Detection method of multi-layer composite high-strength anti-aging agricultural mulching film

Abstract

The invention relates to the technical field of agricultural intelligent monitoring, and discloses a method for detecting a multi-layer composite high-strength anti-aging agricultural mulching film, and the method comprises the steps: carrying out the high-frequency noise suppression and edge preservation smoothing of a mulching film image, and obtaining a smooth image; generating a degraded area mask based on the color distribution characteristics and morphological processing; calculating a roughness degree score in combination with the standard texture value; fusing the roughness and color data to perform clustering mapping to obtain a comprehensive aging map; calculating an intensity reduction proportion according to the aging map, and calculating a recovery risk and a pollution level in combination with edge fragmentation characteristics to obtain a final aging quantification index; and inputting the index into a neural network model for deviation calibration to obtain a target aging detection result. According to the invention, accurate quantification of the aging degree of the mulching film and scientific evaluation of the pollution risk in a complex environment are realized.

Description

A test method for multi-layer composite high-strength anti-aging agricultural mulch film Technical Field

[0001] This invention relates to the field of agricultural intelligent monitoring technology, and in particular to a detection method for multi-layer composite high-strength anti-aging agricultural mulch film. Background Technology

[0002] Currently, agricultural mulch film covering technology has a significant effect on warming and moisture retention in modern agricultural production. However, during long-term field use, mulch film inevitably ages due to environmental factors such as strong ultraviolet radiation, temperature fluctuations, and mechanical operations. This manifests as decreased light transmittance, yellowing and darkening of the surface color, and roughness and cracking of the micro-texture. This not only reduces the functionality of the mulch film, but in severe cases, it can also produce microplastic fragments that are difficult to recycle, causing long-term soil pollution.

[0003] In existing technologies, the lifespan of mulch film is typically assessed using physical and mechanical performance tests (such as tensile strength tests) under laboratory conditions or visual inspections conducted manually in the field. While laboratory tests provide accurate data, they are destructive tests with limited sampling points, making it difficult to cover the actual aging differences across large areas of farmland. Traditional visual inspections rely primarily on experience to judge the damage and discoloration of the mulch film, and in scenarios lacking high-precision image recognition technology, they are easily affected by complex lighting conditions in the field, soil background color, and dust accumulation on the mulch film surface. This approach cannot deeply analyze the microscopic relationship between color degradation and texture roughening of the mulch film surface, resulting in assessments that are often lagging and subjective, making it difficult to establish standardized quantitative indicators of aging, and consequently, failing to accurately predict the risk of embrittlement and breakage of the mulch film.

[0004] In summary, existing technologies have the problem of making it difficult to accurately, in real time and non-destructively quantify and assess the degree of aging of plastic film in complex field environments. Summary of the Invention

[0005] This invention provides a detection method for multi-layer composite high-strength anti-aging agricultural mulch film, in order to solve the problem of accurately, in real time and non-destructively quantifying the degree of aging of mulch film in complex field environments.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides a method for detecting the aging of high-strength agricultural mulch film, comprising: acquiring surface image data of the mulch film; performing high-frequency noise suppression and edge-preserving smoothing on the surface image data to obtain a smoothed image; based on the smoothed image, performing contrast enhancement and extracting color distribution parameters; identifying color anomaly regions whose color distribution parameters satisfy a preset degradation significance threshold; and performing connected component filtering and boundary filling on the color anomaly regions to obtain a degradation region mask; based on the degradation region mask, and combined with a preset standard texture value, performing local grayscale data extraction and texture roughness quantification analysis to obtain... The roughness score is obtained; based on the roughness score and the color anomaly region, a multidimensional feature vector is constructed and clustered to obtain aging level partitions, and the aging level partitions are mapped onto the smoothed image to obtain a comprehensive aging map; based on the comprehensive aging map, the overall intensity reduction ratio is calculated; if the overall intensity reduction ratio is higher than a preset intensity reduction threshold, the risk estimation of recyclable debris and the integral accumulation of potential soil pollution level are performed based on the smoothed image to obtain the final aging quantification index; the final aging quantification index is input into a pre-trained neural network verification model for bias calibration to obtain the target aging detection result.

[0007] Secondly, the present invention provides an aging detection system for high-strength agricultural mulch film, comprising: an image acquisition and preprocessing module for acquiring image data of the mulch film surface, performing high-frequency noise suppression and edge-preserving smoothing on the mulch film surface image data to obtain a smoothed image; a feature extraction and mask generation module for performing contrast enhancement and extracting color distribution parameters based on the smoothed image, identifying color anomaly regions whose color distribution parameters meet a preset degradation significance threshold, and performing connected component filtering and boundary filling on the color anomaly regions to obtain a degradation region mask; and a texture analysis module for performing local grayscale data extraction and texture roughness quantification analysis based on the degradation region mask and a preset standard texture value to obtain a roughness measurement. The system comprises the following modules: a roughness score module; a clustering and map generation module, used to construct and cluster multi-dimensional feature vectors based on the roughness score and the color anomaly areas to obtain aging level partitions, and map the aging level partitions onto the smoothed image to obtain a comprehensive aging map; an intensity assessment and calculation module, used to calculate the overall intensity reduction ratio based on the comprehensive aging map, and if the overall intensity reduction ratio is higher than a preset intensity reduction threshold, then based on the smoothed image, it performs risk estimation of reclaimed debris and accumulation of potential soil pollution levels to obtain a final aging quantification index; and a model verification module, used to input the final aging quantification index into a pre-trained neural network verification model for bias calibration to obtain the target aging detection result.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention collects images in the field and performs high-frequency noise suppression and edge preservation smoothing, extracts the color distribution parameters and texture roughness index (based on gray-level co-occurrence matrix) of the mulch film surface, and uses the K-means clustering algorithm to fuse color changes and texture deterioration features to generate a comprehensive aging map. This method can comprehensively capture the aging characteristics of mulch film under the action of light and oxygen from two dimensions: micro-texture and macro-color. It uses the deterioration area mask to accurately limit the analysis range and effectively eliminates environmental background interference, thereby realizing the accurate quantification and visual classification of the degree of mulch film aging in complex field environments, and solving the problems of strong subjectivity and inaccurate single-sided feature analysis of the traditional visual method.

[0009] (2) This invention calculates the overall strength reduction ratio based on a comprehensive aging map and triggers a risk assessment of recycled fragments when the ratio exceeds the standard. Then, it calculates the potential soil pollution level through an integral accumulation method. This method establishes a logical connection between the aging characteristics of the mulch film surface and the mechanical strength decay and environmental pollution risk. It can transform the difficult-to-observe risk of additive leaching into a quantifiable numerical indicator, thus providing agricultural managers with a scientific basis for recycling or replacement decisions. This effectively reduces the risk of soil microplastic pollution and chemical residues caused by broken and scattered mulch film, ensuring the sustainability of farmland ecology.

[0010] (3) This invention uses a convolutional neural network to validate and train the final aging quantification index, constructs a validation set using historical field data, and dynamically adjusts the network weights based on prediction bias. The system can continuously correct the detection model using accumulated historical data, adapt to detection needs under different soil backgrounds, light conditions, and crop shading, and determine the reliable detection accuracy of the quantification index, thereby improving the robustness and accuracy of the aging detection model in long-term applications, reducing subjective errors from human intervention, and realizing self-optimization and performance improvement of the detection system. Attached Figure Description

[0011] Figure 1 is a schematic flowchart of the aging detection method for high-strength agricultural mulch film provided in the first embodiment of the present invention; Figure 2 is a schematic structural diagram of the aging detection system for high-strength agricultural mulch film provided in the second embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Referring to Figure 1, the first embodiment of the present invention provides a method for detecting the aging of high-strength agricultural mulch film, comprising the following steps: S11, acquiring surface image data of the mulch film, and performing high-frequency noise suppression and edge preservation smoothing on the surface image data to obtain a smoothed image; S12, based on the smoothed image, performing contrast enhancement and extracting color distribution parameters, identifying color anomaly regions whose color distribution parameters meet a preset degradation significance threshold, and performing connected component filtering and boundary filling on the color anomaly regions to obtain a degradation region mask; S13, based on the degradation region mask, combined with a preset standard texture value, performing local grayscale data extraction and texture roughness quantification analysis to obtain... S14. Based on the roughness score and the color anomaly region, construct and cluster multidimensional feature vectors to obtain aging level partitions, and map the aging level partitions onto the smoothed image to obtain a comprehensive aging map; S15. Based on the comprehensive aging map, calculate the overall intensity reduction ratio. If the overall intensity reduction ratio is higher than the preset intensity reduction threshold, calculate the risk of recycling debris and accumulate the potential soil pollution level integral based on the smoothed image to obtain the final aging quantification index; S16. Input the final aging quantification index into a pre-trained neural network verification model for bias calibration to obtain the target aging detection result.

[0014] In step S11, acquiring surface image data of the plastic film and performing high-frequency noise suppression and edge preservation smoothing on the surface image data to obtain a smoothed image includes: acquiring surface image data of the plastic film using field acquisition equipment; performing two-dimensional pixel matrix analysis and high-frequency abrupt change region identification based on the surface image data of the plastic film to obtain the distribution of high-frequency noise regions; calculating the weighted average value of surrounding adjacent pixels and replacing values ​​based on the distribution of high-frequency noise regions to obtain preliminary noise reduction data; and performing edge locking and Gaussian convolution smoothing operations based on the preliminary noise reduction data to obtain a smoothed image.

[0015] First, a field data collection device (such as a low-altitude drone equipped with a global shutter industrial camera or a ground inspection robot) is used to take vertical overhead photos of farmland covered with mulch under natural lighting conditions to obtain analog photoelectric signals including the surface of the mulch. The analog photoelectric signals are then quantized into discrete digital signals by an analog-to-digital converter inside the camera. These digital signals are mapped into a two-dimensional pixel matrix according to the spatial arrangement of the photosensitive elements. The row index of the matrix corresponds to the vertical coordinate of the physical space, the column index corresponds to the horizontal coordinate of the physical space, and the value of the matrix element represents the gray level (e.g., 0 to 255) of the reflected light intensity at that coordinate point.

[0016] Next, based on the image data of the mulch film surface, two-dimensional pixel matrix analysis and high-frequency abrupt change region identification are performed. Specifically, a sliding window (e.g., a 3×3 pixel area) is used to traverse the two-dimensional pixel matrix, and the gray-level difference statistics between the center pixel and neighboring pixels are calculated. Taking the local variance statistics as an example, the arithmetic mean of the gray-level values ​​of all pixels in the sliding window is first calculated, and then the square of the difference between the gray-level value of each pixel in the window and the arithmetic mean is calculated. The squares of all differences are summed to obtain the total, and finally the total is divided by the total number of pixels in the window to obtain the local variance value. If the local variance value exceeds the preset high-frequency judgment threshold, it is determined that there is a discontinuous brightness jump at the position of the center pixel, and it is marked as a high-frequency abrupt change region, thereby obtaining the high-frequency noise region distribution. Subsequently, based on the distribution of the high-frequency noise region, the weighted average value of the surrounding neighboring pixels is calculated and the value is replaced. For the marked noise pixel, its neighboring pixels that are not marked as noise are selected as references. The weighted average gray value is calculated based on the inverse Euclidean distance weighting method, that is, the neighboring pixels that are closer to the noise point are assigned a larger weight. The gray value of the original noise pixel is replaced by the calculated weighted average gray value to obtain the preliminary noise reduction data.

[0017] It should be noted that the preset high-frequency judgment threshold is not a fixed constant. Its setting is based on the statistical analysis of the noise distribution of film mulch images under different historical lighting conditions (such as direct sunlight on sunny days and diffuse sunlight on cloudy days). The specific setting method is as follows: select historical image samples containing typical noise points (such as dust reflection and sensor thermal noise), calculate the local variance distribution curves of the manually marked noise area and the normal area in the sample, select the variance value corresponding to the intersection of the two distribution curves as the benchmark threshold (for example, variance value 400), and make dynamic fine adjustments according to the ambient light intensity sensor readings during actual detection. The stronger the light, the higher the threshold is adjusted to adapt to high-contrast reflection.

[0018] Finally, based on the preliminary denoising data, edge locking and Gaussian convolution smoothing operations are performed, and the image gradient is calculated using the Sobel operator. Specifically, horizontal and vertical templates are used to convolve with the target pixel and its neighborhood. The horizontal calculation involves multiplying the pixel values ​​in the right column of the neighborhood by positive weights and the pixel values ​​in the left column by negative weights, and then summing the results. The vertical calculation involves multiplying the pixel values ​​in the lower row of the neighborhood by positive weights and the pixel values ​​in the upper row by negative weights, and then summing the results. The square of the calculated horizontal gradient value is added to the square of the calculated vertical gradient value, and the arithmetic square root is taken to obtain the gradient magnitude. Pixels with gradient magnitudes greater than the edge detection threshold are locked as edge information. For pixels in flat regions that are not locked, a Gaussian convolution kernel is applied for smooth convolution. Specifically, a weight matrix with the highest weight at the center and decreasing weight towards the surrounding areas is generated based on a two-dimensional Gaussian function. This weight matrix is ​​then applied to the target pixel and its neighborhood. The product of each element value in the weight matrix and the corresponding pixel gray value is calculated. All product results are summed and divided by the sum of all elements in the weight matrix to obtain the smoothed gray value of the target pixel, thereby outputting a smooth image.

[0019] It should be noted that the edge detection threshold is set based on statistical analysis of the gradient magnitude distribution characteristics of the current image. The specific setting rules and methods are as follows: the gradient magnitude histogram of the entire preliminary noise reduction data is statistically analyzed. This histogram reflects the distribution of the number of pixels with different gradient intensities. The Otsu algorithm is used to automatically calculate a segmentation value that divides the histogram into a foreground edge class and a background flat class, with the gray-level variance between the two classes being the largest. This segmentation value is then determined as the edge detection threshold. This method can adaptively adjust the locking range according to the clarity of the mulch texture, avoiding the loss of weak edges in blurry images or the introduction of too much noise in clear images.

[0020] It is worth noting that the parsing process of the two-dimensional pixel matrix must follow the common data transmission standards for industrial vision, such as the GigE Vision or USB3 Vision protocols, to ensure that no packet loss or misalignment occurs during the mapping process from the camera cache to the memory matrix. At the same time, the selection of the Gaussian convolution kernel size (such as 5×5 or 7×7) and standard deviation parameter must refer to the definition of image sharpness in the ISO 12233 resolution test standard, and be matched according to the ratio of the minimum width of the expected crack on the mulch film surface (such as 0.5 mm) to the spatial resolution of the image, to ensure that the smoothing operation does not eliminate the actual micro-crack features.

[0021] In step S12, based on the smoothed image, contrast enhancement and color distribution parameters are extracted to identify color anomaly regions whose color distribution parameters meet a preset degradation significance threshold. Connected component filtering and boundary filling are then performed on these color anomaly regions to obtain a degradation region mask. This includes: constructing an original grayscale histogram and establishing a grayscale mapping lookup table based on preset red, green, and blue channels using the smoothed image to obtain enhanced image data; calculating color offset and brightness attenuation based on the enhanced image data to obtain a color difference feature matrix; determining the corresponding image region as a color anomaly region if the brightness attenuation is greater than a preset darkness threshold or the color offset is greater than a preset yellowing threshold based on the color difference feature matrix; and merging connected components and extracting contours and filling the interior based on preset area ratio conditions using the color anomaly regions to obtain a degradation region mask.

[0022] First, based on the smoothed image, an original grayscale histogram is constructed and a grayscale lookup table is established using preset red, green, and blue three channels. Specifically, the smoothed image is separated into three single-color channels: red, green, and blue. For each pixel in the image, its red channel value is multiplied by a weight of 0.299, its green channel value by a weight of 0.587, and its blue channel value by a weight of 0.114. These three products are then added together and rounded to obtain the single-channel grayscale value of that pixel. The entire image is then traversed to count the number of pixels at each grayscale level from 0 to 255. This count is divided by the total number of pixels in the image to obtain the probability of each grayscale level. Then, the pixels are sorted in ascending order of grayscale level. The probability of occurrence of the current gray level and all previous gray levels is accumulated to obtain the cumulative distribution probability of the current gray level. The cumulative distribution probability of each gray level is multiplied by the maximum value of the gray level, 255, and rounded to the nearest integer to obtain the new mapping value corresponding to that gray level. This establishes a gray-level mapping lookup table (LUT) that corresponds one-to-one between the original input gray levels and the new output gray levels. Then, the gray-level mapping lookup table is applied back to the original red, green, and blue three-channel data. That is, each pixel value of each channel of the smoothed image is traversed, and the corresponding new value is retrieved in the lookup table as an index and replaced in place. This achieves synchronous contrast stretching of the three channels to obtain enhanced image data.

[0023] Subsequently, based on the enhanced image data, color shift and brightness attenuation are calculated. Using pixels or local image blocks of a set size (e.g., 16×16 pixels) as units, the current average brightness value and average chromaticity values ​​of the red, green, and blue channels are obtained. These are then compared with the corresponding reference values ​​of a pre-stored, unaged standard mulch film image. This unaged standard mulch film image is a reference image acquired and stored during the initial stage of mulch film installation or using a brand-new sample of the same model under the same light intensity and shooting height. Next, the brightness attenuation is obtained by subtracting the current area's average brightness value from the standard reference brightness value. The color shift is then calculated using the chromaticity values ​​of each channel in the current area and the standard reference chromaticity value. Specifically, a three-dimensional color vector is constructed, and the Euclidean distance between the current pixel and the reference value is calculated, which is then used as the color shift. Alternatively, the image can be converted to the CIELAB color space, and the color difference value is calculated as the color shift. After traversing the entire image, these difference values ​​are combined to form a color difference feature matrix. Subsequently, based on the color difference feature matrix, a threshold determination operation based on a preset darkness threshold and a preset yellowing threshold is performed. Each element in the matrix is ​​traversed. If the brightness attenuation value corresponding to a certain position is greater than the preset darkness threshold (e.g., gray level difference value of 50), or its color offset value is greater than the preset yellowing threshold (e.g., chromaticity distance of 15), it indicates that the area has undergone blackening or yellowing caused by photo-oxidative aging. Then, the position is marked as a candidate abnormal point, thereby determining the color abnormal area.

[0024] It should be noted that the preset darkness threshold and yellowing threshold are based on the testing specifications for light transmittance and haze of mulch film in industry standards such as GB / T 35795 "Fully Biodegradable Agricultural Ground Cover Film" or GB / T 13735, combined with historical field experimental data. Specifically, multiple sets of mulch film sample images at different aging stages (such as 30 days, 60 days, and 90 days of coverage) are collected, and the color difference distribution and brightness difference distribution of non-aged samples and severely aged samples in Lab color space are statistically analyzed. The statistical boundary point that can distinguish between normal and aged samples (such as the lower limit of the 95% confidence interval) is selected as the initial threshold, and it is dynamically corrected according to the light intensity change curve of the crop growth cycle in actual application.

[0025] Finally, based on the color anomaly regions, connected component merging and contour extraction and internal filling based on preset area ratio conditions are performed. The 8-neighborhood connectivity algorithm is used to scan all candidate anomalies, merging spatially adjacent or contacting anomalies into an independent connected component object. The proportion of the total number of pixels contained in each connected component object to the total number of pixels in the image is calculated. If this proportion is less than the preset area ratio threshold (e.g., 0.8%), it is considered noise and is removed, retaining the connected components that meet the conditions. Then, morphological closing operations are used to smooth the boundaries of the retained connected components and fill the empty areas inside the connected components, finally generating a binarized degraded region mask, where the degraded region is marked as a valid value (e.g., 1) and the background region is marked as an invalid value (e.g., 0).

[0026] It should be noted that the weighting of the red channel (0.299), green channel (0.587), and blue channel (0.114) is based on the brightness formula for human visual characteristics in the ITU-R BT.601 standard. Since the human eye is most sensitive to green light, followed by red light, and least sensitive to blue light, this weighting ratio can most realistically reflect the human eye's perception of changes in brightness on the surface of the mulch film, ensuring that the grayscale image retains the main aging texture features. The preset area percentage threshold (e.g., 0.05% to 0.1%) is based on the statistical analysis of the size of non-aging disturbances such as dust particles and flying insects in the field. Specifically, by collecting field images containing typical disturbances, statistically analyzing the area distribution histogram of the connected regions of the disturbances, and selecting the upper limit of the value covering 99% of the disturbance area as the threshold, any area smaller than this threshold is judged as environmental noise rather than aging patches of the mulch film itself.

[0027] It is worth noting that the structuring element used in the morphological closing operation is usually set as a 3×3 or 5×5 pixel circular or square matrix. The size selection needs to refer to the average width of the cracks on the mulch film surface. If the structuring element is too large, it may cause adjacent independent aging areas to be incorrectly merged. If it is too small, it will be unable to repair tiny crack breaks. Those skilled in the art can determine the optimal structuring element size based on the resolution of the actual acquired image (e.g., each pixel represents 0.1 mm) and the physical characteristics of the aging cracks in the mulch film (e.g., the crack width is usually above 0.5 mm) to ensure the geometric accuracy of the mask in the deteriorated area.

[0028] In step S13, based on the degraded region mask and combined with preset standard texture values, local grayscale data extraction and texture roughness quantification analysis are performed to obtain a roughness score. This includes: extracting local regions from the smoothed image based on the degraded region mask to obtain a local grayscale data matrix; generating a grayscale co-occurrence matrix and extracting energy eigenvalues ​​and moments of inertia values ​​based on the local grayscale data matrix to obtain texture feature parameters; and performing weighted summation and numerical deviation calculation based on the texture feature parameters and the preset standard texture values ​​to obtain a roughness score.

[0029] First, based on the deteriorated area mask, local regions are extracted from the smoothed image to obtain a local grayscale data matrix. Specifically, the deteriorated area mask is used as a binary gating filter and a bitwise AND operation is performed with the smoothed image. Only the grayscale pixel values ​​at the corresponding positions in the mask that are marked as valid are retained, while the pixel values ​​of the background area are set to zero or marked as invalid data, thereby constructing a local grayscale data matrix that only contains the aging feature areas of the plastic film.

[0030] Next, based on the local grayscale data matrix, a grayscale co-occurrence matrix is ​​generated, and energy eigenvalues ​​and moment of inertia values ​​are extracted to obtain texture feature parameters. Specifically, a distance step size (e.g., 1 pixel) and a scanning direction (e.g., 0 degrees horizontal direction) are set to define the spatial relationship between pixels. A square matrix with dimensions equal to the number of grayscale levels in the image is constructed. The row index of the square matrix represents the grayscale value of the reference pixel, and the column index represents the grayscale value of the neighboring pixels. Each valid pixel in the local grayscale data matrix is ​​traversed to find its neighboring pixels at a specified distance and direction. If the grayscale value of the reference pixel is A and the grayscale value of the neighboring pixel is B, then the element in the A-th row and B-th column of the square matrix is... Increment the value by one; after traversal, divide the value of each element in the matrix by the total number of pixel pairs to obtain the normalized gray-level co-occurrence matrix. Each value in this matrix represents the joint probability of the corresponding gray-level pair appearing simultaneously in space. Calculate the energy eigenvalue based on this matrix, which is to sum the squares of the values ​​of all elements in the matrix to characterize the uniformity of the texture distribution. Simultaneously, calculate the moment of inertia value, which is to traverse each element of the matrix, calculate the difference between the row index and column index of the element, square the difference, multiply the squared value by the probability value of the element itself, and finally sum all the product results to characterize the sharpness and contrast of the texture grooves.

[0031] Finally, based on the texture feature parameters and the preset standard texture values, a weighted summation and numerical deviation calculation are performed to obtain the roughness score. Specifically, the calculated current energy feature value and current moment of inertia value are first Z-score standardized or max-min normalized to map them to the same dimension range (e.g., 0 to 1). Then, the absolute value of the difference between the standardized current moment of inertia and the standardized preset standard moment of inertia is calculated to obtain the normalized moment of inertia deviation. The absolute value of the difference between the standardized current energy feature value and the standardized preset standard energy feature value is also calculated to obtain the normalized energy deviation. The normalized moment of inertia deviation is multiplied by a first weighting coefficient (e.g., 0.7) and then by a second weighting coefficient (e.g., 0.3). These two products are then added to obtain the final roughness score. This normalization step eliminates the feature overload problem caused by the large difference in magnitude between the energy feature value (usually less than 1) and the moment of inertia value (usually greater than 100), ensuring the scientific validity of multi-dimensional feature fusion.

[0032] It should be noted that the preset standard texture values ​​are derived from the average gray-level co-occurrence matrix features obtained by photographing and calculating brand-new, unlaid mulch film of the same model and batch under a standard D65 light source environment using the same acquisition equipment. In addition, before constructing the gray-level co-occurrence matrix, in order to reduce the computational complexity of the sparse matrix and improve noise resistance, the gray levels of the local gray-level data matrix are usually compressed and quantized from the original 256 levels to 32 or 16 levels. At the same time, in order to eliminate the influence of the shooting angle on the texture directionality, the gray-level co-occurrence matrices in four directions of 0 degrees, 45 degrees, 90 degrees, and 135 degrees are usually calculated separately, and the arithmetic mean of the energy feature values ​​and moment of inertia values ​​calculated in these four directions is taken as the final texture feature parameters used for calculation. The rules for setting the first and second weighting coefficients are based on the Pearson correlation analysis results between the characteristic parameters and the physical tensile strength of the mulch film. Specifically, a batch of mulch film samples with different aging degrees are collected, and their physical tensile strength, moment of inertia deviation, and energy deviation are measured respectively. The correlation coefficient between each deviation and the tensile strength value is calculated. Since cracks caused by aging of the mulch film will cause drastic changes in texture contrast, the moment of inertia feature usually has a stronger positive correlation with the degree of aging (the correlation coefficient is usually greater than 0.8), while the correlation of the energy feature is relatively weak. Therefore, the weights are allocated according to the proportion of the correlation coefficient. For example, the first weighting coefficient is set to 0.7 and the second weighting coefficient is set to 0.3 to ensure that the roughness score can more sensitively reflect the crack features that have a greater impact on the strength of the mulch film.

[0033] In step S14, based on the roughness score and the color anomaly region, multidimensional feature vectors are constructed and clustered to obtain aging level partitions. These aging level partitions are then mapped onto the smoothed image to obtain a comprehensive aging map. This includes: performing pixel spatial distribution density statistics based on the color anomaly region to obtain pixel distribution density data; combining the pixel distribution density data and the roughness score using multidimensional data to obtain multidimensional feature vectors; performing iterative clustering calculations based on the multidimensional feature vectors using a preset K-Means algorithm until a preset convergence condition is met to obtain aging level labels; and generating partition masks and mapping aging correlation weights within partitions based on the aging level labels to obtain a comprehensive aging map.

[0034] First, based on the color anomaly regions, pixel spatial distribution density statistics are performed to obtain pixel distribution density data. Specifically, a sliding window of the same size as the texture analysis step (e.g., a 32×32 pixel block) is used to traverse the degraded region mask. The number of pixels marked as anomalies (i.e., a value of 1) within each window is counted, and this number is divided by the total number of pixels within the window to obtain the color anomaly pixel density value of that window region. This value is between 0 and 1, representing the density of macroscopic color degradation. Next, based on the pixel distribution density data and the roughness score, multidimensional data is combined to obtain a multidimensional feature vector. Since the density data (0-1 interval) and the roughness score (usually a larger value) have different dimensions, they need to be normalized using the Z-Score standardization formula to adjust their mean to 0 and their standard deviation to 1. Then, the normalized density value and the roughness score are concatenated to construct a two-dimensional feature vector V=[Normalized_Density,Normalized_Roughness] corresponding to each local window.

[0035] Subsequently, based on the multidimensional feature vectors, clustering iterations are performed using the preset K-Means algorithm until the preset convergence condition is met. The number of clusters k is set (e.g., 3, corresponding to mild, moderate, and severe aging). k feature vectors are randomly selected as initial cluster centers. The Euclidean distance between the feature vector of each window and each cluster center is calculated, and the feature vector is assigned to the nearest cluster. The arithmetic mean of all vectors in each cluster is recalculated as the new cluster center. The above allocation and update process is repeated until the offset of the cluster center position in the two iterations is less than the preset convergence threshold. At this point, the iteration is stopped and the aging level label (e.g., 0, 1, 2) corresponding to each window is output.

[0036] Finally, based on the aging level labels, a partition mask is generated and an aging correlation weight mapping is performed within each partition to obtain a comprehensive aging map. Specifically, the magnitude (i.e., the combined strength of density and roughness) of each cluster center vector is calculated. The label corresponding to the cluster center with the largest magnitude is defined as "severe aging" and assigned the highest aging correlation weight (e.g., 1.0). The label corresponding to the cluster center with the smallest magnitude is "mild aging" and assigned the lowest weight (e.g., 0.2). Each window is filled with the corresponding weight value according to its label and mapped back to a matrix of the same size as the original smoothed image. The boundary between windows is smoothly transitioned using bilinear interpolation, and finally a gray-level comprehensive aging map reflecting the spatial distribution of the aging degree of the entire field is generated.

[0037] It should be noted that the preset convergence threshold is based on the gradient analysis of the clustering loss function (i.e., the sum of squared errors within clusters, SSE) as a function of the number of iterations. During system initialization, a representative set of aging plastic film images is selected to construct a calibration dataset. The K-Means algorithm is run, and the centroid displacement is recorded after each iteration. A decay curve of "iteration number - centroid displacement" is plotted. The inflection point where the rate of change of the slope on the curve approaches zero is identified, and the displacement value corresponding to this inflection point (e.g., ...) is recorded. () is used as the convergence threshold.

[0038] Furthermore, the preset number of clusters k for the K-Means algorithm is based on the classification description of residual mulch film morphology in GB / T 25413 "Limits and Determination of Residual Mulch Film in Farmland". The aging degree is typically divided into three typical stages: "intact", "surface cracked", and "broken and scattered". In addition, considering that the mulch film in the field may be in a single aging stage (e.g., all new film), if the number of samples in a certain cluster is zero or extremely small (e.g., less than 1% of the total samples) during clustering, the system will automatically trigger a K-value adaptive mechanism to reduce the number of clusters k (e.g., to k=2 or k=1) to avoid misjudgments caused by forcibly segmenting similar features. The convergence condition is set to balance computational accuracy and system response speed. Through analysis of convergence curves on historical datasets, the value corresponding to the inflection point where the slope of the loss function tends to flatten is selected as the convergence threshold. The assignment of aging-related weights within a zone is determined based on the principal component analysis (PCA) results of the eigenvectors. In the specific implementation, the Mahalanobis distance of the center point of each cluster relative to the origin (i.e., the non-aging state) in the density-roughness feature space is calculated. This distance directly reflects the degree to which the zone deviates from the normal mulch film state. The system directly assigns the normalized distance value as a weight to the corresponding aging level zone.

[0039] In step S15, based on the comprehensive aging map, the overall intensity reduction ratio is calculated. If the overall intensity reduction ratio is higher than a preset intensity reduction threshold, the risk estimation of reclaimed fragments and the integration accumulation of potential soil pollution levels are performed based on the smoothed image to obtain the final aging quantification index. This includes: based on the comprehensive aging map, performing pixel grayscale mean statistics and attenuation percentage calculation to obtain the overall intensity reduction ratio; if the overall intensity reduction ratio is higher than the preset intensity reduction threshold, performing edge tear fragment number statistics and length detection based on the smoothed image to obtain the reclaimed fragment risk value; based on the reclaimed fragment risk value, performing the specific surface area increment estimation of plastic film fragments and the conversion of chemical additive leaching amount per unit soil area to obtain the single-area pollution level; based on the single-area pollution level, performing full-area integration accumulation calculation to obtain the final aging quantification index.

[0040] First, based on the comprehensive aging map, pixel grayscale average values ​​are statistically analyzed and attenuation percentages are calculated to obtain the overall intensity reduction ratio. Specifically, each pixel in the comprehensive aging map is traversed, and the arithmetic mean of all pixel grayscale values ​​is calculated to obtain the current grayscale average value. This value is then compared with the preset initial state mulch film grayscale average value. The specific calculation process is as follows: the initial state mulch film grayscale average value is subtracted from the current grayscale average value, the difference is divided by the initial state mulch film grayscale average value, and the quotient is multiplied by 100% to obtain the overall intensity reduction ratio. This ratio directly reflects the overall degree of decrease in light transmittance and increase in surface haze caused by aging of the mulch film.

[0041] Next, a threshold determination is performed. If the overall strength reduction ratio is lower than or equal to the preset strength reduction threshold, it is determined that although the mulch film is aging, its structure is still intact and has not yet reached the risk stage of generating microplastic fragments. In this case, the risk value of the recycled fragments is directly set to zero, and the subsequent fragment statistics steps are skipped, directly entering the final index synthesis stage. If the overall strength reduction ratio is higher than the preset strength reduction threshold, the number and length of edge-torn fragments are counted and detected based on the smoothed image. Specifically, the Canny edge detection operator is used to perform a second scan on the smoothed image to extract all edge contours and calculate the compactness of each contour, that is, calculate the square of the contour perimeter and divide it by the contour area. If the compactness of a contour is greater than the preset shape factor threshold, it is determined to be a torn fragment. The total number of torn fragments is counted, and the total pixel path length of all torn edges is calculated. Then, the risk value of the recycled fragments is calculated, that is, the total number of torn fragments is multiplied by the first normalized weight coefficient, the total pixel path length is multiplied by the second normalized weight coefficient, and the two products are added together to obtain the risk value of the recycled fragments.

[0042] It should be noted that the preset strength reduction threshold is based on the aging retention rate requirements for "nominal strain at break" and "tensile load" in GB / T 13735-2017 "Agricultural Polyethylene Blown Ground Cover Film". Generally, when photoaging causes the mechanical properties of the material to decrease to 50% of the initial value (i.e., the reduction ratio reaches 50%), the mulch film can no longer be completely recycled by mechanical means. Therefore, in this embodiment, combined with field measurement data, the image grayscale attenuation ratio corresponding to the critical point of mechanical properties (e.g., 35% to 40%) is set as the strength reduction threshold. This threshold can be obtained by conducting tensile tests and image grayscale correlation analysis on mulch film samples with different degradation cycles.

[0043] Subsequently, based on the risk values ​​of the recovered fragments, the increase in the specific surface area of ​​the plastic film fragments is calculated and converted into the leaching amount of chemical additives per unit soil area. Specifically, based on geometric assumptions, the torn edge is considered as a newly exposed lateral area, and the increase in specific surface area is calculated. This is achieved by multiplying the total length of the pixel path by the nominal thickness of the plastic film, where the nominal thickness of the plastic film is derived from the specifications on the outer packaging of the plastic film product or the nominal value specified in GB 13735 standard (e.g., 0.01 mm). Then, the pollution level of a single area is calculated by multiplying the increase in specific surface area by the leaching rate coefficient per unit area, and then by the initial areal density of the chemical additives added to the plastic film. The leaching rate coefficient per unit area is defined as the rate constant of additive migration into the soil per unit time per unit exposed area under specific soil moisture and pH conditions. The initial added areal density is the mass of chemical additives contained in a unit area of ​​the mulch film, which can be calculated by multiplying the nominal thickness of the mulch film, the material density, and the initial added concentration, with the unit being milligrams per square millimeter. The initial added concentration is the mass percentage of plasticizers (such as phthalates) or antioxidants added during the production of the mulch film. Specifically, samples of mulch film from the same batch are taken, cut into standard sizes, and placed in a standard soil extract with preset humidity and pH values. An accelerated aging soaking experiment is conducted under constant temperature conditions. The concentration of the target chemical additive (such as phthalates) in the extract is periodically sampled and tested. The leaching rate coefficient per unit area is calculated by dividing the product of the concentration and the volume of the extract by the product of the exposed area of ​​the mulch film sample and the soaking time. The final value is the average value of multiple experiments during the stable phase. This coefficient can be pre-established in a database for different typical soil types (such as clay, loam, and sandy soil) and climatic regions, and the corresponding reference value can be called according to the farmland geographical information during actual testing. The nominal thickness and material density can be obtained from the Material Safety Data Sheet (MSDS) or product specification sheet provided by the mulch film manufacturer.

[0044] Finally, based on the pollution levels of each individual zone, a total area integral summation is performed to obtain the final aging quantification index. Specifically, the pollution levels of each individual zone in all monitored areas are summed to obtain the total pollution amount for the entire area. The total pollution amount for the entire area is mapped to a pollution risk score between 0 and 100 using the Sigmoid function or the maximum value normalization method. At the same time, the overall intensity reduction ratio is also normalized to the range of 0 to 100. The normalized pollution risk score and the normalized overall intensity reduction ratio are then weighted and combined, that is, multiplied by the corresponding preset influence factors (e.g., physical performance weight 0.6, environmental risk weight 0.4) and then added together to obtain a final aging quantification index that comprehensively reflects the physical intensity decay and chemical pollution risk.

[0045] Specifically, the preset influencing factors are used to balance the weights of physical performance degradation and environmental pollution risk in the final evaluation. Their setting is based on statistical analysis of historical plastic film recycling cases and the expert Delphi method. Data from multiple sets of plots that have experienced breakage and pollution are collected, recording their image characteristics before failure, i.e., the proportion of strength reduction, and the pollution severity level assessed afterward. Logistic regression analysis is used to determine the correlation weight between the proportion of strength reduction and the pollution severity level. For example, the analysis may show that mechanical strength failure is a necessary but not sufficient condition for pollution, therefore assigning it a higher weight (e.g., 0.6), while the pollution risk weight is correspondingly 0.4. These influencing factors can be fine-tuned before implementation based on local policy priorities (whether prioritizing recycling rates or pollution control).

[0046] It should be noted that the preset initial state grayscale average value of the mulch film is based on benchmark measurements of brand new mulch film of the same brand, specification, and batch. Specifically, multiple images of brand new mulch film are taken using the same field acquisition equipment in the early stage of mulch film laying in the field or in a standard light source box (D65 light source) environment in the laboratory, and the arithmetic mean of the grayscale values ​​of these images is calculated and stored as a fixed benchmark constant in the system database. The first normalized weight coefficient and the second normalized weight coefficient are based on the analytic hierarchy process (AHP) assessment model of the difficulty of mulch film recycling. The number of fragments directly affects the time efficiency of manual picking, while the fragment length (i.e., the degree of fragmentation) is directly related to the probability of microplastic formation. For example, the first normalized weight coefficient is set to 0.4 (emphasizing the number), the second normalized weight coefficient is set to 0.6 (emphasizing the degree of microplasticization), and the sum of the two coefficients is 1.

[0047] In step S16, the final aging quantization index is input into a pre-trained neural network verification model for deviation calibration to obtain the target aging detection result. This includes: inputting the final aging quantization index into a pre-built convolutional neural network model for feature mapping to obtain a predicted aging value; calculating the difference deviation value between the predicted aging value and the final aging quantization index to obtain a predicted deviation value; and performing a weighted correction on the final aging quantization index based on the predicted deviation value to obtain the target aging detection result.

[0048] First, based on the final aging quantization index, it is input into a pre-trained convolutional neural network validation model for feature mapping to obtain the predicted aging value. Specifically, the final aging quantization index calculated in step S15 is concatenated with the multidimensional feature vector (containing pixel distribution density data and coarsening score) generated in step S14 to construct a high-dimensional input tensor. Specifically, the final aging quantization index (scalar) calculated in step S15 is dimensionally expanded so that it has the same time step or feature dimension as the multidimensional feature vector (sequence data) generated in step S14, or the feature vector is first processed through a flattening layer and then concatenated with the aging quantization index before a fully connected layer. The high-dimensional input tensor is fed into the input layer of the convolutional neural network model, and features are extracted sequentially through several one-dimensional convolutional layers. A sliding operation is performed on the feature sequence using a convolutional kernel of a preset size to capture the local correlation between different feature components. Then, a max pooling layer is used for downsampling to reduce the data dimensionality and retain significant features. Finally, a fully connected layer maps the extracted abstract features to a regression output, which is the predicted aging value. This value represents the theoretical aging degree inferred by the model based on historical data patterns.

[0049] It should be noted that the training dataset for the neural network validation model comes from a hybrid database of laboratory accelerated aging experiments and field in-situ monitoring. The input data consists of the quantization index and feature vector calculated by the method of this invention, and the label data is the true aging degree converted from the retention rate of the film fracture elongation measured by the standard mechanical tensile test (according to GB / T 1040.3 standard). The model architecture uses a lightweight one-dimensional convolutional neural network (1D-CNN), with specific parameters including two convolutional layers (kernel size set to 3, filter numbers of 32 and 64 respectively, and ReLU activation function), one max pooling layer (stride of 2), and two fully connected layers (number of neurons of 128 and 1 respectively). During training, mean squared error (MSE) is used as the loss function, and the Adam optimizer (learning rate set to 0.001, batch size of 32) is used for backpropagation to iteratively update the weights until the loss value on the validation set no longer decreases or reaches the preset number of iterations (e.g., 500 iterations), completing the model training and saving the optimal weight parameters.

[0050] Next, based on the predicted aging value and the final aging quantification index, a difference deviation value is calculated to obtain the prediction deviation value. Specifically, the predicted aging value is subtracted from the final aging quantification index to obtain an algebraic difference value with positive and negative signs, i.e., the prediction deviation value. This value reflects the deviation of the index calculated based on physical rules from the big data experience model. Finally, based on the prediction deviation value, the final aging quantification index is weighted and corrected to obtain the target aging detection result. Specifically, the prediction deviation value is multiplied by a preset calibration trust weight coefficient to obtain a correction increment. The final aging quantification index is added to the correction increment to obtain the target aging detection result that has undergone double verification.

[0051] It should be noted that the calibration trust weight coefficient is set based on the evaluation of the model's prediction accuracy on the validation set. Specifically, the root mean square error (RMSE) of the model on the validation set is calculated. If the RMSE is small, it indicates that the model prediction is very accurate, and a higher calibration trust weight coefficient is assigned (e.g., 0.8), meaning that the final result mainly refers to the model's prediction. If the RMSE is large, it indicates that the model has high uncertainty in the current scenario, and a lower weight coefficient is assigned (e.g., 0.2), meaning that the final result mainly maintains the original physical calculation value. This dynamic weight mechanism ensures that the system can still degenerate into rule-based conservative detection in extreme cases where there are insufficient training samples (such as rare lighting environments), thus guaranteeing the robustness and safety of the system.

[0052] In summary, this invention achieves accurate quantification of the degree of aging of plastic film and prediction of pollution risk by denoising preprocessing of field images and extracting multidimensional features of color and texture, integrating K-means clustering to construct a comprehensive aging map, and combining convolutional neural networks for verification training.

[0053] Referring to Figure 2, a second embodiment of the present invention provides an aging detection system for high-strength agricultural mulch film, comprising: an image acquisition and preprocessing module for acquiring image data of the mulch film surface, performing high-frequency noise suppression and edge preservation smoothing on the mulch film surface image data to obtain a smoothed image; a feature extraction and mask generation module for performing contrast enhancement and extracting color distribution parameters based on the smoothed image, identifying color anomaly regions whose color distribution parameters meet a preset degradation significance threshold, and performing connected component filtering and boundary filling on the color anomaly regions to obtain a degradation region mask; and a texture analysis module for performing local grayscale data extraction and quantitative analysis of texture roughness based on the degradation region mask and a preset standard texture value to obtain... The system includes a roughness score module; a clustering and map generation module, used to construct and cluster multi-dimensional feature vectors based on the roughness score and the color anomaly areas to obtain aging level partitions, and map the aging level partitions onto the smoothed image to obtain a comprehensive aging map; an intensity assessment and calculation module, used to calculate the overall intensity reduction ratio based on the comprehensive aging map, and if the overall intensity reduction ratio is higher than a preset intensity reduction threshold, then based on the smoothed image, it performs risk estimation of recyclable debris and accumulation of potential soil pollution levels to obtain a final aging quantification index; and a model verification module, used to input the final aging quantification index into a pre-trained neural network verification model for bias calibration to obtain the target aging detection result.

[0054] It should be noted that the aging detection system for high-strength agricultural mulch film provided in this embodiment of the invention is used to perform all the process steps of the aging detection method for high-strength agricultural mulch film in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0055] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0056] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for testing the aging of high-strength agricultural mulch film, characterized in that, include: The process involves acquiring surface image data of the plastic film, performing high-frequency noise suppression and edge-preserving smoothing on the image data to obtain a smoothed image, enhancing contrast and extracting color distribution parameters based on the smoothed image, identifying color anomaly regions whose color distribution parameters meet a preset degradation significance threshold, and performing connected component filtering and boundary filling on these color anomaly regions to obtain a degradation region mask, extracting local grayscale data and performing texture roughness quantification analysis based on the degradation region mask and a preset standard texture value to obtain a roughness score, constructing multidimensional feature vectors and performing clustering based on the roughness score and the color anomaly regions to obtain aging level partitions, and mapping these aging level partitions onto the smoothed image to obtain a comprehensive aging map, and calculating the overall intensity reduction ratio based on the comprehensive aging map. If the overall intensity reduction ratio is higher than a preset intensity reduction threshold, the risk estimation of recyclable debris and the accumulation of potential soil pollution levels are performed based on the smoothed image to obtain the final aging quantification index. The final aging quantification index is input into a pre-trained neural network verification model for bias calibration to obtain the target aging detection result.

2. The aging test method for high-strength agricultural mulch film according to claim 1, characterized in that, The process of acquiring mulch film surface image data and performing high-frequency noise suppression and edge preservation smoothing on the mulch film surface image data to obtain a smoothed image includes: acquiring mulch film surface image data using field acquisition equipment; performing two-dimensional pixel matrix analysis and high-frequency abrupt region identification based on the mulch film surface image data to obtain the high-frequency noise region distribution; calculating the weighted average value of surrounding adjacent pixels and replacing values ​​based on the high-frequency noise region distribution to obtain preliminary noise reduction data; and performing edge locking and Gaussian convolution smoothing operations based on the preliminary noise reduction data to obtain a smoothed image.

3. The aging test method for high-strength agricultural mulch film according to claim 1, characterized in that, The process of enhancing contrast and extracting color distribution parameters based on the smoothed image, identifying color anomaly regions whose color distribution parameters meet a preset degradation significance threshold, and performing connected component filtering and boundary filling on the color anomaly regions to obtain a degradation region mask includes: constructing an original grayscale histogram and establishing a grayscale mapping lookup table based on preset red, green, and blue channels based on the smoothed image to obtain enhanced image data; calculating color offset and brightness attenuation based on the enhanced image data to obtain a color difference feature matrix; determining the corresponding image region as a color anomaly region if the brightness attenuation is greater than a preset darkness threshold or the color offset is greater than a preset yellowing threshold based on the color difference feature matrix; and performing connected component merging and contour extraction and internal filling based on preset area ratio conditions on the color anomaly regions to obtain a degradation region mask.

4. The aging test method for high-strength agricultural mulch film according to claim 1, characterized in that, The step of extracting local grayscale data and performing quantitative analysis of texture roughness based on the degraded region mask and a preset standard texture value to obtain a roughness score includes: extracting local regions from the smoothed image based on the degraded region mask to obtain a local grayscale data matrix; generating a grayscale co-occurrence matrix and extracting energy eigenvalues ​​and moments of inertia values ​​based on the local grayscale data matrix to obtain texture feature parameters; and performing weighted summation and numerical deviation calculation based on the texture feature parameters and the preset standard texture value to obtain a roughness score.

5. The aging test method for high-strength agricultural mulch film according to claim 3, characterized in that, The step of constructing and clustering multidimensional feature vectors based on the roughness score and the color anomaly regions to obtain aging level partitions, and mapping the aging level partitions onto the smoothed image to obtain a comprehensive aging map, includes: performing pixel spatial distribution density statistics based on the color anomaly regions to obtain pixel distribution density data; combining the pixel distribution density data and the roughness score into multidimensional data to obtain multidimensional feature vectors; performing iterative clustering calculations based on the multidimensional feature vectors using a preset K-Means algorithm until a preset convergence condition is met to obtain aging level labels; and generating partition masks and mapping aging correlation weights within partitions based on the aging level labels to obtain a comprehensive aging map.

6. The aging test method for high-strength agricultural mulch film according to claim 1, characterized in that, The process involves calculating the overall intensity reduction ratio based on the comprehensive aging map. If the overall intensity reduction ratio exceeds a preset intensity reduction threshold, the process then uses the smoothed image to perform risk estimation of reclaimed fragments and integral accumulation of potential soil pollution levels to obtain the final aging quantification index. This includes: calculating the pixel grayscale mean and attenuation percentage based on the comprehensive aging map to obtain the overall intensity reduction ratio; if the overall intensity reduction ratio exceeds the preset intensity reduction threshold, calculating the number and length of edge-torn fragments based on the smoothed image to obtain a reclaimed fragment risk value; calculating the increase in the specific surface area of ​​plastic film fragments and converting the leaching amount of chemical additives per unit soil area based on the reclaimed fragment risk value to obtain the pollution level of a single area; and performing an integral accumulation calculation across the entire region based on the pollution level of the single area to obtain the final aging quantification index.

7. The aging test method for high-strength agricultural mulch film according to claim 1, characterized in that, The step of inputting the final aging quantification index into a pre-trained neural network verification model for deviation calibration to obtain the target aging detection result includes: inputting the final aging quantification index into a pre-built convolutional neural network model for feature mapping to obtain a predicted aging value; calculating the difference deviation value between the predicted aging value and the final aging quantification index to obtain a predicted deviation value; and performing a weighted correction on the final aging quantification index based on the predicted deviation value to obtain the target aging detection result.

8. An aging detection system for high-strength agricultural mulch film, characterized in that, include: The image acquisition and preprocessing module is used to acquire image data of the mulch film surface, perform high-frequency noise suppression and edge preservation smoothing on the mulch film surface image data to obtain a smooth image; the feature extraction and mask generation module is used to perform contrast enhancement and extract color distribution parameters based on the smooth image, identify color abnormal regions whose color distribution parameters meet a preset degradation significance threshold, and perform connected component filtering and boundary filling on the color abnormal regions to obtain a degradation region mask; The texture analysis module is used to extract local grayscale data and perform quantitative analysis of texture roughness based on the deteriorated area mask and the preset standard texture value to obtain a roughness score. The clustering and map generation module is used to construct and cluster based on the roughness score and the color anomaly area to obtain aging level partitions, and map the aging level partitions onto the smoothed image to obtain a comprehensive aging map; the intensity assessment and calculation module is used to calculate the overall intensity reduction ratio based on the comprehensive aging map. If the overall intensity reduction ratio is higher than the preset intensity reduction threshold, the module performs risk estimation of recyclable debris and accumulation of potential soil pollution level based on the smoothed image to obtain the final aging quantification index. The model validation module is used to input the final aging quantification index into a pre-trained neural network validation model for bias calibration, thereby obtaining the target aging detection result.

Images