Image recognition-based mulch film recovery rate measurement method and system

Abstract

The present application relates to the technical field of image recognition, in particular to a mulch recovery rate measurement method and system based on image recognition, comprising the following steps: collecting farmland images to extract pixel matrices, extracting mulch distribution maps and extracting contours, calculating overlap rates to generate inhibition coefficients, reversing maps to extract gap areas to calculate dominant degrees, combining inhibition coefficients and dominant degrees to correct coverage rates, and obtaining recovery rate information. The present application identifies mulch residue areas by performing semantic segmentation on the ground surface pixel matrix to extract connected domain contour coordinates, determines the range of morphological operations in combination with the compactness of the fragments, reversely extracts non-mulch areas, calculates their structure dominant degree indexes based on the Euclidean distance and area, constructs a coverage correction function in combination with the indexes to adjust the coverage area, avoids measurement errors caused by fragment overlap and gap interference, and improves the adaptability and accuracy of the measurement results to complex mulch distribution.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a method and system for calculating the recycling rate of plastic film based on image recognition. Background Technology

[0002] Image recognition technology is a technical field in which computers identify and judge objective things by acquiring, analyzing and understanding targets in images. Its core aspects include image acquisition devices acquiring two-dimensional or multi-dimensional image information, representing images in grayscale or color, locating and distinguishing target areas based on pixel distribution characteristics, quantifying the shape, area and quantity of targets, and outputting structured information based on recognition results. This technical field is widely used in application scenarios such as agricultural production monitoring, environmental perception and resource management.

[0003] The traditional image recognition-based method and system for calculating the plastic film recycling rate involves taking pictures of the ground surface with a camera or drone after farmland operations are completed to obtain image data containing plastic film residue. By distinguishing the exposed soil crop residue and plastic film fragments in the image by color or brightness differences, the outline of the plastic film area is extracted and its pixel ratio in the image is calculated. Then, the actual area information of the shooting area is combined to convert the residual area of ​​plastic film, thereby completing the calculation of the plastic film recycling rate based on the ratio between the area of ​​plastic film laid before recycling and the area of ​​residual plastic film after recycling.

[0004] Traditional calculation methods rely on the color or brightness differences between the plastic film and the background soil or crop residue in the image to extract the contour. In scenarios with severe plastic film pollution, large color changes, or unevenly distributed broken plastic film, it is difficult to ensure the accuracy of identification. There are problems such as incomplete extraction of plastic film areas due to blurred image features or ambient light interference. In addition, calculating the residual amount by combining the pixel ratio with the area ignores the spatial superposition of plastic film fragments, interference between fragments, and the structural distribution characteristics of non-plastic film areas, which can easily lead to significant deviations between the estimated results and the actual situation, affecting the accuracy and reliability of the recovery rate calculation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose a method and system for calculating the plastic film recycling rate based on image recognition.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for calculating the plastic film recycling rate based on image recognition, comprising the following steps: S1: Collect farmland surface images, extract two-dimensional pixel matrices, use U-Net to classify the two-dimensional pixel matrices to generate a distribution map of plastic film fragments, and extract the geometric contour coordinates of the connected domains of plastic film fragments; S2: Call the distribution map and geometric contour coordinates of the plastic film fragments, calculate the shape compactness parameters, set the radius of the structural elements and extend the geometric contour coordinates to enclose the area, and generate a single spatial interference domain. S3: Perform pairwise intersection operation on the single spatial interference domain to obtain the area of ​​the overlapping region, calculate the ratio of the area of ​​the overlapping region to the area of ​​the single spatial interference domain to obtain the spatial overlap rate, and generate the coverage contribution suppression coefficient. S4: Invert the distribution map of the plastic film fragments, extract the non-plastic film void areas, calculate the distance value from the pixel of the non-plastic film void area to the plastic film boundary, and calculate the void structure dominance index by combining the distance value and the connected area of ​​the non-plastic film void area. S5: Based on the weighted summation of pixels in the distribution map of plastic film fragments according to the coverage contribution inhibition coefficient, generate the plastic film coverage area. Combine the void structure dominance index to construct the coverage rate correction function. Use the coverage rate correction function and the plastic film coverage area to calculate and obtain the agricultural plastic film recycling rate information.

[0007] The present invention is improved in that the geometric contour coordinates include the boundary vertex sequence, the centroid position of the polygon, and the contour closure identifier; the single-unit spatial interference domain includes the expanded boundary mask, the virtual morphology mapping layer, and the neighborhood occupancy matrix; the coverage contribution suppression coefficient includes the overlapping area weight, the density penalty factor, and the effective contribution ratio; the void structure dominance index includes the regional connectivity score, the longitudinal penetration feature, and the average distance gradient value; and the farmland mulch film recycling rate information includes the corrected coverage percentage and the mulch film residual density level.

[0008] The present invention is improved in that the step of obtaining the geometric contour coordinates is specifically as follows: S111: A two-dimensional pixel matrix of the farmland surface is obtained by an image acquisition device deployed above the farmland operation area. Multi-level feature extraction and non-linear transformation are performed on the red, green and blue channel values ​​of each pixel in the matrix to obtain a deep feature vector including the edge and texture information of the land cover. The spatial dimension of the feature vector is reconstructed to restore the image resolution. The confidence value of each pixel position belonging to the residual plastic film category is calculated to generate a probability matrix of the surface pixel category. S112: Call the surface pixel category probability matrix, compare the confidence value of the plastic film residue category of each pixel in the matrix with the confidence value of the background soil category, define the pixel whose confidence value of the plastic film residue category exceeds the preset confidence judgment threshold as the plastic film target pixel, construct a raster image that only includes the foreground and background values ​​based on the row and column position distribution of the plastic film target pixels on the two-dimensional plane, and generate a plastic film fragment distribution map; S113: Based on the distribution map of the plastic film fragments, perform connected component labeling on the pixels in the foreground area, aggregate the plastic film target pixels that are spatially adjacent and continuous to form independent fragment regions, perform boundary tracking, sequentially traverse the edge pixels around each fragment region, obtain the horizontal and vertical coordinate sequences of the edge pixels in the image coordinate system, and construct vectorized polygon data to obtain the geometric contour coordinates of the connected domain of the plastic film fragments.

[0009] The present invention is improved in that the step of obtaining the single-unit spatial interference domain is specifically as follows: S211: Call the distribution map and geometric contour coordinates of the plastic film fragments, traverse each independent connected domain object, use coordinate analytical geometry logic to calculate the total number of pixels inside the closed area enclosed by the contour vertex sequence as the area, accumulate the Euclidean distance between adjacent vertices on the contour line to obtain the boundary perimeter value, construct the dimensionless ratio of the area value to the square of the perimeter value, calculate the measurement index reflecting the degree of edge jaggedness and morphological regularity of the plastic film fragments, and obtain the plastic film fragment shape compactness parameter. S212: Based on the shape compactness parameter of the plastic film fragments, construct an inverse nonlinear mapping relationship table, associate the edge morphology with the value below the preset shape compactness threshold with the high-weight expansion coefficient, perform a product operation on the expansion coefficient and the preset farmland soil environment lateral influence distance benchmark value, calculate the pixel size value, and independently match the corresponding morphological processing kernel size for each fragment, and establish the structural element radius of the morphological expansion operation. S213: Call the morphological dilation operation's structural element radius and geometric contour coordinates, construct a circular structural element corresponding to the radius in the two-dimensional pixel matrix, perform morphological dilation logic operation along the geometric boundary trajectory of the mulch film fragments with the center of the structural element, extend the area defined by the original contour coordinates outward in the normal direction and fill the neighboring pixels swept by the structural element, forming a binary mask layer including the original area and the extended area, and generating a single spatial interference domain.

[0010] The present invention is improved in that the step of obtaining the coverage contribution suppression coefficient is specifically as follows: S311: For the single spatial interference domain, construct a spatial index tree structure, retrieve adjacent mask object pairs that have geometric bounding box intersections, perform a pixel-by-pixel Boolean logical AND operation on the binarized matrix of each pair of adjacent masks, locate the set of pixel coordinates in the common coverage area of ​​the two, and use an accumulator counter to count the total number of pixels in the set that are in the active state, and generate the area of ​​the overlapping region of adjacent interference domains. S312: Call the area of ​​the overlapping region of the adjacent interference domain and the spatial interference domain of the single unit, perform a full-domain scan on the binarized layer of each single unit interference domain, count the total number of pixels effectively covered by itself, construct a ratio calculation model with the total number of pixels as the denominator and the cumulative value of the area of ​​all overlapping regions involved in the single unit as the numerator, perform division operation, quantify the degree of spatial overlap of the single unit due to the surrounding fragment space, and obtain the spatial overlap rate value. S313: Based on the spatial overlap rate value, construct an inverse weighted attenuation mapping model, set a response logic in which the output weight decreases non-linearly and monotonically when the overlap rate index increases, map the high overlap rate state to a low confidence weight value through the model, assign unit benchmark weights to fragments in the spatial isolation state, calculate the effective proportion factor of each fragment in the final coverage statistics, and generate a coverage contribution suppression coefficient.

[0011] The present invention is improved in that the step of obtaining the void structure dominance index is specifically as follows: S411: Call the plastic film fragment distribution map, perform binarization logic inversion on the plastic film pixels and background pixels, use the connected component labeling algorithm to extract independent void blocks, traverse each void block to count the total number of pixels inside it, and establish a set of non-plastic film void connected domain objects including spatial location index and area attribute data. S412: Based on the binarized mask data in the non-mulch film void connected domain object set, the Euclidean distance transformation algorithm is applied to calculate the Euclidean distance from each background pixel in the void region to the nearest foreground mulch film boundary, and a gray-level matrix reflecting the depth change gradient inside the void is constructed to obtain the pixel distance distribution field inside the void. S413: Call the pixel distance distribution field inside the gap and the set of non-membrane gap connected domain objects, extract the area value of the connected domain and the distance pixel value within the corresponding region, and calculate to obtain the gap structure dominance index.

[0012] The present invention is improved in that the formula for calculating the dominance index of the void structure is specifically as follows: ; in, Represents the dominance index of the void structure. This represents the total number of pixels within the current gap region. Representing the Normalized distance value of each pixel, The normalized area value representing the non-mulch film void region, Represents the normalized maximum distance value within the region. The preset weighted adjustment coefficient represents the logarithmic difference term.

[0013] The present invention is improved in that the steps for obtaining the agricultural plastic film recycling rate information are specifically as follows: S511: Call the coverage contribution inhibition coefficient and the distribution map of plastic film fragments, traverse each connected domain object marked as plastic film category in the map, count the total number of pixels covered by each connected domain, perform a product operation on the total number and the corresponding inhibition coefficient to obtain the effective pixel count of a single unit, accumulate the effective pixel count of the connected domain, and multiply the accumulated value by the physical area of ​​the ground surface corresponding to the single pixel to generate the corrected effective coverage area of ​​plastic film. S512: Based on the void structure dominance index, construct an inverse nonlinear compensation model, set the confidence weight of residual amount estimation according to void characteristic data, adjust the compensation curve parameters, establish a mathematical mapping relationship for dynamically adjusting the area estimation value according to the input index, and obtain the coverage correction mapping function. S513: Call the coverage correction mapping function and the corrected effective coverage area of ​​the mulch film, substitute the area value into the mapping function to perform calibration calculation, obtain the absolute value of the residual amount after compensation by the void feature parameter, perform a division comparison operation with the total surface area of ​​the farmland operation area, calculate the percentage index describing the current residual state of the mulch film, and generate farmland mulch film recycling rate information.

[0014] A system for calculating the plastic film recycling rate based on image recognition, the system being used to implement the aforementioned method for calculating the plastic film recycling rate based on image recognition, the system comprising: The connectivity contour analysis module collects farmland surface images, extracts a two-dimensional pixel matrix, uses U-Net to classify the two-dimensional pixel matrix to generate a distribution map of plastic film fragments, and extracts the geometric contour coordinates of the connected domains of the plastic film fragments. The spatial interference analysis module calls the distribution map and geometric contour coordinates of the plastic film fragments, calculates the shape compactness parameters, sets the radius of the structural elements and extends the area enclosed by the geometric contour coordinates to generate a single spatial interference domain. The spatial overlap assessment module performs pairwise intersection operations on the individual spatial interference domains to obtain the area of ​​the overlapping region, calculates the ratio of the area of ​​the overlapping region to the area of ​​the individual spatial interference domain to obtain the spatial overlap rate, and generates the coverage contribution suppression coefficient. The void impact assessment module reverses the distribution map of the plastic film fragments, extracts the non-plastic film void areas, calculates the distance value from the pixel of the non-plastic film void area to the plastic film boundary, and calculates the void structure dominance index by combining the distance value and the connected area of ​​the non-plastic film void area. The recycling rate analysis module generates the mulch film coverage area by weighted summing of pixels in the mulch film fragment distribution map based on the coverage contribution inhibition coefficient. It then constructs a coverage rate correction function by combining the void structure dominance index and uses the coverage rate correction function and the mulch film coverage area to calculate and obtain farmland mulch film recycling rate information.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, semantic segmentation of the surface pixel matrix is ​​used to identify residual mulch film areas and extract the contour coordinates of connected components. The morphological operation range is adaptively determined by combining the compactness of fragment shapes. After generating the possible interference range of each fragment, Boolean superposition is performed to obtain overlap information and a coverage contribution suppression coefficient is introduced to avoid overestimation of area caused by fragment overlap. At the same time, non-mulch film areas are extracted in reverse and their structural dominance index is calculated based on Euclidean distance and area. The coverage correction function is constructed in combination with the index to adjust the coverage area, avoiding the measurement errors caused by fragment overlap and gap interference, and improving the adaptability and accuracy of the measurement results to complex mulch film distribution. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart of the process for obtaining geometric contour coordinates in this invention; Figure 3 This is a flowchart illustrating the process of obtaining the spatial interference domain of a single entity in this invention; Figure 4 This is a flowchart illustrating how the coverage contribution suppression coefficient is obtained in this invention. Figure 5 This is a flowchart for obtaining the void structure dominance index according to the present invention; Figure 6 This is a flowchart illustrating the process of obtaining farmland plastic film recycling rate information according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0019] Please see Figure 1 This invention provides a technical solution: a method for calculating the plastic film recycling rate based on image recognition, comprising the following steps: S1: Obtain a two-dimensional pixel matrix of the farmland surface by image acquisition equipment deployed above the farmland operation area, and use the U-Net semantic segmentation model to perform pixel-level classification of the residual plastic film area and the background soil area of ​​the two-dimensional pixel matrix, generate a distribution map of plastic film fragments and extract the geometric contour coordinates of the connected domain of the plastic film fragments; S2: Call the distribution map and geometric contour coordinates of the plastic film fragments, calculate the shape compactness parameter of the connected domain of the plastic film fragments, set the structural element radius of the morphological dilation operation based on the shape compactness parameter, perform morphological dilation processing on the area enclosed by the geometric contour coordinates, and generate the corresponding single spatial interference domain. S3: Perform Boolean intersection operation on the generated individual spatial interference domains to obtain the area of ​​the overlapping region between adjacent individual spatial interference domains. Calculate the spatial overlap rate based on the ratio of the overlapping region area to the total area of ​​the individual spatial interference domains. Generate the coverage contribution inhibition coefficient for each mulch film fragment connected domain based on the spatial overlap rate. S4: Perform binarization and inversion processing on the distribution map of plastic film fragments to extract non-plastic film void areas. Use the Euclidean distance transform algorithm to calculate the distance value from each pixel in the non-plastic film void area to the nearest plastic film boundary. Calculate the void structure dominance index based on the distance value and the connected area of ​​the non-plastic film void area. S5: Based on the coverage contribution inhibition coefficient, the number of plastic film pixels in the distribution map of plastic film fragments is weighted and summed to obtain the corrected plastic film coverage area. The coverage rate correction function is constructed by combining the void structure dominance index. The final farmland plastic film recycling rate information is calculated using the coverage rate correction function and the plastic film coverage area. The geometric contour coordinates include the boundary vertex sequence, the centroid position of the polygon, and the contour closure identifier. The individual spatial interference domain includes the expanded boundary mask, the virtual morphology mapping layer, and the neighborhood occupancy matrix. The coverage contribution suppression coefficient includes the overlapping area weight, the density penalty factor, and the effective contribution ratio. The void structure dominance index includes the regional connectivity score, the longitudinal penetration feature, and the average distance gradient value. The farmland mulch film recycling rate information includes the corrected coverage percentage and the residual mulch film density level.

[0020] Please see Figure 2 The specific steps for obtaining the geometric contour coordinates are as follows: S111: A two-dimensional pixel matrix of the farmland surface is obtained by an image acquisition device deployed above the farmland operation area. Multi-level feature extraction and non-linear transformation are performed on the red, green and blue channel values ​​of each pixel in the matrix to obtain a deep feature vector including the edge and texture information of the land cover. The spatial dimension of the feature vector is reconstructed to restore the image resolution. The confidence value of each pixel position belonging to the residual plastic film category is calculated to generate a probability matrix of the surface pixel category. Data acquisition was conducted using a high-resolution visible light camera mounted on a drone. The drone's flight altitude was set at 30 meters, a height chosen to balance coverage and ground resolution. Combined with the camera's 35mm focal length lens, it could stably acquire images with a ground resolution of 0.5cm / pixel, sufficient to capture mulch film fragments with a physical diameter greater than 1cm. The acquired raw images, after orthorectification and stitching, were cropped to a size of [size missing]. A two-dimensional pixel matrix is ​​used as the basic input data for subsequent processing. For processing the two-dimensional pixel matrix, an improved U-Net semantic segmentation network model is employed. This model does not rely on off-the-shelf API calls but is constructed through an explicit hierarchical structure. Specifically, the encoder part of the model contains four downsampling modules, each consisting of two... A convolutional layer and a The system consists of max pooling layers. The red, green, and blue channel values ​​of the input data are first normalized and then mapped to... The first convolutional layer sets the number of kernels to 64, a number chosen based on the complexity of the mulch film texture, sufficient to cover the three basic features: edges, wrinkles, and reflections. A modified linear unit (MRU) activation function is used to introduce non-linear features. Through layer-by-layer downsampling, the model extracts deep feature vectors containing mulch film edge texture, reflective properties, and morphological structure. In the decoder, four upsampling modules reconstruct the spatial dimensions of the feature maps. Each upsampling step includes a... The deconvolutional layers are used to enlarge the feature map size and are then concatenated with the feature maps of the corresponding layers in the encoder via skip connections to fuse shallow spatial location information with deep semantic feature information. During model training, a dedicated dataset containing 5000 labeled images was constructed, where the plastic film area was manually and meticulously labeled as a foreground mask. The training process employs a weighted combination of binary cross-entropy loss and Dice coefficient loss to address the sample imbalance problem caused by the small proportion of plastic film pixels on the ground surface (typically less than 15%). The optimizer uses the Adam algorithm, with an initial learning rate set to [value missing]. The learning rate value was determined based on the convergence curve of gradient descent, effectively avoiding oscillations and achieving rapid convergence. After 100 epochs of iterative training, the model parameters converged. During the inference phase, for the input test image, the model output layer uses the Sigmoid activation function, mapping the feature response value of each pixel to... The probability value of the interval. This probability value is the confidence score, representing the likelihood that the pixel belongs to the "plastic film residue category". The final output surface pixel category probability matrix has the same size as the input image, and the positions in the matrix are... The value at the location This directly corresponds to the probability that the pixel at that location in the original image is a plastic film.

[0021] S112: Call the surface pixel category probability matrix, compare the confidence value of the plastic film residue category of each pixel in the matrix with the confidence value of the background soil category, define the pixel whose confidence value of the plastic film residue category exceeds the preset confidence judgment threshold as the plastic film target pixel, construct a raster image that only includes the foreground and background values ​​based on the row and column position distribution of the plastic film target pixels on the two-dimensional plane, and generate the plastic film fragment distribution map; Set confidence threshold The specific threshold was set with reference to statistical data on the differences in spectral reflectance between the mulch film and soil under different lighting conditions. In the experiment, when the solar altitude angle was... to In this case, the confidence score for the high reflectivity area of ​​the mulch film is typically greater than 0.9, while the confidence score for the semi-transparent edge area ranges from 0.6 to 0.8. Setting it to 0.75 is based on the F1-Score maximization principle, aiming to filter out false positive noise caused by soil salinity spots (confidence scores typically between 0.4 and 0.6) while retaining most of the mulch film edge information. This involves iterating through each element of the probability matrix. Execute logical judgment: If If the location is determined to be a target pixel of the mulch film, it is assigned a value of 1 (foreground); if If a pixel is not found in the background soil area, it is identified as a background soil pixel and assigned a value of 0 (background). This operation constructs a distribution map of plastic film fragments. This distribution map is a binary raster image in which the white foreground area accurately delineates the physical morphology of the residual plastic film on the ground surface, while the black background area represents exposed soil or crop straw.

[0022] S113: Based on the distribution map of plastic film fragments, perform connected component labeling on the pixels in the foreground area, aggregate the plastic film target pixels that are spatially adjacent and continuous to form independent fragment regions, perform boundary tracking, sequentially traverse the edge pixels around each fragment region, obtain the horizontal and vertical coordinate sequences of the edge pixels in the image coordinate system, construct vectorized polygon data, and obtain the geometric contour coordinates of the connected domain of the plastic film fragments. For the distribution map of plastic film fragments, an 8-neighborhood connected component labeling algorithm was used to scan the image matrix, grouping contacting plastic film target pixels into the same independent fragment region, and assigning a unique index ID to each region. Experimental data showed that in In the sample plot images, typically 50 to 200 independent fragments are identified. Then, the Moore-Neighbor boundary tracking algorithm is performed on each independent fragment. Starting from the top-left pixel of the fragment, the algorithm probes the surrounding pixels in a clockwise direction until it returns to the starting point. During this process, the x-axis coordinates of all pixels located on the boundary are recorded sequentially. with the vertical axis coordinate Combine these coordinate points into an ordered sequence. ,in The fragments are numbered. This sequence represents the geometric contour coordinates of the connected domains of the plastic film fragments, realizing the conversion from raster data to vectorized polygon data, and providing an accurate data foundation for subsequent geometric feature calculations.

[0023] Please see Figure 3 The specific steps for obtaining the single-unit spatial interference domain are as follows: S211: Call the distribution map and geometric contour coordinates of the plastic film fragments, traverse each independent connected domain object, use coordinate analytical geometry logic to calculate the total number of pixels inside the closed area enclosed by the sequence of contour vertices as the area, accumulate the Euclidean distance between adjacent vertices on the contour line to obtain the boundary perimeter value, construct the dimensionless ratio of the area value to the square of the perimeter value, calculate the measurement index reflecting the degree of edge jaggedness and morphological regularity of the plastic film fragments, and obtain the plastic film fragment shape compactness parameter. For each fragment Geometric properties are calculated using its geometric contour coordinates. Area This is obtained by counting the number of all pixels within the connected component, for example, a certain fragment. Pixels. Boundary perimeter The Euclidean distance between adjacent coordinate points in the contour sequence is calculated by accumulating the distances. The calculation is set to obtain... Pixels. To quantify the morphological regularity of the fragments, a parameter for the compactness of the film fragment shape was constructed. The calculation formula is defined as follows: This index is dimensionless, and its physical meaning lies in measuring the degree of deviation of the target object from an ideal circle. For a standard circle, its value is 1; the rougher the edges, the more elongated or fragmented the shape, the closer the value is to 0. Substituting the above example data into the calculation: This parameter effectively reflects the jaggedness of the edges of the plastic film fragments. The lower the value, the more severe the fragment is due to wind erosion or mechanical pulling, and the easier it is for its edges to be covered by soil, thus having a higher potential range of spatial disturbance.

[0024] S212: Based on the shape compactness parameter of the plastic film fragments, a reverse nonlinear mapping relationship table is constructed. The edge morphology with a value lower than the preset shape compactness threshold is associated with a high-weight expansion coefficient. The expansion coefficient is multiplied with the preset farmland soil environment lateral influence distance benchmark value to calculate the pixel size value of the circular structural element used to define the spatial expansion range. The corresponding morphological processing kernel size is matched independently for each fragment to establish the structural element radius of the morphological expansion operation. The process of constructing the inverse nonlinear mapping relationship table is as follows: set the upper and lower bounds of the expansion coefficient, and construct a negative exponential decay logical mapping relationship with the plastic film fragment shape compactness parameter as the independent variable; in the negative exponential decay logical mapping relationship, when the function parameter is configured so that the plastic film fragment shape compactness parameter approaches zero, the corresponding expansion coefficient increases nonlinearly and approaches the upper bound of the value; when the plastic film fragment shape compactness parameter approaches one, the corresponding expansion coefficient decreases nonlinearly and converges to the lower bound of the value. Based on the obtained plastic film fragment shape compactness parameters It is necessary to determine the potential interference range of each fragment in space. A reverse nonlinear mapping table is constructed, the core logic of which lies in: shape fragmentation ( Small fragments are often accompanied by irregular fracture zones. In reality, these areas are easily covered by trace amounts of soil, resulting in a visual "disconnection." Therefore, a larger expansion radius is needed to restore their potential physical connectivity. An upper limit boundary is set for the expansion coefficient. This value is set based on the physical tensile limit of polyethylene mulch film, meaning that the film can undergo a maximum deformation displacement of approximately 3 times its original value before breaking; a lower limit boundary is set for this value. This indicates that no additional expansion is performed for perfectly circular fragments. Construct the negative exponential decay logistic mapping function: in Let be the decay rate constant; in this embodiment, we take . The constant The specific settings were based on the statistical distribution of plastic film breakage patterns. At this time, it can ensure that typical fragments with a compactness between 0.2 and 0.6 obtain significant extended gain, while the gain of regular fragments with a compactness greater than 0.8 rapidly decays to 0, which is in accordance with physical laws. For example, for the aforementioned Calculate the expansion factor for the fragments: Establish baseline values ​​for the lateral influence distance of farmland soil environment. This value is strictly dependent on the texture type of farmland soil. By consulting a soil physical property database, for the sandy loam environment in this embodiment, due to the weak adhesion between soil particles (internal friction angle approximately...),... The edges of the plastic film have a wide-ranging effect on loosening the surrounding soil, therefore, a setting is necessary. Pixels (corresponding to a physical distance of 5cm). For clay, this baseline value will be adjusted down to 6 pixels. Finally, the radius of the structuring element for the morphological dilation operation is calculated. : Pixels. This process independently matches the processing kernel size, which is negatively correlated with the morphological characteristics, to each fragment, ensuring the physical plausibility of the spatial interference simulation.

[0025] S213: Call the morphological dilation operation on the radius and geometric contour coordinates of the structural element, construct a circular structural element corresponding to the radius in the two-dimensional pixel matrix, perform morphological dilation logic operation on the center of the structural element along the geometric boundary trajectory of the film fragment, extend the area defined by the original contour coordinates to the external normal direction and fill the neighboring pixels swept by the structural element, form a binary mask layer including the original area and the extended area, and generate a single spatial interference domain. The process of performing morphological dilation logic operation on the geometric boundary trajectory of the structural element center along the film fragment is as follows: traverse each discrete coordinate point in the geometric contour coordinate sequence and take it as the center anchor point of the circular structural element; retrieve the background pixel position covered by the circular structural element at the current anchor point in the two-dimensional pixel matrix and update the pixel value state of the background pixel position to the effective interference domain identifier; after traversing all anchor points, perform union fusion on the pixel regions marked as effective interference domain identifiers to establish a connected single spatial interference domain. Using the calculated radius For each pixel, a circular structuring element with a diameter of 31 pixels is constructed. The geometric contour coordinates of this fragment are traversed, aligning the center of the structuring element with each pixel on the contour in turn, and marking all background pixels covered by the structuring element as the "interference domain". This process is mathematically morphologically equivalent to performing a dilation operation on the original binary image. The final generated single-unit spatial interference domain... It is an expanded binary mask. It not only includes the original area of ​​plastic film fragments, but also includes a ring-shaped area extending outwards by about 15 pixels. This area simulates the range of shading or entanglement effects that plastic film fragments may have on the surrounding area under mechanical operation or wind action, that is, it defines the spatial logic that "if other fragments exist within this range, the two may belong to the same physical residual source or affect each other."

[0026] Please see Figure 4 The specific steps for obtaining the coverage contribution suppression coefficient are as follows: S311: For a single spatial interference domain, construct a spatial index tree structure, retrieve adjacent mask object pairs that have geometric bounding box intersections, perform a pixel-by-pixel Boolean AND operation on the binarized matrix of each pair of adjacent masks, locate the set of pixel coordinates in the common coverage area of ​​the two, and use an accumulator counter to count the total number of active pixels in the set to generate the area of ​​the overlapping region of adjacent interference domains. Construct an R-Tree spatial index for all individual spatial interference domains. Query the R-Tree to quickly filter out adjacent fragment pairs whose bounding boxes intersect. To avoid processing all fragments A global traversal is performed. For each selected pair of fragments, its corresponding binary mask is extracted. and Perform a pixel-by-pixel Boolean AND operation: .statistics The total number of pixels with a median value of 1 represents the area of ​​the overlapping region between adjacent interference domains. For example, fragment A has an interference domain area of ​​2000 pixels, fragment B has an interference domain area of ​​1800 pixels, and the pixel count of the overlapping part is... Pixel.

[0027] S312: Call the area of ​​the overlapping region of the adjacent interference domain and the spatial interference domain of the single unit, perform a full-domain scan on the binarized layer of each single unit interference domain, count the total number of pixels effectively covered by itself, construct a ratio calculation model with the total number of pixels as the denominator and the cumulative value of the area of ​​all overlapping regions involved in the single unit as the numerator, perform division operation, quantify the degree of spatial overlap of the single unit due to the surrounding fragment space, and obtain the spatial overlap rate value. For a certain fragment Its total interference domain area is Iterate through all neighboring fragments that overlap with it. Cumulative overlapping area: Calculate the spatial overlap rate. Continuing the previous example, if fragment A only overlaps with fragment B, then... This value quantifies the degree to which fragment A is crowded out by the surrounding environment. The higher the overlap rate, the more dense the residual area, making it more difficult to recover individually. Furthermore, it is easily counted repeatedly or underestimated due to shading in traditional area statistics.

[0028] S313: Based on the spatial overlap rate, a reverse weighted attenuation mapping model is constructed. The output weight is set to decrease non-linearly and monotonically when the overlap rate index increases. The model maps the high overlap rate state to a low confidence weight value and assigns a unit benchmark weight to the fragments in the spatial isolation state. The effective proportion factor of each fragment in the final coverage statistics is calculated, and the coverage contribution suppression coefficient for the connected domain of each film fragment is generated. The process of constructing the inverse weighted decay mapping model is as follows: select the exponential decay function as the mapping basis, input the spatial overlap rate value as the independent variable into the exponential term of the exponential function, and introduce a preset decay control constant to linearly scale the exponential term; set the boundary constraints of the inverse weighted decay mapping model so that when the input spatial overlap rate value is zero, the output value of the function is always equal to the unit base weight. The process of setting the response logic that the output weights decrease non-linearly and monotonically when the overlap rate index increases is as follows: Configure the base and sign of the exponential function so that as the spatial overlap rate increases, the output value of the inverse weighted attenuation mapping model shows a decreasing trend in the absolute value of the slope in the range from zero to the unit benchmark weight; Set a minimum weight truncation threshold for the high overlap rate range, and when the calculated weight is lower than the minimum weight truncation threshold, the minimum weight truncation threshold is directly used as the output; The process of calculating the effective proportion factor of each fragment in the final coverage statistics is as follows: extract the original geometric area value of the connected domain of each film fragment, multiply the original geometric area value with the corresponding model output weight value; define the weighted area value obtained by the calculation as the effective proportion factor, and store the model output weight value as the coverage contribution suppression coefficient.

[0029] A reverse weighted decay mapping model is constructed to correct the area contribution of dense regions. An exponential decay function is selected, and a decay control constant is set. Minimum weight cutoff threshold The calculation formula is: .in, The setting is based on the transmittance experiment of densely stacked mulch film: when the overlap rate reaches 100% (that is, the two layers of mulch film completely overlap), its visual area contribution should rapidly decrease to about 20% of a single layer (only the edges are visible). It perfectly matches this decay curve. The setting is to prevent the weights from dropping to zero, ensuring that any physically existing fragment retains at least 20% of the basic statistical weight. When input... (Without overlap) This refers to the unit baseline weight, indicating that the fragment is completely independent and its area contribution is 100% effective. When input... When calculating weights: The calculation result This is the coverage contribution suppression coefficient of fragment A. This means that in the final statistics, the effective area of ​​fragment A will be discounted to offset the impact of the visually inflated area caused by spatial overlap or the increased difficulty of data recovery.

[0030] Please see Figure 5 The specific steps for obtaining the void structure dominance index are as follows: S411: Call the distribution map of plastic film fragments, perform binary logic inversion on the plastic film pixels and background pixels, use the connected component labeling algorithm to extract independent void blocks, traverse each void block to count the total number of pixels inside it, and establish a set of non-plastic film void connected domain objects including spatial location index and area attribute data. The distribution map is inverted, i.e., the original mulch film pixels (1) become 0, and the original background pixels (0) become 1. Connected component labeling is then performed again on the inverted image to extract the bare soil areas, i.e., independent void blocks. A set of non-mulch film void connected domain objects is established, recording the total number of pixels (area) and location index of each void object. For example, a large area of ​​bare soil voids is extracted, with a total number of pixels of... Pixel.

[0031] S412: Based on the binary mask data in the non-mulch film void connected domain object set, the Euclidean distance transformation algorithm is applied to calculate the Euclidean distance from each background pixel in the void region to the nearest foreground mulch film boundary, and a gray-level matrix reflecting the depth change gradient inside the void is constructed to obtain the pixel distance distribution field inside the void. Apply a Euclidean distance transform to each gap region. For any pixel within the gap... Calculate the pixel closest to the mulch film boundary. Euclidean distance The generated distance matrix is ​​the pixel distance distribution field inside the void. In this field, the pixel distance values ​​near the boundary of the mulch film are smaller, while the pixel distance values ​​located in the center of large areas of bare soil are larger, forming a gradient field that reflects the "depth" of the void.

[0032] S413: Call the pixel distance distribution field inside the gaps and the set of connected domain objects for non-mulch film gaps, extract the area values ​​of the connected domains and the corresponding distance pixel values ​​within the regions, using the following formula: ; The dominance index of the void structure is obtained through computation. in, Represents the dominance index of the void structure. This represents the total number of pixels within the current gap region. It is obtained by traversing the connected component mask and counting the coordinates of the active points within it. Representing the The normalized distance value of each pixel is obtained by calculating the Euclidean distance from that pixel to the nearest boundary of the plastic film and dividing it by the length of the image diagonal pixels. The normalized area value representing the non-mulch film void region is obtained by calculating the total pixel coverage of the current connected region and dividing it by the total pixel area of ​​the entire image. This represents the normalized maximum distance value within the region. It is obtained by filtering the peak values ​​of the distance transformation values ​​within the current region and dividing by the length of the image diagonal pixels. The preset weighted adjustment coefficient representing the logarithmic difference term is obtained by calling a fixed dimensionless constant from the configuration database based on the expected complexity of the farmland surface texture. For a specific gap region, extract the following parameters: total number of pixels. 0. Normalized distance value Set the diagonal length of the image. Pixels. If the distance to a point is 72.4 pixels, then Normalized area value : Normalized maximum distance value : Set the maximum distance between the center of the region to 200 pixels, then Weighted adjustment coefficient Based on the complexity of farmland texture, preset The coefficient is set based on the fact that when the farmland surface is relatively flat, the shape of the voids is more dominant. Take the smaller value; if the surface is rough, the weight of the logarithmic difference term needs to be increased. It can be adjusted to 0.2. In this example, which is for leveled cotton fields, we take 0.1.

[0033] Substitute the above parameters into the formula: ; The calculation logic and meaning of the formula are explained below: Formula Part 1 This study utilizes the ratio of the accumulated distance field value to the root mean square (RMS) and combines it with area normalization. This term primarily measures the topological compactness and internal homogeneity of the void region. If the voids are regularly shaped and concentrated (i.e., large areas of continuous bare soil), the internal distance field is uniformly distributed, and this ratio tends to stabilize; if the voids are fragmented by finely broken mulch film, the distance value... The generally small and highly fluctuating values ​​cause changes in the root mean square term in the denominator, thus altering the ratio. (Formula Part Two) The logarithmic difference between area and maximum depth was calculated. This term measures the scale uniformity of the voids. For normal, large voids, the area should be proportional to the maximum depth (distance from center to edge), with a small difference. However, for narrow or irregular voids (often caused by mulch film tears), the area may be large but the depth shallow, leading to a larger logarithmic difference and thus increasing the risk of voids becoming larger. The value of .

[0034] set up , The calculated result for the (root mean square distance) is 0.03.

[0035] First calculation: 2.5 Note: The numerical example here is for demonstrating logic. In actual normalized numerical systems, this term usually converges to... The interval is set to 2.5 after normalization adjustment.

[0036] Second calculation: .

[0037] .

[0038] The result is the porosity index. The value indicates that the void region analyzed exhibits a moderate degree of regularity in structure. This means that although the region is mainly bare soil, its edges are still subject to some degree of erosion or segmentation by the mulch film. It is not a perfectly rectangular or circular void, but it is also not an extremely fragmented or noisy region. This indicator directly guides the subsequent area compensation strategy.

[0039] Please see Figure 6 The specific steps for obtaining information on agricultural plastic film recycling rates are as follows: S511: Call the coverage contribution inhibition coefficient and the distribution map of plastic film fragments, traverse each connected domain object marked as plastic film category in the map, count the total number of pixels covered by each connected domain, perform a product operation with the corresponding inhibition coefficient to obtain the effective pixel count of a single unit, accumulate the effective pixel count of the connected domain, and multiply the accumulated value by the physical area of ​​the land surface corresponding to the single pixel to generate the corrected effective coverage area of ​​plastic film. Traverse all identified plastic film fragments, for example, fragment A: original pixel count 450, suppression coefficient 0.713. Effective pixel count per unit. The total effective pixel count is obtained by summing all fragments in the entire image. Given that GSD is 0.5cm / pixel, the physical area of ​​a single pixel is... Correcting the effective coverage area of ​​the plastic film. .

[0040] S512: Based on the void structure dominance index, an inverse nonlinear compensation model is constructed. The confidence weight of the residual amount estimation is set according to the void characteristic data, the compensation curve parameters are adjusted, and a mathematical mapping relationship for dynamically adjusting the area estimation value according to the input index is established to obtain the coverage correction mapping function. The process of constructing the inverse nonlinear compensation model is as follows: the void structure dominance index is compared with the preset void dispersion benchmark interval. If the void structure dominance index is within the void dispersion benchmark interval, the linear regression logic is called to construct the basic architecture. If the void structure dominance index exceeds the void dispersion benchmark interval, the exponential decay logic is called to construct the basic architecture. The process of adjusting the compensation curve parameters is as follows: obtain a sample dataset including known residual amounts, perform least squares iterative operation on the sample dataset, calculate the numerical combination that minimizes the sum of squared fitted residuals, and lock the numerical combination as the slope and intercept parameters of the compensation curve. The process of establishing the coverage correction mapping function is as follows: the slope and intercept parameters are substituted into the constructed basic architecture to generate a functional relationship with the void structure dominance index as the input variable and the area correction factor as the output variable. To further eliminate the impact of unidentified micro-fragments, the porosity dominance index was used. Compensation is performed. A compensation model is constructed and fitted using a sample dataset containing known residual amounts (obtained through manual collection and weighing). A baseline interval for the porosity dispersion is defined. The interval was set based on statistics from hundreds of standard sample plots: when Within this range, the amount of residual plastic film shows a significant linear correlation with porosity characteristics; if This indicates that the pores are extremely large and pure, with very little plastic film present; if This indicates that the surface is extremely fragmented. In this example... Located within the interval, linear regression logic is used. The slope obtained from the fitting is... ,intercept These parameters are derived from least squares regression using historical training data. A coverage correction mapping function is constructed. Calculate the area correction factor: This factor indicates that, based on the inference of gap features, the current recognition area needs to be increased by 11% to compensate for minor omissions.

[0041] S513: Call the coverage correction mapping function and correct the effective coverage area of ​​the mulch film, substitute the area value into the mapping function to perform calibration calculation, obtain the absolute value of the residual amount after compensation by the void feature parameter, perform a division comparison operation with the total surface area of ​​the farmland operation area, calculate the percentage index describing the current residual state of the mulch film, and generate farmland mulch film recycling rate information. Substitute the corrected area into the mapping function for calibration: final residual area The total area of ​​the farmland operation area (sample plot) is... Calculate the residual rate: If the initial film coverage rate is set at 90%, then the agricultural film recycling rate will be... The final agricultural plastic film recycling rate was 82.4%.

[0042] The table below shows a comparison of experimental data from three typical quadrats in this embodiment, verifying the effectiveness of the method: Table 1 Comparison of Calculation Results of Mulch Film Recycling Rate for Different Formulas As shown in Table 1, by introducing shape compactness, spatial overlap suppression and void structure compensation mechanisms, the error between the calculation results and the manual measured values ​​is controlled within 1% in scenarios with different degrees of fragmentation (differences in the number and compactness of fragments), which is significantly better than the traditional method based solely on pixel statistics (the error is usually 5%-10%). This verifies the high accuracy and robustness of the technical solution in complex farmland environments.

[0043] A system for calculating the plastic film recycling rate based on image recognition is provided. This system implements the aforementioned method for calculating the plastic film recycling rate based on image recognition. The system includes: The connectivity contour analysis module collects farmland surface images, extracts a two-dimensional pixel matrix, uses U-Net to classify the two-dimensional pixel matrix to generate a distribution map of plastic film fragments, and extracts the geometric contour coordinates of the connected domains of the plastic film fragments. The spatial interference analysis module calls the distribution map and geometric contour coordinates of the plastic film fragments, calculates the shape compactness parameters, sets the radius of the structural elements and extends the geometric contour coordinates to enclose the area, and generates a single spatial interference domain. The spatial overlap assessment module performs intersection operations on the individual spatial interference domains to obtain the area of ​​the overlapping region, calculates the ratio of the area of ​​the overlapping region to the area of ​​the individual spatial interference domains to obtain the spatial overlap rate, and generates the coverage contribution suppression coefficient. The void impact assessment module reverses the distribution map of plastic film fragments, extracts non-plastic film void areas, calculates the distance value from the pixel of the non-plastic film void area to the plastic film boundary, and calculates the void structure dominance index by combining the distance value and the connected area of ​​the non-plastic film void area. The recycling rate analysis module generates the mulch film coverage area by weighting and summing the pixels in the mulch film fragment distribution map based on the coverage contribution inhibition coefficient. It then constructs a coverage rate correction function by combining the void structure dominance index and uses the coverage rate correction function and the mulch film coverage area to calculate and obtain farmland mulch film recycling rate information.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for calculating the plastic film recycling rate based on image recognition, characterized in that, Includes the following steps: S1: Collect farmland surface images, extract two-dimensional pixel matrices, use U-Net to classify the two-dimensional pixel matrices to generate a distribution map of plastic film fragments, and extract the geometric contour coordinates of the connected domains of plastic film fragments; S2: Call the distribution map and geometric contour coordinates of the plastic film fragments, calculate the shape compactness parameters, set the radius of the structural elements and extend the geometric contour coordinates to enclose the area, and generate a single spatial interference domain. S3: Perform pairwise intersection operation on the single spatial interference domain to obtain the overlapping area, calculate the ratio of the overlapping area to the single spatial interference domain area to obtain the spatial overlap rate, and generate the coverage contribution suppression coefficient. S4: Invert the distribution map of the plastic film fragments, extract the non-plastic film void areas, calculate the distance value from the pixel of the non-plastic film void area to the plastic film boundary, and calculate the void structure dominance index by combining the distance value and the connected area of ​​the non-plastic film void area. S5: Based on the weighted summation of pixels in the distribution map of plastic film fragments according to the coverage contribution inhibition coefficient, generate the plastic film coverage area. Combine the void structure dominance index to construct the coverage rate correction function. Use the coverage rate correction function and the plastic film coverage area to calculate and obtain the agricultural plastic film recycling rate information.

2. The method for calculating the plastic film recycling rate based on image recognition according to claim 1, characterized in that, The geometric contour coordinates include the boundary vertex sequence, the centroid position of the polygon, and the contour closure identifier. The single-unit spatial interference domain includes the expanded boundary mask, the virtual morphology mapping layer, and the neighborhood occupancy matrix. The coverage contribution suppression coefficient includes the overlapping area weight, the density penalty factor, and the effective contribution ratio. The void structure dominance index includes the regional connectivity score, the longitudinal penetration feature, and the average distance gradient value. The farmland mulch film recycling rate information includes the corrected coverage percentage and the mulch film residual density level.

3. The method for calculating the plastic film recycling rate based on image recognition according to claim 1, characterized in that, The specific steps for obtaining the geometric contour coordinates are as follows: S111: A two-dimensional pixel matrix of the farmland surface is obtained by an image acquisition device deployed above the farmland operation area. Multi-level feature extraction and non-linear transformation are performed on the red, green and blue channel values ​​of each pixel in the matrix to obtain a deep feature vector including the edge and texture information of the land cover. The spatial dimension of the feature vector is reconstructed to restore the image resolution. The confidence value of each pixel position belonging to the residual plastic film category is calculated to generate a probability matrix of the surface pixel category. S112: Call the surface pixel category probability matrix, compare the confidence value of the plastic film residue category of each pixel in the matrix with the confidence value of the background soil category, define the pixel whose confidence value of the plastic film residue category exceeds the preset confidence judgment threshold as the plastic film target pixel, construct a raster image that only includes the foreground and background values ​​based on the row and column position distribution of the plastic film target pixels on the two-dimensional plane, and generate a plastic film fragment distribution map; S113: Based on the distribution map of the plastic film fragments, perform connected component labeling on the pixels in the foreground area, aggregate the plastic film target pixels that are spatially adjacent and continuous to form independent fragment regions, perform boundary tracking, sequentially traverse the edge pixels around each fragment region, obtain the horizontal and vertical coordinate sequences of the edge pixels in the image coordinate system, and construct vectorized polygon data to obtain the geometric contour coordinates of the connected domain of the plastic film fragments.

4. The method for calculating the plastic film recycling rate based on image recognition according to claim 3, characterized in that, The specific steps for obtaining the individual spatial interference domain are as follows: S211: Call the distribution map and geometric contour coordinates of the plastic film fragments, traverse each independent connected domain object, use coordinate analytical geometry logic to calculate the total number of pixels inside the closed area enclosed by the contour vertex sequence as the area, accumulate the Euclidean distance between adjacent vertices on the contour line to obtain the boundary perimeter value, construct the dimensionless ratio of the area value to the square of the perimeter value, calculate the measurement index reflecting the degree of edge jaggedness and morphological regularity of the plastic film fragments, and obtain the plastic film fragment shape compactness parameter. S212: Based on the shape compactness parameter of the plastic film fragments, construct an inverse nonlinear mapping relationship table, associate the edge morphology with the value below the preset shape compactness threshold with the high-weight expansion coefficient, perform a product operation on the expansion coefficient and the preset farmland soil environment lateral influence distance benchmark value, calculate the pixel size value, and independently match the corresponding morphological processing kernel size for each fragment, and establish the structural element radius of the morphological expansion operation. S213: Call the morphological dilation operation's structural element radius and geometric contour coordinates, construct a circular structural element corresponding to the radius in the two-dimensional pixel matrix, perform morphological dilation logic operation along the geometric boundary trajectory of the mulch film fragments with the center of the structural element, extend the area defined by the original contour coordinates outward in the normal direction and fill the neighboring pixels swept by the structural element, forming a binary mask layer including the original area and the extended area, and generating a single spatial interference domain.

5. The method for calculating the plastic film recycling rate based on image recognition according to claim 4, characterized in that, The specific steps for obtaining the coverage contribution suppression coefficient are as follows: S311: For the single spatial interference domain, construct a spatial index tree structure, retrieve adjacent mask object pairs that have geometric bounding box intersections, perform a pixel-by-pixel Boolean logical AND operation on the binarized matrix of each pair of adjacent masks, locate the set of pixel coordinates in the common coverage area of ​​the two, and use an accumulator counter to count the total number of pixels in the set that are in the active state, and generate the area of ​​the overlapping region of adjacent interference domains. S312: Call the area of ​​the overlapping region of the adjacent interference domain and the spatial interference domain of the single unit, perform a full-domain scan on the binarized layer of each single unit interference domain, count the total number of pixels effectively covered by itself, construct a ratio calculation model with the total number of pixels as the denominator and the cumulative value of the area of ​​all overlapping regions involved in the single unit as the numerator, perform division operation, quantify the degree of spatial overlap of the single unit due to the surrounding fragment space, and obtain the spatial overlap rate value. S313: Based on the spatial overlap rate value, construct an inverse weighted attenuation mapping model, set a response logic in which the output weight decreases non-linearly and monotonically when the overlap rate index increases, map the high overlap rate state to a low confidence weight value through the model, assign unit benchmark weights to fragments in the spatial isolation state, calculate the effective proportion factor of each fragment in the final coverage statistics, and generate a coverage contribution suppression coefficient.

6. The method for calculating the plastic film recycling rate based on image recognition according to claim 5, characterized in that, The specific steps for obtaining the void structure dominance index are as follows: S411: Call the plastic film fragment distribution map, perform binarization logic inversion on the plastic film pixels and background pixels, use the connected component labeling algorithm to extract independent void blocks, traverse each void block to count the total number of pixels inside it, and establish a set of non-plastic film void connected domain objects including spatial location index and area attribute data. S412: Based on the binarized mask data in the non-mulch film void connected domain object set, the Euclidean distance transformation algorithm is applied to calculate the Euclidean distance from each background pixel in the void region to the nearest foreground mulch film boundary, and a gray-level matrix reflecting the depth change gradient inside the void is constructed to obtain the pixel distance distribution field inside the void. S413: Call the pixel distance distribution field inside the gap and the set of non-membrane gap connected domain objects, extract the area value of the connected domain and the distance pixel value within the corresponding region, and calculate to obtain the gap structure dominance index.

7. The method for calculating the plastic film recycling rate based on image recognition according to claim 6, characterized in that, The specific formula for obtaining the void structure dominance index is as follows: ； in, Represents the dominance index of the void structure. This represents the total number of pixels within the current gap region. Representing the Normalized distance value of each pixel, The normalized area value representing the non-mulch film void region, Represents the normalized maximum distance value within the region. The preset weighted adjustment coefficient represents the logarithmic difference term.

8. The method for calculating the plastic film recycling rate based on image recognition according to claim 6, characterized in that, The specific steps for obtaining the farmland plastic film recycling rate information are as follows: S511: Call the coverage contribution inhibition coefficient and the distribution map of plastic film fragments, traverse each connected domain object marked as plastic film category in the map, count the total number of pixels covered by each connected domain, perform a product operation on the total number and the corresponding inhibition coefficient to obtain the effective pixel count of a single unit, accumulate the effective pixel count of the connected domain, and multiply the accumulated value by the physical area of ​​the ground surface corresponding to the single pixel to generate the corrected effective coverage area of ​​plastic film. S512: Based on the void structure dominance index, construct an inverse nonlinear compensation model, set the confidence weight of residual amount estimation according to void characteristic data, adjust the compensation curve parameters, establish a mathematical mapping relationship for dynamically adjusting the area estimation value according to the input index, and obtain the coverage correction mapping function. S513: Call the coverage correction mapping function and the corrected effective coverage area of ​​the mulch film, substitute the area value into the mapping function to perform calibration calculation, obtain the absolute value of the residual amount after compensation by the void feature parameter, perform a division comparison operation with the total surface area of ​​the farmland operation area, calculate the percentage index describing the current residual state of the mulch film, and generate farmland mulch film recycling rate information.

9. A system for calculating the plastic film recycling rate based on image recognition, characterized in that, The system is used to implement the image recognition-based method for calculating the plastic film recycling rate as described in any one of claims 1-8, and the system comprises: The connectivity contour analysis module collects farmland surface images, extracts a two-dimensional pixel matrix, uses U-Net to classify the two-dimensional pixel matrix to generate a distribution map of plastic film fragments, and extracts the geometric contour coordinates of the connected domains of the plastic film fragments. The spatial interference analysis module calls the distribution map and geometric contour coordinates of the plastic film fragments, calculates the shape compactness parameters, sets the radius of the structural elements and extends the area enclosed by the geometric contour coordinates to generate a single spatial interference domain. The spatial overlap assessment module performs pairwise intersection operations on the individual spatial interference domains to obtain the area of ​​the overlapping region, calculates the ratio of the area of ​​the overlapping region to the area of ​​the individual spatial interference domain to obtain the spatial overlap rate, and generates the coverage contribution suppression coefficient. The void impact assessment module reverses the distribution map of the plastic film fragments, extracts the non-plastic film void areas, calculates the distance value from the pixel of the non-plastic film void area to the plastic film boundary, and calculates the void structure dominance index by combining the distance value and the connected area of ​​the non-plastic film void area. The recycling rate analysis module generates the mulch film coverage area by weighted summing of pixels in the mulch film fragment distribution map based on the coverage contribution inhibition coefficient. It then constructs a coverage rate correction function by combining the void structure dominance index and uses the coverage rate correction function and the mulch film coverage area to calculate and obtain farmland mulch film recycling rate information.

Images