A method and system for monitoring the quality of cultivated land

Abstract

The application relates to the technical field of image data processing, in particular to a farmland quality monitoring method and system. The method comprises the following steps: acquiring an overhead image of a farmland to be monitored, performing connected domain division on the overhead image to obtain a plurality of connected domains, and obtaining a regularity degree value of the connected domains; determining a complexity degree value of the connected domains; determining a feature value of the connected domains according to the regularity degree value and the complexity degree value of the connected domains; for any two different pixel points in the overhead image, determining a feature distance between the two pixel points according to the difference between the two pixel points in the feature value of the connected domain to which the two pixel points belong; and clustering the pixel points in the overhead image by using the feature distance between the different pixel points in the overhead image to obtain a farmland segmentation result, so as to monitor the farmland to be monitored according to the plot segmentation result of the farmland to be monitored at different time points. Through the above technical scheme, the monitoring of the farmland can be more accurately realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, and particularly relates to a farmland quality monitoring method and system. BACKGROUND

[0002] With the population growth and the acceleration of urbanization process, farmland resources are increasingly scarce. Monitoring farmland helps to maintain the quantity and quality of farmland, so as to ensure the stability of food production. For example, it can find the problems of degradation or pollution existing in farmland in time, or it can understand the growth of crops in time.

[0003] For example, in order to realize the monitoring of farmland, a farmland protection scene data processing method is provided in the Chinese patent application file with the publication number CN118762215A. The method includes: through the identification and feature extraction of farmland images, the farmland can be divided into multiple farmland protection areas, and then the respective monitoring and the respective collection of farmland scene data can be performed, so as to effectively reduce the complexity of collected data. In addition, the protection levels of each farmland protection area are finely divided, so as to adjust the data collection frequency of different farmland protection areas, and to strengthen the protection and management of the farmland protection area after the protection level of the farmland protection area changes.

[0004] When the farmland is divided into regions, the farmland actually belonging to different regions may be divided into the same region, or the farmland actually belonging to the same region may be divided into different regions, thereby affecting the monitoring result of the farmland.

[0005] For example, when the division result of the overhead image of the farmland is used to monitor the farmland area or the crop area in the farmland, if there is improper division of the overhead image, the monitoring result of the farmland area or the crop area will be affected. Therefore, the related technology is difficult to realize effective monitoring of the farmland. SUMMARY

[0006] In order to overcome the problem that the related technology is difficult to realize effective monitoring of the farmland, the present application provides a farmland quality monitoring method and system.

[0007] According to a first aspect of the embodiment of the present application, a method for monitoring farmland quality is provided, which comprises: obtaining an overhead image of a farmland to be monitored, and performing connected domain division on the overhead image to obtain a plurality of connected domains, so as to obtain a regularity degree value of the connected domains; the regularity degree value is used to represent the continuity of the shape of the edge line in the connected domain; determining a complexity degree value of the connected domain, which is used to represent the difference degree of the complexity of the connected domain and the overhead image; determining a feature value of the connected domain according to the regularity degree value and the complexity degree value of the connected domain, determining a feature distance between any two different pixel points in the overhead image according to the difference between the feature values of the two pixel points in the connected domain to which the two pixel points belong; and clustering the pixel points in the overhead image by using the feature distances between different pixel points in the overhead image to obtain a farmland segmentation result, so as to monitor the farmland to be monitored according to the plot segmentation results of the farmland to be monitored at different time points.

[0008] In this way, the monitoring of the farmland to be monitored can be accurately realized according to the plot segmentation results of the farmland to be monitored at different time points.

[0009] Optionally, the regularity degree value of the connected domain is determined by the following method: determining a first variance of the curvature of the pixel points on the outer contour line of the connected domain, determining a second variance of the curvature of all edge pixel points in the connected domain, determining a first product of the normalized result of the first variance and the normalized result of the second variance, and taking the difference between the first preset value and the first product as the regularity degree value.

[0010] In this way, the obtained regularity degree value can better represent the continuity of the shape of the edge line in the connected domain.

[0011] Optionally, the complexity degree value of the connected domain is determined by the following method: determining a first ratio of the number of pixel points in the connected domain to the number of pixel points in the overhead image; determining a third variance of the gray value of the pixel points in the connected domain, determining a fourth variance of the gray value of the pixel points in the overhead image, and determining a second ratio between the third variance and the fourth variance; and taking the product of the first ratio and the second ratio as the complexity degree value of the connected domain.

[0012] In this way, by using the ratio of the number of pixel points in the connected domain to the number of pixel points in the entire image and the ratio between the variance of the gray value of the pixel points in the connected domain and the variance of the gray value of the pixel points in the entire image, the complexity degree value of the connected domain can be used to distinguish different types of plots.

[0013] Optionally, the feature value of the connected domain is determined according to the regularity degree value and the complexity degree value of the connected domain, which comprises: determining a first exponential operation result of the regularity degree value of the connected domain by using an exponential function with a natural constant as the base, and taking the ratio of the complexity degree value of the connected domain to the first exponential operation result as the feature value of the connected domain.

[0014] In this way, the eigenvalue of the connected domain can combine the regularity value and the complexity value of the connected domain, so as to reduce the probability of misidentification of different types of plots.

[0015] Optionally, the feature distance between the two pixel points is determined according to the difference between the eigenvalues of the connected domains to which the two pixel points belong, including: determining a color difference value between the two pixel points, and normalizing the color difference value to obtain a normalized color difference value; using an exponential function with a natural constant as the base to process the absolute value of the difference between the eigenvalues of the connected domains to which the two pixel points belong to obtain a second exponential operation result, and determining the sum of the reciprocal of the second exponential operation result and a second preset numerical value to determine a second product of the sum and the normalized color difference value; obtaining a normalized distance value of the two pixel points in the overhead image, and taking the sum of the distance value and the second product as the feature distance between the two pixel points.

[0016] In this way, the difference between the eigenvalues of the connected domains to which the two pixel points belong can adaptively adjust the contribution of the color difference value to the feature distance.

[0017] Optionally, the color difference value between the two pixel points is determined by: determining the pixel value difference of the two pixel points in three different color channels, i.e., the red channel, the green channel, and the blue channel, and determining the square sum of the pixel value difference in different color channels, and taking the square root of the square sum as the color difference value.

[0018] Optionally, the cultivated land to be monitored is monitored according to the plot segmentation results of the cultivated land to be monitored at different time points, including: determining target monitoring information of the cultivated land to be monitored at different time points according to the plot segmentation results of the cultivated land to be monitored at different time points, so as to monitor the cultivated land to be monitored by using the target monitoring information at different time points.

[0019] Optionally, the target monitoring information includes at least one of crop type, crop growth, landform feature, land health status, and green plant health status.

[0020] Optionally, the method further includes: inputting the image regions corresponding to different plots in the plot segmentation result into a pre-trained plot classification model respectively to obtain plot type labels of the corresponding image regions output by the plot classification model.

[0021] According to a second aspect of the embodiments of the present application, a cultivated land quality monitoring system is provided, including: a processor and a memory, the memory stores computer program instructions, and the computer program instructions are executed by the processor to realize the steps of the cultivated land quality monitoring method provided in the first aspect of the present application.

[0022] The technical scheme provided by the embodiment of the present application can include the following beneficial effects: according to the regularity value and the complexity value of the connected domain in the overhead image of the to-be-monitored cultivated land, the characteristic value of the connected domain can be determined, according to the difference between different pixel points in the overhead image of the to-be-monitored cultivated land in the characteristic value of the connected domain to which the different pixel points belong, the feature distance between the different pixel points can be obtained, and thus a more accurate plot segmentation result of the pixel points in the overhead image of the to-be-monitored cultivated land can be obtained, and therefore, according to the plot segmentation result of the to-be-monitored cultivated land at different time points, the monitoring of the to-be-monitored cultivated land can be more accurately realized.

[0023] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and are not limiting to the present application. BRIEF DESCRIPTION OF DRAWINGS

[0024] The accompanying drawings, which are incorporated into the specification and constitute part of it, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the application.

[0025] Figure 1 is a flowchart of a cultivated land quality monitoring method according to an exemplary embodiment;

[0026] Figure 2 is a structural schematic diagram of a cultivated land quality monitoring system according to an exemplary embodiment. DETAILED DESCRIPTION

[0027] First, the application scenario of the embodiment of the present application is simply introduced. In the application scenario of the present application, the pixel points in the overhead image of the to-be-monitored cultivated land can be clustered, or the overhead image can be regionally divided to obtain the plot segmentation result of the overhead image of the to-be-monitored cultivated land. However, in the related art, the image regions actually belonging to the same cultivated land region can be divided into different image regions, or the image regions actually not belonging to the same cultivated land region can be divided into the same image region.

[0028] For example, when monitoring the cultivated land area of the to-be-monitored cultivated land, due to the inaccuracy of the plot segmentation result, the monitoring result of the cultivated land area of the cultivated land can be inaccurate, causing the underreporting or misreporting of the cultivated land area. For another example, when monitoring the crop growth of the to-be-monitored cultivated land, due to the inaccuracy of the plot segmentation result, the monitoring result of the growth stage or growth area of the crop can be inaccurate, causing the inaccuracy of the monitoring result of the crop growth. The specific type of information monitored in the embodiment of the present application does not constitute a limitation.

[0029] It should be noted that all the actions of obtaining signals, information or data in the present application are carried out in accordance with the corresponding data protection regulations and policies of the country where the device is located, and with the authorization given by the owner of the corresponding device.

[0030] To solve the above technical problems, the embodiments of the present application provide a farmland quality monitoring method and system, Figure 1 is a flow chart of a farmland quality monitoring method according to an exemplary embodiment, as Figure 1 shown, the method comprises the following steps.

[0031] In step S101, an overhead image of the farmland to be monitored is obtained, and the overhead image is subjected to connected component division to obtain a plurality of connected components, so as to obtain a regularity value of the connected components.

[0032] The overhead image of the farmland to be monitored can be obtained by satellite remote sensing image or unmanned aerial vehicle photographing; or an image acquisition device can be arranged at a high place of a tower arranged near the farmland, and the overhead image of the farmland to be monitored is obtained by the image acquisition device; the embodiments of the present application do not limit the collection mode of the overhead image of the farmland to be monitored.

[0033] By dividing the overhead image into connected components, pixel points with consistent color characteristics and adjacent positions can be divided into the same connected component. Since the division of the connected components is mainly based on the color of the pixel points and the positions between them, the obtained connected components may include pixel points actually belonging to different farmland regions, or different pixel points belonging to the same farmland region may be divided into different connected components, which affects the monitoring result of the farmland.

[0034] In order to facilitate personnel walking in the farmland and irrigating or fertilizing crops in the farmland, a passageway is usually arranged in the farmland in a transverse or longitudinal direction, so that the edge of the internal passageway in the same farmland region corresponds to the edge line in the overhead image of the farmland.

[0035] Compared with other regions except farmland, the edge line inside the farmland region is more flat, so that the connected component with the edge line inside the connected component being more flat in the overhead image has a higher probability of belonging to the farmland region.

[0036] Or, the connected component with the edge line inside the connected component being more flat in the overhead image has a lower probability of belonging to the non-farmland region; in the overhead image of the farmland to be monitored, the non-farmland region can be a river, a building, a forest, a mountain, etc.

[0037] Meanwhile, in order to facilitate the management personnel of the farmland to stack materials or tools such as fertilizers at the junctions between different farmlands, or to facilitate the management personnel of the farmland to walk at the junctions between different farmlands, or to facilitate the vehicles to drive at the junctions between farmlands, a road with a certain width is usually arranged between the junctions of different farmlands.

[0038] Compared with the junctions between the uncultivated areas and the uncultivated areas, or the junctions between the uncultivated areas and the cultivated areas, the edges of the junctions between the cultivated areas and the cultivated areas are smoother, so that the edges of the junctions between different connected domains in the top view image are smoother, and there is a higher probability that the edges correspond to the junctions between the cultivated areas and the cultivated areas.

[0039] The regularity degree value is used to represent the continuity of the shape of the edge line in the connected domain; for example, the regularity degree value can represent the shape continuity of the internal edge line of the connected domain, and the continuity of the shape of the external contour line; the greater the regularity degree value of the connected domain, the greater the probability that the connected domain corresponds to the cultivated area in the farmland to be monitored, and vice versa, the smaller the regularity degree value of the connected domain, the greater the probability that the connected domain corresponds to the uncultivated area in the farmland to be monitored.

[0040] In a possible implementation, the regularity degree value of the connected domain is determined by: determining a first variance of the curvature of the pixel points on the external contour line of the connected domain, and determining a second variance of the curvature of all edge pixel points in the connected domain, determining a first product of the normalized result of the first variance and the normalized result of the second variance, and taking the difference between the first preset value and the first product as the regularity degree value.

[0041] The first preset value can be greater than or equal to 1, for example, when the first preset value is 1, the regularity degree value of the connected domain wherein norm is a normalization processing function, is the variance of the curvature of the pixel points on the external contour line of the connected domain, is the variance of the curvature of all edge pixel points in the connected domain.

[0042] The curvature of the pixel points on the external contour line of the connected domain can reflect the rate of change of the shape of the pixel points on the external contour line of the connected domain; the variance of the curvature of the pixel points on the external contour line of the connected domain can reflect the difference degree of the rate of change of the shape of the pixel points on the external contour line of the connected domain; the smaller the difference degree of the rate of change of the shape of the pixel points on the external contour line of the connected domain, the more likely the connected domain is a cultivated area in the top view image.

[0043] The curvature of pixels on the internal edge lines of a connected domain reflects the rate of change of the shape of pixels on the internal edge lines of the connected domain; the variance of the curvature of pixels on the internal edge lines of the connected domain reflects the degree of difference in the rate of change of the shape of pixels on the internal edge lines of the connected domain; the smaller the degree of difference in the rate of change of the shape of pixels on the internal edge lines of the connected domain, the more likely the connected domain is to be farmland in the top-view image.

[0044] The edge lines in a top-view image can be determined by the gradient values ​​of the pixels in the grayscale image of the top-view image. The gradient values ​​of the pixels in the grayscale image can be determined by the Canny edge detection algorithm, the Sobel operator, and the Prewitt operator.

[0045] Normalization functions can normalize variables to the range of 0 to 1, facilitating operations and comparisons between variables. as well as Normalization process, after normalization as well as The values ​​of are all between 0 and 1, making as well as The product of these is also between 0 and 1, thus making the regularity value between 0 and 1.

[0046] For example, normalization functions can be min-max normalization, logarithmic transformation, arctangent function, and sigmoid function, etc.

[0047] In this way, the regularity value obtained by using the variance of the curvature of the pixels on the outer contour line of the connected domain and the variance of the curvature of all edge pixels in the connected domain can better characterize the continuity of the shape of the edge line in the connected domain, so as to characterize the probability that the connected domain belongs to the cultivated land area.

[0048] In step S102, the complexity value of the connected components is determined.

[0049] Different image regions in the top-view image of the farmland to be monitored may correspond to farmland areas with different crop types or non-farmland areas with different landform types. The complexity of the image regions corresponding to these different types of land parcels varies. In order to distinguish these different types of land parcels and to achieve a more refined division of pixels in the connected domain, the complexity value of the connected domain can be determined.

[0050] The complexity value is used to represent the difference between the connected domain and the overhead image in complexity, and the complexity of the overhead image can reflect the complexity of the whole plot included in the farmland to be monitored. By the difference between the connected domain and the overhead image in complexity, the probability of the connected domain corresponding to the non-cultivated region in the farmland to be monitored can be reflected.

[0051] In a possible implementation, the complexity value of the connected domain is determined by: determining a first ratio of the number of pixel points in the connected domain to the number of pixel points in the overhead image; determining a third variance of the gray value of the pixel points in the connected domain, determining a fourth variance of the gray value of the pixel points in the overhead image, and determining a second ratio between the third variance and the fourth variance; and taking the product of the first ratio and the second ratio as the complexity value of the connected domain.

[0052] For example, the complexity value of the connected domain , wherein, is the regularity value of the connected domain, m is the number of pixel points in the connected domain, M is the number of pixel points in the overhead image, n is the variance of the gray value of the pixel points in the connected domain, and N is the variance of the gray value of the pixel points in the overhead image.

[0053] The complexity value considers the proportion of the number of pixel points in the connected domain to the number of pixel points in the whole image, and considers the proportion between the variance of the gray value of the pixel points in the connected domain and the variance of the gray value of the pixel points in the whole image, which helps to identify the non-cultivated region in the farmland to be monitored, which has regular shape but complex internal texture, thereby reducing the probability of misidentifying different regions in the farmland to be monitored.

[0054] The proportion of the number of pixel points in the connected domain to the number of pixel points in the whole image realizes the normalization processing of the number of pixel points; and the proportion between the variance of the gray value of the pixel points in the connected domain and the variance of the gray value of the pixel points in the whole image realizes the normalization processing of the variance of the gray value of the pixel points, which is convenient for comparison or calculation between parameters after normalization processing.

[0055] The complexity value of the non-cultivated region is different from the complexity value of the cultivated region; or, the complexity value of the cultivated region planted with different crops is different, different crops can refer to different types of crops, or refer to the same type of crops in different growth stages, for example, different plots are planted with corns in different growth stages; or, the complexity value of the cultivated region planted with crops is also different from the complexity value of the cultivated region without crops.

[0056] In this way, by considering the proportion of the number of pixel points in the connected domain to the number of pixel points in the whole image, and considering the proportion between the variance of the gray value of the pixel points in the connected domain and the variance of the gray value of the whole image, the complexity value of the connected domain can effectively realize the differentiation of different types of plots, so as to further realize the re-division of the overhead image.

[0057] In step S103, according to the regularity value and the complexity value of the connected domain, a feature value of the connected domain is determined, and according to the difference between the feature values of any two different pixel points in the connected domain, a feature distance between the two pixel points is determined.

[0058] Since the regions with different regularity values correspond to different probabilities of cultivated land regions, and the connected domains with different complexity values belong to different types of plots with a high probability, the regularity value and the complexity value of the connected domain can be combined to realize the re-division of the overhead image by the feature value of the connected domain to obtain different plots.

[0059] In a possible implementation, according to the regularity value and the complexity value of the connected domain, the feature value of the connected domain is determined, including: determining a first exponential operation result of an exponential function with a natural constant as the base number on the regularity value of the connected domain, and taking the ratio of the complexity value of the connected domain to the first exponential operation result as the feature value of the connected domain.

[0060] For example, the feature value of the connected domain is wherein, is the complexity value of the connected domain, is the regularity value of the connected domain.

[0061] Compared with the non-cultivated land region that may exist in the overhead image of the cultivated land to be monitored, the regularity value of the cultivated land region that may exist in the overhead image of the cultivated land to be monitored is larger; and the complexity values of the connected domains corresponding to different types of plots in the overhead image of the cultivated land to be monitored are different, so that the feature value of the connected domain can effectively realize the differentiation between different types of plots.

[0062] For two cultivated land regions with the same regularity value, since the complexity values of the cultivated land regions planted with different crops are different; or the complexity value of the cultivated land region planted with crops is different from the complexity value of the cultivated land region without crops, the feature values of the connected domains corresponding to different types of cultivated land regions are different.

[0063] In this way, the eigenvalue of the connected domain can combine the regularity value and the complexity value of the connected domain, the eigenvalue can distinguish different types of land plots, improve the ability to distinguish different types of land plots, and help to reduce the probability of misidentification of different types of land plots.

[0064] In a possible implementation, the feature distance between the two pixel points is determined according to the difference between the eigenvalues of the connected domains to which the two pixel points belong, including: determining a color difference value between the two pixel points, and normalizing the color difference value to obtain a normalized color difference value; using an exponential function with a natural constant as a base to process an absolute value of a difference between the eigenvalues of the connected domains to which the two pixel points belong to obtain a second exponential operation result, and determining a sum of an inverse of the second exponential operation result and a second preset numerical value to determine a second product of the sum and the normalized color difference value; obtaining a normalized distance value of the two pixel points in the top view image, and taking a sum of the distance value and the second product as the feature distance between the two pixel points.

[0065] The second preset numerical value can be greater than or equal to 1, for example, when the second preset numerical value is 1, the feature distance between two different pixel points a and b in the top view image wherein, is the eigenvalue of the connected domain in which the pixel point a is located, is the eigenvalue of the connected domain to which the pixel point b belongs, is the Euclidean distance between the pixel point a and the pixel point b, is the color difference value between the pixel point a and the pixel point b, is the diagonal length of the top view image, and exp is an exponential function with a natural constant as a base.

[0066] The feature distance between the two pixel points considers not only the position information and the color difference between the pixel points in the top view image, but also the difference between the eigenvalues of the connected domains to which the pixel points belong. Since the eigenvalue of the connected domain is determined according to the complexity value and the regularity value of the connected domain, the eigenvalue can better distinguish different types of land plots, and thus the feature distance between the pixel points can more comprehensively reflect the difference between the pixel points.

[0067] The pixel point a and the pixel point b are any two different pixel points in the top view image of the farmland to be monitored; in the calculation formula of the feature distance between the two pixel points, the Euclidean distance between the pixel point a and the pixel point b is divided by the diagonal length of the top view image, which can realize the normalization processing of the Euclidean distance between the pixel points.

[0068] In the calculation formula of the feature distance between two pixel points, since the pixel value of the pixel point is in the range of 0 to 255, 255 is the maximum color difference that can occur between two pixel points in the top view image, and dividing the color difference value by 255 can realize the normalization processing of the color difference value between two pixel points.

[0069] When the difference in the feature value of the connected domain to which the two pixel points in the top view image respectively belong is large, for example, tends to 1, the two pixel points have a large probability of belonging to different land types, and the difference between the feature values can be used to better distinguish the two pixel points, the value of tends to 0, and the value of tends to 1, which can reduce the contribution of the color difference value between the pixel point a and the pixel point b to the feature distance, and avoid that the pixel points of different land types are divided into the same region.

[0070] When the difference in the feature value of the connected domain to which the two pixel points in the top view image respectively belong is small, for example, tends to 0, the land types of the two pixel points determined according to the feature value are relatively consistent, and it can be difficult to distinguish the different land types only by the feature value.

[0071] In order to more accurately distinguish the land types of the pixel points, in the calculation formula of the feature distance, when tends to 0, the value of tends to 1, and the value of tends to 2, which improves the contribution of the color difference value between the pixel point a and the pixel point b to the feature distance, improves the distinguishing effect of the color features of the pixel points on the pixel points, and avoids that the pixel points of different land types are divided into the same region. In this way, by the difference in the feature value of the connected domain to which the two pixel points respectively belong, the contribution of the color difference value to the calculation of the feature distance can be adaptively adjusted, and therefore, the pixel points belonging to different land types can be better distinguished, and the feature distance obtained can better reflect the difference in the position information and the features of the pixel points.

[0072] In a possible implementation manner, the color difference value between the two pixel points is determined by the following manner: the pixel value difference of the two pixel points in three different color channels, i.e., the red channel, the green channel and the blue channel, is determined, the sum of squares of the pixel value differences in different color channels is determined, and the square root of the sum is taken as the color difference value.

[0073]

[0074] ​​For example, the color difference value between two different pixels a and b in the overhead image is , , and are the pixel values of the pixel a in the red, green and blue channels respectively, , and are the pixel values of the pixel b in the red, green and blue channels respectively.

[0075] The overhead image of the farmland to be monitored includes three layers corresponding to the red, green and blue channels respectively; since different types of plots may have different performances in the red, green and blue channels, the pixel values of the pixels in the overhead image in the three layers can be used to distinguish different positions in the farmland to be monitored, so as to distinguish different types of plots.

[0076] The color difference value between two pixels can represent the difference degree of the two pixels in color features; the greater the color difference value between the two pixels, the higher the difference degree of the two pixels in color features, and the two pixels have a higher probability of belonging to different cultivated regions or uncultivated regions; on the contrary, the smaller the color difference value between the two pixels, the lower the difference degree of the two pixels in color features, and the two pixels have a higher probability of belonging to the same cultivated region or uncultivated region.

[0077] In this way, the color difference value between two pixels can combine the differences of the two pixels in the red, green and blue channels, so as to better represent the difference degree of the two pixels in color features, so as to distinguish different plots in the farmland to be monitored.

[0078] In step S104, the pixels in the overhead image are clustered using the feature distance between different pixels in the overhead image to obtain a farmland segmentation result, so as to monitor the farmland to be monitored according to the plot segmentation results of the farmland to be monitored at different times.

[0079] Since the feature distance between two pixels is determined according to the difference between the feature values of the two pixels in the connected domain to which they belong, the feature distance between the two pixels can better reflect the difference between the feature values of the two pixels in the connected domain to which they belong, and the feature values can better distinguish different types of plots, so that the clustering of the overhead image using the feature distance between different pixels in the overhead image can divide the position points belonging to the same type of plot and adjacent to each other into the same clustering cluster, so as to divide different clustering clusters as different plots.

[0080] For example, the K-means clustering algorithm and the feature distance between different pixels can be used to cluster the pixels in the overhead image, and the pixels in the overhead image of the farmland to be monitored are divided into K categories, K is a positive integer, and the specific value of K can be specified according to the actual needs of the user, or can be determined according to the actual situation of the overhead image of the farmland to be monitored. The embodiments of the present application will not be repeated here.

[0081] For example, by clustering the overhead image, a plurality of clustering clusters corresponding to farmland, forest, building and river, etc. in the farmland to be monitored can be obtained. For at least one clustering cluster corresponding to farmland in the plurality of clustering clusters, a clustering cluster corresponding to different crop types can be further obtained.

[0082] In one possible implementation, monitoring the farmland to be monitored according to the plot segmentation results of the farmland to be monitored at different times includes: determining target monitoring information of the farmland to be monitored at different times according to the plot segmentation results of the farmland to be monitored at different times, so as to monitor the farmland to be monitored by using the target monitoring information at different times.

[0083] The target monitoring information includes at least one of crop type, crop growth, landform feature, land health status and green plant health status.

[0084] The image area corresponding to each plot in the plot segmentation result can be input into the pre-trained plot classification model to obtain the plot type label of the corresponding image area output by the plot classification model.

[0085] For example, the output plot type label can include farmland without planting crops, farmland planting corn, farmland planting rice, farmland contaminated, plot of forest, plot of building and river area, etc.

[0086] The plot classification model can be trained by: obtaining a plurality of sample plot images, the plurality of sample plot images respectively corresponding to a plurality of sample plots, and the plurality of sample plots being different in at least one of crop type, crop growth, landform feature, land health status and green plant health status; pre-labeling the sample images with corresponding plot type labels; inputting the plurality of sample plot images and the corresponding plot type labels into a pre-constructed network model to train the network model; and using the trained network model as the plot classification model.

[0087] According to the plot segmentation result of the to-be-monitored cultivated land at different time points, the monitoring information of the to-be-monitored cultivated land at different time points can be determined. For example, when the monitoring information is the land health condition, the change of the land health condition of the plot at different time points can be determined according to the land health condition of different plots in the to-be-monitored cultivated land at different time points, so as to determine whether to process the land with abnormality.

[0088] The plot segmentation result of the to-be-monitored cultivated land includes a plurality of plots with different plot types. When the plot type of a plot is determined according to the image area where the plot is located, the target monitoring information of the to-be-monitored cultivated land can be determined according to the area or position of the plots with the same plot type at different time points.

[0089] For example, if the plot segmentation result includes a plot planted with corn and a plot not planted with crops, when the target monitoring information is the growth of crops, the growth change of the planted corn can be obtained according to the height change of the corn in the plot planted with corn at different time points.

[0090] In this way, since the plot segmentation result of the to-be-monitored cultivated land is more accurate, the monitoring of the to-be-monitored cultivated land according to the plot segmentation result of the to-be-monitored cultivated land at different time points can better achieve the monitoring of the monitoring information of the to-be-monitored cultivated land.

[0091] Figure 2 FIG. 1 is a structural schematic diagram of a cultivated land quality monitoring system 1000 according to an example embodiment. Referring to FIG. 1, Figure 2 The cultivated land quality monitoring system 1000 includes a processor 1100 and a memory 1200, and the memory 1200 stores computer program instructions, and the computer program instructions are executed by the processor 1100 to implement all or part of the steps of the cultivated land quality monitoring method in the present application.

[0092] It should be understood that the features of various embodiments of the present application described herein can be combined with each other unless otherwise specifically indicated.

Claims

1. A method of monitoring the quality of a ploughed field, characterized in that, The method comprises the following steps: An overhead image of the farmland to be monitored is acquired, and the overhead image is subjected to connected domain division to obtain a plurality of connected domains, so as to obtain a regularity degree value of the connected domain, which comprises: determining a first variance of the curvature of the pixel points on the outer contour line of the connected domain, and determining a second variance of the curvature of all edge pixel points in the connected domain, determining a first product of the normalized result of the first variance and the normalized result of the second variance, and taking the difference between a first preset value and the first product as the regularity degree value; the regularity degree value is used to represent the continuity of the shape of the edge line in the connected domain; A complexity degree value of the connected domain is determined, which comprises: determining a first ratio of the number of pixel points in the connected domain to the number of pixel points in the overhead image; determining a third variance of the gray value of the pixel points in the connected domain, determining a fourth variance of the gray value of the pixel points in the overhead image, and determining a second ratio between the third variance and the fourth variance, and taking the product of the first ratio and the second ratio as the complexity degree value of the connected domain; the complexity degree value is used to represent the difference degree of the complexity of the connected domain and the overhead image; According to the regularity degree value and the complexity degree value of the connected domain, a feature value of the connected domain is determined, which comprises: determining a first exponential operation result of an exponential function with a natural constant as the base on the regularity degree value of the connected domain, and taking the ratio of the complexity degree value of the connected domain to the first exponential operation result as the feature value of the connected domain; for any two different pixel points in the overhead image, a feature distance between the two pixel points is determined according to the difference between the feature values of the two pixel points in the connected domain to which the two pixel points belong; The pixel points in the overhead image are clustered to obtain a farmland segmentation result by using the feature distances between different pixel points in the overhead image, so as to monitor the farmland to be monitored according to the plot segmentation results of the farmland to be monitored at different time.

2. The method of claim 1, wherein According to the difference between the feature values of the two pixel points in the connected domain to which the two pixel points belong, the feature distance between the two pixel points is determined, which comprises: A color difference value between the two pixel points is determined, and the color difference value is normalized to obtain a normalized color difference value; An exponential function with a natural constant as the base is used to process the absolute value of the difference between the feature values of the connected domains to which the two pixel points belong to obtain a second exponential operation result, and the sum of the reciprocal of the second exponential operation result and a second preset value is determined to determine a second product of the sum and the normalized color difference value; A distance value of the two pixel points in the overhead image after normalization is obtained, and the sum of the distance value and the second product is taken as the feature distance between the two pixel points.

3. The method of claim 2, wherein The color difference value between the two pixel points is determined by the following method: The pixel value difference of the two pixel points in three different color channels, i.e., the red channel, the green channel and the blue channel, is determined, and the square sum of the pixel value difference in different color channels is determined, and the square root of the square sum is taken as the color difference value.

4. The method of claim 1, wherein According to the plot segmentation results of the farmland to be monitored at different time, the farmland to be monitored is monitored, which comprises: According to the plot segmentation results of the farmland to be monitored at different time, target monitoring information of the farmland to be monitored at different time is determined, so as to monitor the farmland to be monitored by using the target monitoring information at different time.

5. The method of claim 4, wherein The target monitoring information comprises at least one of crop type, crop growth, landform feature, land health status and green plant health status.

6. The method of claim 1, wherein The method further comprises: inputting image regions corresponding to different plots in the plot segmentation result into a pre-trained plot classification model respectively to obtain plot type labels of the corresponding image regions output by the plot classification model.

7. A cultivated ground quality monitoring system, characterized in that The method comprises: A processor and a memory, the memory stores computer program instructions, and the computer program instructions are executed by the processor to implement the cultivated land quality monitoring method according to any one of claims 1-6.

Images