviation characteristics and adjacent in space are grouped into one class to form several irrigation anomaly areas. This algorithm does not preset the number of clusters, but automatically determines the number and range of clusters according to the density distribution characteristics of the data. For each identified anomaly area, characteristic parameters such as its area, average deviation value, and deviation direction (excess or deficiency) are calculated to construct the characteristic vector of the anomaly area. At the same time, the spatial distribution pattern of the anomaly area is analyzed, such as whether it shows a strip-shaped, block-shaped, or gradient-shaped distribution, and whether it is related to the terrain characteristics of the field block, to form the characteristic map of the water depth anomaly area.

[0095] The characteristic map of the water depth anomaly area is input into the Markov decision model for diagnosing the reasons for irrigation anomalies. This model is constructed based on an expert knowledge base, which contains common irrigation anomaly problems and their characteristic patterns, such as the typical manifestations of problems like uneven terrain, channel leakage, and uneven water flow distribution. The model adopts the framework of the Markov decision process, represents the state of the irrigation system as the characteristic vector of the anomaly area, and identifies the most likely anomaly reason by calculating the similarity with the patterns in the knowledge base. For complex situations, the model can perform multi-step reasoning, considering the comprehensive influence of multiple factors to improve the diagnostic accuracy. The diagnostic results include the type of anomaly reason, the confidence score, and the problem severity rating, forming the classification result of the irrigation anomaly reason.

[0096] The classification result of the irrigation anomaly reason is combined with the characteristic map of the water depth anomaly area to establish the corresponding relationship between the problem and the area. First, on the spatial expression of the characteristic map of the water depth anomaly area, each anomaly area is labeled with its corresponding anomaly reason type to form a problem distribution map. Then, an association matrix between the anomaly type and the area characteristics is constructed. The rows of the matrix represent different anomaly reason types, the columns represent different anomaly areas, and the cell values represent the confidence that this area belongs to this anomaly type. In addition, for each anomaly type, typical areas are extracted as representative examples, and their characteristic parameters and spatial ranges are recorded. Through these processes, an irrigation heterogeneity distribution map and an irrigation system fault diagnosis matrix are formed, intuitively reflecting the spatial distribution law of the irrigation problems in the field block and the results of the cause analysis.

[0097] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0098] Integrate the data of the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix, and obtain the irrigation regulation demand data by establishing the dynamic relationship of the soil moisture in the fields.

[0099] Input the irrigation regulation demand data into the multi-objective optimization algorithm, and obtain the candidate irrigation control schemes by simultaneously considering the irrigation uniformity, water resource utilization rate, and operation cost.

[0100] Conduct scheme classification processing based on the candidate irrigation control schemes, and obtain the differentiated irrigation optimization schemes by distinguishing different types of irrigation problems.

[0101] Rank the priorities of the differentiated irrigation optimization schemes, and obtain the irrigation execution task list by comprehensively considering the implementation difficulty and expected effect.

[0102] Match the irrigation execution task list with the field characteristic parameters, and obtain the sub-region irrigation parameters by calculating the optimal irrigation time and flow rate.

[0103] Conduct timing arrangement processing on the sub-region irrigation parameters, and obtain the sub-region irrigation control parameter set by generating the operation instruction sequence and execution schedule.

[0104] Specifically, integrate the data of the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix, and obtain the irrigation regulation demand data by establishing the dynamic relationship of the soil moisture in the fields; input the irrigation regulation demand data into the multi-objective optimization algorithm, and obtain the candidate irrigation control schemes by simultaneously considering the irrigation uniformity, water resource utilization rate, and operation cost; conduct scheme classification processing based on the candidate irrigation control schemes, and obtain the differentiated irrigation optimization schemes by distinguishing different types of irrigation problems; rank the priorities of the differentiated irrigation optimization schemes, and obtain the irrigation execution task list by comprehensively considering the implementation difficulty and expected effect; match the irrigation execution task list with the field characteristic parameters, and obtain the sub-region irrigation parameters by calculating the optimal irrigation time and flow rate; conduct timing arrangement processing on the sub-region irrigation parameters, and obtain the sub-region irrigation control parameter set by generating the operation instruction sequence and execution schedule.

[0105] In the specific implementation process, first, data integration is carried out on the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix to establish a dynamic relationship model of field moisture. During the integration process, the spatial information in the irrigation heterogeneity distribution map is associated with the problem type information in the fault diagnosis matrix to form a three-dimensional associated data structure of problem-region-cause. For each type of irrigation problem, key parameters such as its influence range, severity, and duration are extracted to quantitatively evaluate its impact on crop growth. At the same time, combined with the field soil type, crop water demand characteristics, and current growth stage, the ideal water replenishment amount and replenishment rate of each region are calculated. The integration results form structured irrigation regulation demand data, including spatial partition information, problem types, regulation target values, and time window constraints. The irrigation regulation demand data is input into a multi-objective optimization algorithm to optimize the irrigation control scheme. The non-dominated sorting genetic algorithm (NSGA-II) is used as the optimization framework, and three objective functions are considered simultaneously: maximizing irrigation uniformity, maximizing water resource utilization rate, and minimizing operation cost. The decision variables set in the algorithm include the flow distribution ratio of each water inlet, the total irrigation duration, the phased irrigation time sequence arrangement, and the treatment method for special regions. The constraint conditions include total water volume limit, irrigation time window, equipment capacity limit, etc. Through multi-generation evolutionary calculation, the algorithm generates a series of non-dominated solutions, that is, the Pareto optimal solution set, and each solution represents a feasible irrigation control candidate scheme. These schemes have different focuses and form an optimal balance surface of uniformity-efficiency-cost.

[0106] Based on the irrigation control candidate schemes, scheme classification processing is carried out, and differentiated strategies are formulated for different types of irrigation problems. First, the candidate schemes are divided into several categories according to the reasons for irrigation anomalies: terrain unevenness category, uneven water flow distribution category, ridge leakage category, and time control category. For problems caused by uneven terrain, select schemes that can compensate for the impact of height differences, such as adjusting the water inlet flow distribution and extending the drainage time in low-lying areas; for problems of uneven water flow distribution, focus on optimizing the flow distribution ratio of water inlets and controlling the time sequence; for ridge leakage problems, adopt a zoned irrigation strategy to control the water level difference between adjacent regions. Through this classification processing, differentiated irrigation optimization schemes for the specific problems of the current field are screened out from the candidate scheme set.

[0107] Priority ranking is carried out on the differentiated irrigation optimization schemes to determine the implementation order. Four factors are comprehensively considered in the ranking process: the difficulty of scheme implementation, the expected improvement effect, resource consumption, and time urgency. Quantitative scores are given to these four factors for each scheme, and then the comprehensive priority score is calculated by weighted summation. The weight parameters are dynamically adjusted according to the sensitivity of the current crop to water stress and the water resource supply situation. The ranking results form a hierarchical irrigation execution task list, and each task item in the list includes information such as the operation object, operation type, control parameter range, and execution time window, and is arranged from high to low according to the priority, providing guidance for specific implementation.

[0108] Match the irrigation execution task list with the field characteristic parameters to calculate accurate irrigation control parameters. The field characteristic parameters include soil infiltration characteristics, terrain slope, crop canopy coverage, etc. For each task item, combine the characteristic parameters of its action area to perform fine irrigation parameter calculation. For the irrigation flow rate, consider the soil infiltration rate and the risk of surface runoff to determine the optimal irrigation intensity per unit area; for the irrigation time, combine the soil water storage capacity and the water depth requirement to calculate the accurate duration required to reach the target water depth. These calculation results form sub-region irrigation parameters, and each region is equipped with independent irrigation flow rate values and duration values.

[0109] Perform a timing arrangement process on the sub-region irrigation parameters to form an execution plan. Considering the hydraulic condition limitations of the irrigation system, such as pipeline network pressure changes, total flow rate constraints, etc., the irrigation operations in different regions need to be reasonably sequenced and combined. Use a time series optimization algorithm to sort the irrigation tasks in each region according to the priority and hydraulic condition constraints to generate an optimal operation timing arrangement. At the same time, for each operation step, generate detailed control instructions, including valve opening, pump station flow rate, duration, and switching conditions, etc. These instructions are organized into an operation instruction sequence according to the execution order and are equipped with a detailed execution schedule to form a sub-region irrigation control parameter set, providing direct input for the automatic control of the irrigation system.

[0110] It should be noted that when matching the irrigation execution task list with the field characteristic parameters, the irrigation execution task list can also be input into a computer-aided decision-making system. By simulating the water infiltration curve under different parameter combinations, a sub-region irrigation control plan for the field can be obtained; perform soil water content prediction and analysis on the sub-region irrigation control plan for the field. By considering the soil infiltration characteristics and the crop water absorption dynamics, obtain the optimized irrigation parameters for the region; match and calculate the optimized irrigation parameters for the region with the hydraulic condition constraints. By solving the flow balance equation, obtain the water distribution model of the irrigation system; perform water volume distribution calculation based on the water distribution model of the irrigation system. By minimizing the energy consumption loss function, obtain the flow rate values of each control unit; perform field verification simulation on the flow rate values of each control unit. By calculating the water uniformity index, obtain the flow compensation coefficient; combine the flow compensation coefficient with the flow rate values of the control units. By generating a time period division irrigation sequence, obtain the sub-region irrigation parameters.

[0111] Specifically, the irrigation execution task list is input into the computer-aided decision-making system for processing. This system is constructed based on physical models and includes three sub-modules: moisture movement, heat transfer, and crop growth. Information such as the operation object, time window, and target water depth in the irrigation execution task list is parsed by the system and converted into model input parameters. The system uses the finite element method to simulate the soil-plant-atmosphere continuum under different combinations of irrigation parameters and calculates the infiltration curve of moisture in the vertical profile. Multiple boundary conditions are considered in the simulation calculation, including surface irrigation intensity, bottom drainage conditions, and lateral moisture movement. By comparing the moisture infiltration performance under different irrigation parameters, the system screens out the parameter combination that can meet the water demand of the crop and has the highest water use efficiency, forming an irrigation control plan for sub-regions of the field. When predicting and analyzing the soil water content for the irrigation control plan of the field sub-regions, the system combines the soil texture and structure characteristics of the field and applies the Richards equation to simulate the movement process of moisture in the porous medium. In the prediction and analysis, the soil profile is divided into multiple calculation layers, and different hydraulic parameters are set for each layer, including saturated hydraulic conductivity, water holding capacity, and water characteristic curve. At the same time, considering the differences in crop root distribution characteristics and growth stages, the root water uptake in each layer of soil is calculated. The system evaluates the water use efficiency of different irrigation schemes based on the predicted changes in soil water content distribution, optimizes the irrigation intensity and duration parameters, and obtains the regional optimized irrigation parameters. When matching the regional optimized irrigation parameters with the hydraulic condition constraints, the system establishes a field irrigation water network model, including components such as inlet channels, water distribution outlets, and field ditch systems. In the matching process, the Bernoulli equation and the continuity equation are used to describe the flow movement law in the channels, and the head loss and velocity distribution under different flow rate conditions are calculated. The system considers physical parameters such as the channel cross-sectional shape, roughness coefficient, and slope, solves the non-linear flow balance equations, and determines the hydraulic conditions that meet the irrigation requirements of each sub-region. Through repeated iterative calculations, an irrigation system water distribution model that balances the hydraulic conditions of each region is obtained. When calculating the water volume distribution based on the irrigation system water distribution model, the system sets a multi-objective optimization function and takes minimizing the energy consumption loss as the main optimization goal. The energy consumption loss function considers factors such as head loss, pipe friction loss, and local head loss, and comprehensively evaluates the operation efficiency of the irrigation system. The optimization calculation uses the Lagrange multiplier method combined with the gradient descent algorithm to solve the optimal water volume distribution scheme under the constraint condition of meeting the minimum flow demand of each region. The calculation results give the accurate flow rate values for each control unit, which not only meet the crop irrigation demand but also minimize the energy consumption. When conducting field verification simulations on the flow rate values of each control unit, the system uses numerical simulation methods to predict the water distribution uniformity under actual irrigation conditions. In the verification simulation, actual factors such as the micro-topography change of the field, the spatial heterogeneity of soil permeability, and the water flow boundary effect are considered, and the two-dimensional diffusion wave equation is used to describe the surface water flow movement process.The system calculates moisture uniformity indices such as the Crystal uniformity coefficient and coefficient of variation to evaluate the deviation between the actual irrigation effect and the theoretical model. Based on the deviation analysis results, the flow compensation coefficient for each control unit is calculated to adjust the theoretical flow value and improve the actual irrigation uniformity. When combining the flow compensation coefficient with the control unit flow value, the system applies the corresponding compensation coefficient to the theoretical flow value of each control unit to obtain the corrected actual control flow. At the same time, the system divides the entire irrigation process into multiple time periods, and arranges different irrigation intensities and durations according to the irrigation demand characteristics at different growth stages. The system generates a detailed time-period division irrigation sequence, including the opening time, closing time, and flow setting value of each control unit in each time period. These parameter combinations form the sub-region irrigation parameters.

[0112] The above describes the method for measuring the water depth on the field surface based on image intelligent recognition in the embodiments of the present application. Next, the system for measuring the water depth on the field surface based on image intelligent recognition in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the system for measuring the water depth on the field surface based on image intelligent recognition in the embodiments of the present application includes:

[0113] A processing module 201, configured to perform multi-time period image acquisition on a field block provided with a water gauge and perform time series processing to obtain time series preprocessed image data;

[0114] A matching module 202, configured to input the time series preprocessed image data into a two-stage feature pyramid network for multi-scale anchor box matching to obtain water gauge three-dimensional coordinate and attitude angle data;

[0115] A segmentation module 203, configured to perform instance segmentation based on the water gauge three-dimensional coordinate and attitude angle data to obtain water gauge scale reading and water surface boundary intersection point data;

[0116] A calibration module 204, configured to combine the water gauge scale reading and water surface boundary intersection point data with the high-density discrete point cloud of the measured field block, and through variational adaptive thin plate spline interpolation and hydraulic simulation calibration processing, obtain a dynamic water depth distribution grid model;

[0117] An analysis module 205, configured to perform Markov decision analysis based on the dynamic water depth distribution grid model to obtain an irrigation heterogeneity distribution map and an irrigation system fault diagnosis matrix;

[0118] An input module 206, configured to input the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into a deep reinforcement learning framework for multi-step decision optimization processing under constraint conditions to obtain a sub-region irrigation control parameter set.

[0119] Through the collaborative cooperation of the above-mentioned various components, by applying image intelligent recognition technology to the field of paddy field water depth measurement, the full process automation from image acquisition to irrigation control is realized, effectively solving the technical problems of low efficiency and poor accuracy of traditional water depth measurement methods. At the same time, the combination of the deep reinforcement learning framework and Markov decision analysis provides intelligent decision-making support for irrigation optimization. Specifically, the multi-period image acquisition and time-series processing technology ensure the continuity and reliability of the original data, providing a high-quality data basis for subsequent analysis; the introduction of the two-stage feature pyramid network greatly improves the accuracy and robustness of water gauge positioning. Especially in complex lighting conditions and partial occlusion situations, the water gauge features are effectively extracted through the multi-scale anchor box matching mechanism, and its three-dimensional coordinates and attitude angle data are accurately obtained; the instance segmentation method based on the water gauge coordinates solves the problem of water gauge scale recognition, and the scale value can be accurately read even under the interference of water surface fluctuations; the method combining the variational adaptive thin plate spline interpolation algorithm and hydraulic simulation correction realizes the accurate conversion from discrete points to continuous water depth distribution, overcoming the limitation that traditional point measurements cannot express spatial distribution; Markov decision analysis introduces the ability to process time-series data, and through the comprehensive analysis of historical states and current observations, irrigation anomalies are accurately diagnosed; the deep reinforcement learning framework adaptively generates the optimal irrigation control strategy through multi-step decision optimization under constraints.

[0120] Above Figure 2 From the perspective of modular functional entities, the paddy field water depth measurement system based on image intelligent recognition in the embodiments of the present invention is described in detail. Next, the paddy field water depth measurement device based on image intelligent recognition in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0121] Figure 3FIG. 0 is a schematic structural diagram of a device for measuring paddy field surface water depth based on image intelligent recognition provided by an embodiment of the present invention. The device 300 for measuring paddy field surface water depth based on image intelligent recognition may vary greatly due to configuration or performance differences, and may include one or more processors (central processing units, CPUs) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 for storing application programs 333 or data 332 (for example, one or more mass storage device terminals). Among them, the memory 320 and the storage media 330 may be transient storage or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the device 300 for measuring paddy field surface water depth based on image intelligent recognition. Further, the processor 310 may be configured to communicate with the storage media 330 and execute a series of instruction operations in the storage media 330 on the device 300 for measuring paddy field surface water depth based on image intelligent recognition to implement the steps of the method for measuring paddy field surface water depth based on image intelligent recognition described above.

[0122] The device 300 for measuring paddy field surface water depth based on image intelligent recognition may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art can understand that Figure 3 the shown structural diagram of the device for measuring paddy field surface water depth based on image intelligent recognition does not limit the device for measuring paddy field surface water depth provided by the present invention, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0123] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the method for measuring paddy field surface water depth based on image intelligent recognition.

[0124] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units may refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0125] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing an image intelligent recognition-based field surface water depth measurement device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. A method for measuring the water depth of the field surface based on image intelligent recognition, characterized in that, The method includes: Performing multi - temporal image acquisition on the field with water level gauges and conducting temporal processing to obtain temporally pre - processed image data; Inputting the temporally pre - processed image data into a two - stage feature pyramid network for multi - scale anchor box matching to obtain the three - dimensional coordinates and attitude angle data of the water level gauges; Performing instance segmentation based on the three - dimensional coordinates and attitude angle data of the water level gauges to obtain the water level scale reading and water surface boundary intersection data; Combining the water level scale reading and water surface boundary intersection data with the high - density discrete point cloud of the actually measured sampled field, and through variational adaptive thin - plate spline interpolation and hydraulic simulation correction processing, obtaining a dynamic water depth distribution grid model; Performing Markov decision analysis based on the dynamic water depth distribution grid model to obtain an irrigation heterogeneity distribution map and an irrigation system fault diagnosis matrix; Inputting the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into a deep reinforcement learning framework for multi - step decision - making optimization processing under constraints to obtain a sub - region irrigation control parameter set.

2. The method for measuring the water depth of the field surface based on image intelligent recognition according to claim 1, wherein The performing multi - temporal image acquisition on the field with water level gauges and conducting temporal processing to obtain temporally pre - processed image data includes: Adjusting the lighting parameters of the camera device fixedly installed on the edge of the field, and by controlling the shooting frequency to once per hour, obtaining an original image sequence containing the water level gauges and the water surface; Performing noise filtering processing on the original image sequence, and by eliminating the interference of water surface ripples and weeds, obtaining the water surface image of the field; Inputting the water surface image of the field into a histogram equalization processing unit, and by balancing the image brightness under different lighting conditions, obtaining standardized image data; Performing perspective correction based on the standardized image data, and by using the reference point coordinates at the four corners of the field for coordinate transformation, obtaining an orthographic image; Performing detail enhancement processing on the orthographic image, and by sharpening the edge features of the water level scale, obtaining an enhanced image; Combining the enhanced image with the acquisition time, field number, and geographical location information for combined processing to obtain temporally pre - processed image data.

3. The method for measuring the water depth of the field surface based on image intelligent recognition according to claim 1, wherein The inputting the temporally pre - processed image data into a two - stage feature pyramid network for multi - scale anchor box matching to obtain the three - dimensional coordinates and attitude angle data of the water level gauges includes: Inputting the temporally pre - processed image data into the first - stage network including a backbone network and a feature pyramid structure, and through the extraction of feature maps at different scales, obtaining a set of coordinates of candidate regions of the water level gauges; Performing non - maximum suppression operation on the set of coordinates of candidate regions of the water level gauges, and by setting a confidence threshold of 0.85 for screening, obtaining a high - confidence water level gauge region bounding box; Cropping the image within the high - confidence water level gauge region bounding box and inputting it into the second - stage network, and through fine - grained feature extraction, obtaining the fine - contour data of the water level gauges; Performing Hough transform calculation based on the fine - contour data of the water level gauges, and by detecting the vertical edge lines of the water level gauges, obtaining the direction angle value of the water level gauges in the image plane; Performing mapping transformation on the direction angle value of the water level gauges in the image plane and the pre - calibrated reference scale points, and by establishing the correspondence between the image space and the physical space, obtaining the spatial transformation matrix of the water level gauges; Calculate by combining the spatial transformation matrix with the image resolution scale factor, and obtain the three-dimensional coordinates and attitude angle data of the water gauge through the principle of triangulation.

4. The method for measuring the water depth of the field surface based on image intelligent recognition according to claim 1, wherein, Perform instance segmentation based on the three-dimensional coordinates and attitude angle data of the water gauge to obtain the water gauge scale reading and water surface boundary intersection data, including: Perform image enhancement processing on the water gauge image area according to the three-dimensional coordinates and attitude angle data of the water gauge to obtain a high-contrast water gauge image; Perform adaptive binarization processing on the high-contrast water gauge image to obtain a binary image of scale features; Input the binary image of scale features into a convolutional neural network for feature extraction, and obtain the water gauge scale reference system data by identifying the main scale lines and numbers on the water gauge; Perform edge detection operation based on the water gauge scale reference system data, and obtain the coordinate point set of the water surface boundary line by locating the boundary line between the water surface and the water gauge; Perform robustness analysis processing on the coordinate point set of the water surface boundary line to obtain the water surface line position data; Perform linear interpolation calculation on the water surface line position data and the water gauge scale reference system data, and obtain the water gauge scale reading and water surface boundary intersection data through water gauge zero elevation calibration.

5. The method for measuring the water depth of the field surface based on image intelligent recognition according to claim 1, characterized in that, Combine the water gauge scale reading and water surface boundary intersection data with the high-density discrete point cloud of the measured field block, and obtain the dynamic water depth distribution grid model through variational adaptive thin plate spline interpolation and hydraulic simulation correction processing, including: Perform data preprocessing on the high-density discrete point cloud of the field block obtained by measurement, and obtain the terrain feature point set through sliding window outlier detection and correction; Input the terrain feature point set into the variational adaptive thin plate spline interpolation algorithm, and obtain the continuous digital elevation model of the field surface by setting the resolution parameter of 0.5 m × 0.5 m; Perform regional analysis processing on the continuous digital elevation model of the field surface to obtain the terrain structure data; Perform spatial registration on the water gauge scale reading and water surface boundary intersection data and the terrain structure data to obtain the water surface-terrain correspondence; Perform grid calculation based on the water surface-terrain correspondence, and obtain the initial water depth distribution data by comparing the water level elevation with the field surface elevation; Input the initial water depth distribution data into a hydraulic model for flow field analysis, and obtain the dynamic water depth distribution grid model by considering the water flow direction and velocity parameters to correct the water surface imbalance factor.

6. The method for measuring the water depth of the field surface based on image intelligent recognition according to claim 1, wherein Perform Markov decision analysis based on the dynamic water depth distribution grid model to obtain the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix, including: Perform comparative analysis on the dynamic water depth distribution grid model and the crop irrigation water demand database, and obtain the water depth deviation distribution data by extracting the ideal irrigation water depth parameters of the current crop; Perform uniformity index calculation on the water depth deviation distribution data, and obtain the irrigation uniformity evaluation index by quantitatively evaluating the consistency of the water depth distribution; Combine the irrigation uniformity evaluation index with the irrigation deficiency rate and the irrigation excess rate for processing, and obtain the comprehensive irrigation efficiency index through a comprehensive evaluation method; Perform spatial clustering analysis based on the comprehensive irrigation efficiency index, and obtain the characteristic map of the water depth anomaly area by identifying the distribution law of the water depth anomaly area; Input the characteristic spectrum of the abnormal water depth area into the Markov decision model, and obtain the classification result of the irrigation abnormal cause through pattern matching with the expert knowledge base; Combine and process the classification result of the irrigation abnormal cause and the characteristic spectrum of the abnormal water depth area, and obtain the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix by establishing the corresponding relationship between the irrigation problem and the regional distribution.

7. The method for measuring the water depth of the paddy field surface based on image intelligent recognition according to claim 1, characterized in that Input the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into the deep reinforcement learning framework for multi-step decision optimization under constraints, and obtain the sub-region irrigation control parameter set, including: Integrate the data of the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix, and obtain the irrigation regulation demand data by establishing the dynamic relationship of the field moisture. Input the irrigation regulation demand data into the multi-objective optimization algorithm, and obtain the candidate irrigation control schemes by simultaneously considering the irrigation uniformity, water resource utilization rate and operation cost. Perform scheme classification processing based on the candidate irrigation control schemes, and obtain the differentiated irrigation optimization schemes by distinguishing different types of irrigation problems. Sort the differentiated irrigation optimization schemes by priority, and obtain the irrigation execution task list by comprehensively considering the implementation difficulty and the expected effect. Match the irrigation execution task list with the field characteristic parameters, and obtain the sub-region irrigation parameters by calculating the best irrigation time and flow rate. Perform timing arrangement processing on the sub-region irrigation parameters, and obtain the sub-region irrigation control parameter set by generating the operation instruction sequence and the execution schedule.

8. A system for measuring the water depth of the field surface based on image intelligent recognition, characterized in that For implementing the method for measuring the water depth of the field surface based on image intelligent recognition as described in any one of claims 1-7, the system for measuring the water depth of the field surface based on image intelligent recognition includes: A processing module, configured to perform multi-period image acquisition on the field block provided with a water gauge and perform timing processing to obtain timing preprocessed image data; A matching module, configured to input the timing preprocessed image data into a two-stage feature pyramid network for multi-scale anchor box matching to obtain the three-dimensional coordinates and attitude angle data of the water gauge; A segmentation module, configured to perform instance segmentation based on the three-dimensional coordinates and attitude angle data of the water gauge to obtain the water gauge scale reading and the water surface boundary intersection point data; A calibration module, configured to combine the water gauge scale reading and the water surface boundary intersection point data with the high-density discrete point cloud of the measured field block, and obtain the dynamic water depth distribution grid model through variational adaptive thin plate spline interpolation and hydraulic simulation calibration processing; An analysis module, configured to perform Markov decision analysis based on the dynamic water depth distribution grid model to obtain the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix; An input module, configured to input the irrigation heterogeneity distribution map and the irrigation system fault diagnosis matrix into the deep reinforcement learning framework for multi-step decision optimization under constraints to obtain the sub-region irrigation control parameter set.

9. An equipment for measuring the water depth of the field surface based on image intelligent recognition, characterized in that, Including a memory and a processor, the memory stores a computer program that can run on the processor, and the processor implements the method for measuring the water depth of the field surface based on image intelligent recognition as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by a processor, the processor is caused to execute the method for measuring the water depth of the field surface based on image intelligent recognition according to any one of claims 1 to 7.

Images