Crop planting area segmentation method and system based on remote sensing image information enhancement

Abstract

The present invention relates to the field of regional segmentation technology, and in particular to a crop planting area segmentation method and system based on remote sensing image information enhancement. The method analyzes the crop characteristics of each connected domain, preliminarily determines the mergeable connected domain; updates the mergeable connected domain by the differences in area and crop characteristics between different divided regions in the mergeable connected domain, combined with the differences between historical normalized vegetation indices; determines reference segments for the coincident edge points of each adjacent mergeable connected domain, analyzes the differences on both sides of the reference segment each time to obtain feature correlation and combines the crop characteristics to obtain a merged connected domain; determines the most preferred seed point for the merged connected domain to perform regional growth to obtain a crop area. The present invention divides and merges the crop characteristics between the preliminary connected domains in the remote sensing image, and analyzes the positions of the preferred seed points, so that the selection of seed points in the preliminary crop area is better, and the accuracy and efficiency of regional growth segmentation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of region segmentation, and in particular to a crop planting region segmentation method and system based on remote sensing image information enhancement. Background Art

[0002] With the increasing demand for crop planting, precision agriculture has become an important direction for the development of modern agriculture. Accurately obtaining information on crop planting areas is of vital importance for agricultural production management, optimal resource allocation, yield estimation, and agricultural policy formulation. Remote sensing image technology plays a vital role in the monitoring and management of crop planting areas. How to accurately and efficiently extract crop planting areas from remote sensing image data is an important technical issue facing the field of agricultural remote sensing.

[0003] Generally speaking, remote sensing images cover a rich variety of ground information and complex background environments, such as various vegetation types, different soil textures, towering buildings, vast water bodies, etc. Since the distribution of elements in the image and the regularity of brightness changes are not high, it is greatly difficult to outline the boundaries of crop planting areas. When regional growing is used to perform segmentation tasks on remote sensing images, the complexity of distinguishing crops from backgrounds or different crop planting areas is significantly increased due to the poor recognition of crop planting area boundaries, which ultimately leads to deviations in segmentation results. In addition, the selection of seed points during regional growing segmentation often relies on simple automation strategies, which will lead to inaccurate segmentation of crop planting areas in complex remote sensing image environments, and may cause the regional growing process to terminate prematurely or overgrow, resulting in missed detection or false detection of crop planting areas. Summary of the invention

[0004] In order to solve the technical problem that the selection of seed points in the prior art during region growing segmentation is simple, which leads to inaccurate segmentation of crop planting areas in complex remote sensing image environments, the purpose of the present invention is to provide a crop planting area segmentation method and system based on remote sensing image information enhancement, and the technical solutions adopted are as follows:

[0005] The present invention provides a crop planting area segmentation method based on remote sensing image information enhancement, the method comprising:

[0006] Obtain remote sensing grayscale images of the same geographical area at different historical sampling times; obtain the connected domain of the remote sensing grayscale image and the normalized vegetation index of each connected domain;

[0007] According to the pixel value distribution and texture distribution consistency between the pixels in each connected domain, the crop characteristic index of each connected domain is obtained; based on the approximate crop characteristic index between adjacent connected domains, the connected domains that can be merged are determined;

[0008] According to the distribution of edge points in the mergeable connected domain, different division areas of the mergeable connected domain are obtained; according to the deviation of crop characteristic indicators and areas in different division areas, as well as the difference degree of normalized vegetation index between different historical sampling moments, new mergeable connected domains are determined;

[0009] Through each pixel point on the coincident edge between each adjacent mergible connected domain, the longest line segment intersecting the edge of the adjacent mergible connected domain is used as each reference line segment; according to the closeness of the distribution of pixel values ​​of each adjacent mergible connected domain on both sides of each reference line segment, the feature correlation of each reference line segment is obtained; based on the consistent distribution of the feature correlation of each adjacent mergible connected domain on all reference line segments and the approximate distribution of crop feature indicators, the adjacent mergible connected domain is merged to obtain a merged connected domain;

[0010] In each non-mergeable connected domain and merged connected domain, the seed point is determined according to the pixel value and gradient deviation of each pixel point, combined with the position distribution of the pixel points; regional growth is performed based on the seed points in the current remote sensing grayscale image to obtain the crop area.

[0011] Furthermore, the method for obtaining the crop characteristic index includes:

[0012] For any connected domain, the mode of the pixel values ​​in the connected domain is normalized to obtain the maximum pixel distribution degree of the connected domain; the intersection size between the distribution range of the pixel values ​​in the connected domain and the preset crop pixel value range is used as the pixel distribution characteristic degree of the connected domain; the product of the maximum pixel distribution degree of the connected domain and the pixel distribution characteristic degree is used as the crop pixel characteristic index of the connected domain;

[0013] The variance of the pixel values ​​of all the pixels in the connected domain is used as the pixel chaos distribution degree of the connected domain; the variance of the gradient values ​​of all the pixels in the connected domain is calculated to obtain the texture chaos distribution degree of the connected domain; the product of the pixel chaos distribution degree and the texture chaos distribution degree of the connected domain is negatively correlated and mapped to obtain the crop uniformity characteristic index of the connected domain;

[0014] The crop pixel feature index and the crop uniform feature index of the connected domain are combined to obtain the crop feature index of the connected domain.

[0015] Furthermore, the method for obtaining the mergeable connected domains includes:

[0016] Half of the crop characteristic index of all connected domains is used as the initial difference threshold;

[0017] When the difference between the crop characteristic indicators between adjacent connected domains is less than the initial difference threshold, the corresponding adjacent connected domains are regarded as mergeable connected domains.

[0018] Furthermore, the division according to the distribution of edge points in the mergible connected domain to obtain different division areas of the mergible connected domain includes:

[0019] For any mergible connected domain, the longest edge in the mergible connected domain is obtained as the suspected edge of the mergible connected domain; the vertical direction of the mean value of the gradient direction of all pixels on the suspected edge is used as the extension direction;

[0020] On the suspected edge, the endpoints that do not intersect with the boundary of the mergeable connected domain are extended along the extension direction until they intersect with the boundary of the mergeable connected domain, thereby obtaining the partition edge of the mergeable connected domain; there are two intersection points between the partition edge and the boundary of the mergeable connected domain;

[0021] The two regions divided by the partition edge of the mergeable connected domain are regarded as different partition regions.

[0022] Furthermore, the new mergeable connected domains are determined according to the deviations of crop characteristic indicators and areas in different divided regions and the difference degree of normalized vegetation index between different historical sampling moments, including:

[0023] For any mergable connected domain, the crop characteristic index of each divided area in the mergable connected domain and the normalized vegetation index on the remote sensing grayscale image at different historical sampling times are obtained;

[0024] The product of the negative correlation mapping value of the area difference between different divided regions and the difference of crop characteristic index is used as the regional independent index of the mergeable connected domain;

[0025] The normalized vegetation index of each divided area in the mergeable connected domain is arranged in time series to obtain an NDVI sequence; after calculating the difference of the normalized vegetation index of the same sequence number in the NDVI sequence between two divided areas, the variance of all differences is calculated to obtain the vegetation difference index of the mergeable connected domain;

[0026] Normalizing the product of the regional independence index and the vegetation difference index of the mergible connected domain to obtain the segmentation necessity degree of the mergible connected domain;

[0027] When the segmentation necessity of the mergible connected domain is greater than a preset segmentation threshold, the corresponding mergible connected domain is segmented into different segmentation areas as new mergible connected domains.

[0028] Furthermore, the method for obtaining the reference line segment includes:

[0029] For any two adjacent mergible connected domains, the pixel points on the overlapping edges between the adjacent mergible connected domains are used as reference points in turn;

[0030] All straight lines passing through the reference point are compared with the line segments of the boundary of the adjacent mergeable connected domain as the optional line segments of the reference point; and the longest optional line segment of the reference point is used as a reference line segment.

[0031] Furthermore, the method for obtaining the feature association includes:

[0032] For any two adjacent mergable connected domains, the region after the adjacent mergable connected domains are merged is used as a reference region; each reference line segment in the reference region is used as a target line segment in turn, and the reference region is divided into two associated regions by the target line segments;

[0033] The difference in the mean pixel values ​​between the two associated regions is calculated as the mean deviation of the target line segment; the difference in the variance of the pixel values ​​between the two associated regions is calculated as the variance deviation of the target line segment; the product of the mean deviation and the variance deviation of the target line segment is negatively correlated and normalized to obtain the feature correlation of the target line segment.

[0034] Furthermore, the method for obtaining the merged connected domain includes:

[0035] For any two adjacent mergible connected domains, the mean of the crop characteristic indexes of the adjacent mergible connected domains is used as the characteristic reference index of the adjacent mergible connected domains;

[0036] The product of the normalized value of the length of each reference line segment in the adjacent mergable connected domain and the feature correlation is used as the weighted correlation degree of each reference line segment; the sum of the weighted correlation degrees of all reference line segments is used as the correlation feedback index of the adjacent mergable connected domain;

[0037] After calculating the feature correlation difference between every two reference line segments in the adjacent mergable connected domain, negative correlation mapping is performed on the mean values ​​of all feature correlation differences to obtain the correlation consistency index of the adjacent mergable connected domain;

[0038] The product of the feature reference index, the associated feedback index and the associated consistency index of the adjacent mergable connected domain is used as the merging credibility of the adjacent mergable connected domain;

[0039] The product of the merging credibility of the adjacent merging connected domain and the preset maximum merging threshold is used as the merging threshold of the adjacent merging connected domain;

[0040] When the difference in crop characteristic indicators between the adjacent mergable connected domains is less than the merging threshold, the adjacent mergable connected domains are merged to obtain a merged connected domain.

[0041] Furthermore, the method for obtaining the seed point includes:

[0042] In each non-mergeable connected domain and merged connected domain, for any connected domain, the mean pixel value of all pixels in the connected domain is taken as the pixel mean; the mean gradient value of all pixels in the connected domain is taken as the gradient mean;

[0043] Each pixel point of the connected domain is taken as an analysis point in turn; the difference between the pixel value of the analysis point and the pixel mean is normalized to obtain the regional pixel deviation of the analysis point; the difference between the gradient value of the analysis point and the gradient mean is normalized to obtain the regional gradient deviation of the analysis point; the sum of the regional pixel deviation and the regional gradient deviation of the analysis point is taken as the first representation of the analysis point;

[0044] After calculating the distance between the analysis point and every other pixel in the connected domain, the sum of all distances is calculated and normalized to obtain the second representation of the analysis point;

[0045] Perform negative correlation mapping on the product of the first characterization degree and the second characterization degree of the analysis point to obtain the seed preference degree of the analysis point;

[0046] The pixel point with the maximum seed preference in the connected domain is used as the seed point.

[0047] The present invention also provides a crop planting area segmentation system based on remote sensing image information enhancement, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the above methods when executing the computer program.

[0048] The present invention has the following beneficial effects:

[0049] The present invention analyzes the crop characteristics of each connected domain, preliminarily determines the connected domain that can be merged, considers the edge annexation caused by the blurred boundary of the crop area, and further analyzes the division of the connected domain in the mergeable connected domain. Through the different division areas generated by the edge points of the mergeable connected domain, through the differences in area and crop characteristics between the divided areas, and combined with the differences between the historical normalized vegetation index, the possibility of different planting areas is analyzed, and the mergeable connected domain is updated to reduce the possibility of analyzing the same area due to blurred boundaries even though there are large differences in vegetation and crops. Further, each adjacent mergeable connected domain is divided into reference segments by overlapping edge points, and the difference analysis is performed on the two sides of the division of each reference segment to obtain feature correlation, and the correlation between connected domains is analyzed as a whole. Combined with the crop characteristic index, the merged connected domain is adjusted and merged to improve the accuracy of merging similar crop areas. Finally, in order to obtain a more accurate division result, the selectivity of the seed point is further analyzed, the position, pixel value and gradient of the pixel point in the remaining connected domain are analyzed, and the most preferred seed point is determined for regional growth to obtain the crop area. The present invention divides and merges crop feature analysis between preliminary connected domains in remote sensing images, and analyzes the optimal seed point position, so as to make the selection of seed points in the preliminary crop area better and improve the accuracy and efficiency of regional growth segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A flow chart of a crop planting area segmentation method based on remote sensing image information enhancement provided by an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of a remote sensing grayscale image provided by an embodiment of the present invention;

[0053] Figure 3 A schematic diagram of a suspected edge in a connected domain provided by an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of a curve showing changes in NDVI values ​​of different divided areas at historical sampling times provided by an embodiment of the present invention;

[0055] Figure 5 A schematic diagram of an optional line segment of a reference point provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the crop planting area segmentation method and system based on remote sensing image information enhancement proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0057] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0058] The following is a detailed description of a crop planting area segmentation method and system based on remote sensing image information enhancement provided by the present invention in conjunction with the accompanying drawings.

[0059] See also Figure 1 , which shows a flow chart of a crop planting area segmentation method based on remote sensing image information enhancement provided by an embodiment of the present invention, the method comprising the following steps:

[0060] S1: Obtain remote sensing grayscale images of the same geographical area at different historical sampling times; obtain the connected domain of the current remote sensing grayscale image and the normalized vegetation index of each connected domain.

[0061] In an embodiment of the present invention, the crop areas to be divided are photographed by using an unmanned aerial vehicle equipped with a camera with high spatial resolution to collect remote sensing images. In order to make the division more accurate, remote sensing images of the same area in several historical periods are obtained at the same time, that is, remote sensing images at historical sampling moments. In addition to crop planting areas, remote sensing images usually also contain a variety of other types of land features. These land features and crop planting areas together constitute a complex background of the remote sensing image. The crop planting areas present a relatively consistent pixel representation, and also have relatively uniform texture features and regular geometric shapes.

[0062] In the embodiments of the present invention, the remote sensing image needs to be preprocessed to obtain a remote sensing grayscale image, so as to provide a data basis for subsequent analysis that is more convenient. The image preprocessing may specifically include image grayscale processing and image denoising processing, etc. Image preprocessing is a technology well known in the art and is not limited to this. It should be noted that during the image acquisition process, it is necessary to pay attention to the impact of environmental factors such as weather conditions and lighting conditions on the quality of remote sensing imaging, and take corresponding measures to correct and compensate, such as geometric correction or radiation correction. Please refer to Figure 2, which shows a schematic diagram of a remote sensing grayscale image provided by an embodiment of the present invention.

[0063] In remote sensing images, different landforms such as rivers and vegetation have different characteristics. Crop planting areas usually have obvious characteristics different from other surrounding landforms. Therefore, the basis for the connected domain detection algorithm is the connectivity between pixels in the image. By comparing the pixel values ​​of adjacent pixels and defining the neighborhood relationship, it is preliminarily determined which pixels belong to the same connected domain. Therefore, using the connected domain algorithm to obtain the connected domain of the remote sensing grayscale image, several connected domains can be obtained.

[0064] The normalized vegetation index of each connected domain is further obtained. The normalized vegetation index (NDVI) is one of the important parameters widely used in the field of remote sensing to reflect the growth and nutritional information of crops. By combining the NDVI value, it can help better understand the different distribution conditions of crops in the image.

[0065] S2: Obtain the crop characteristic index of each connected domain according to the pixel value distribution and texture distribution consistency between the pixels in each connected domain; determine the connected domains that can be merged based on the approximation of the crop characteristic indexes between adjacent connected domains.

[0066] In remote sensing images, the boundaries of crop planting areas are sometimes not clear, and the colors and textures vary due to different crop types, growth stages, and environmental conditions, making it difficult to accurately segment crop planting areas. Therefore, connected domain division is difficult to directly distinguish existing crop planting areas.

[0067] For a crop planting area or other surface feature area, due to the complex factors affecting crop growth conditions, even in the same crop planting area, it may be initially divided into multiple connected domains due to the inevitable differences in crop growth conditions. At this time, if the seed point is directly selected, the same feature area will be segmented as a whole, which is not conducive to the subsequent analysis of remote sensing images. Therefore, it is necessary to preliminarily consider the feature similarity between connected domains and merge connected domains to ensure that the same feature area is driven by only one seed point in the subsequent regional growth process, thereby improving the accuracy and consistency of segmentation.

[0068] Based on the distribution of pixel values ​​and textures of each connected domain, the regional crop characteristics are preliminarily analyzed. Preferably, in an embodiment of the present invention, the method for obtaining the crop characteristic index includes:

[0069] First, for any connected domain, the number of pixel values ​​in the connected domain is normalized to obtain the maximum pixel distribution degree of the connected domain, reflecting the degree of consistent distribution of the main distributed pixel values ​​of the connected domain. The size of the intersection between the distribution range of the pixel values ​​in the connected domain and the preset crop pixel value range is used as the pixel distribution characteristic degree of the connected domain, reflecting the consistency between the pixel value fluctuation range in the connected domain and the crop area. It is understandable that the preset crop pixel value range can be set by the implementer according to the crop conditions of the specific real-time scene, and no specific restrictions are made here.

[0070] It should be noted that normalization is a technical means well known to those skilled in the art, and the normalization method may be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0071] Therefore, the product of the maximum pixel distribution degree and the pixel distribution characteristic degree of the connected domain is used as the crop pixel feature index of the connected domain. The larger the maximum pixel distribution degree and the pixel distribution characteristic degree, that is, the larger the crop pixel feature index, the higher the degree of consistent distribution of pixel values ​​in the connected domain and the closer it is to the crop pixel value distribution interval, and the more significant the crop characteristics.

[0072] The variance of the pixel values ​​of all pixels in the connected domain is further used as the pixel chaos distribution degree of the connected domain, reflecting the degree of deviation when the pixel value deviates. The variance of the gradient values ​​of all pixels in the connected domain is calculated to obtain the texture chaos distribution degree of the connected domain, reflecting the degree of texture complexity distribution.

[0073] Therefore, the product of the pixel chaos distribution degree and the texture chaos distribution degree of the connected domain is negatively correlated to obtain the crop uniformity feature index of the connected domain. The smaller the pixel chaos distribution degree and the texture chaos distribution degree are, that is, the larger the crop uniformity feature index is, the higher the uniformity of the pixel value and texture in the connected domain is, and the more significant the crop characteristics are.

[0074] It should be noted that negative correlation mapping is a technical means well known to those skilled in the art, such as using an inverse proportion or a negative exponential power form, etc., which will not be limited or elaborated herein.

[0075] Finally, the crop pixel feature index and the crop uniform feature index of the connected domain are combined to obtain the crop feature index of the connected domain. In the embodiment of the present invention, the product of the crop pixel feature index and the crop uniform feature index of the connected domain is used as the crop feature index of the connected domain. The larger the crop pixel feature index and the crop uniform feature index, the more significant the crop feature performance in the connected domain, so the larger the crop feature index.

[0076] When the crop characteristic index of the connected domain is used as the basis for merging the connected domains, the smaller the difference in the crop characteristic index, the more likely it is that they are separated by the detail features in the same feature area, so they need to be merged to have the same seed point. However, the conventional single threshold for screening has poor adaptability. When there are interference features of roads and rivers in a preliminary connected domain, the crop characteristic analysis of the connected domain will have a certain deviation, so the difference in the degree of crop characteristics of the two adjacent areas may not reach the artificially set fixed threshold, but the two adjacent areas appear to belong to the same feature area as a whole.

[0077] Therefore, the merging of adjacent connected domains is preliminarily analyzed first, and the details in the connected domains and the associated details between the connected domains that may be merged are further analyzed to provide the accuracy of the merging result. Therefore, preferably, in an embodiment of the present invention, the method for obtaining the mergeable connected domains includes:

[0078] Half of the crop characteristic index of all connected domains is used as the initial difference threshold. When the difference between the crop characteristic indexes of adjacent connected domains is less than the initial difference threshold, it means that there is no obvious difference between the adjacent connected domains, and the subsequent merging analysis can be performed, and the corresponding adjacent connected domains are all regarded as merging connected domains. On the contrary, it means that the different characteristic areas are significant, and there is no need for subsequent possible merging analysis.

[0079] S3: Divide the connected domain into different areas according to the distribution of edge points in the mergeable connected domain; determine the new mergeable connected domain according to the deviation of crop characteristic indicators and area in different divided areas, and the difference degree of normalized vegetation index between different historical sampling times.

[0080] Due to the high utilization rate of crop planting areas, some roads and rivers between areas are covered in the remote sensing images. There will be suspected edge situations in the initially divided connected domains, that is, there are shallow edge features inside a crop planting area in the remote sensing image. When dividing the connected domain, it will be swallowed up and its edge features are not significant. Please refer to Figure 3 , which shows a schematic diagram of a suspected edge in a connected domain provided by an embodiment of the present invention, and the circled area is a suspected edge.

[0081] If the vegetation features between the edge corresponding areas are different, only merging will cause the same seed point to be placed for regional growth, resulting in incomplete segmentation. Therefore, according to the distribution of edge points in the mergeable connected domain, the different divided areas of the mergeable connected domain are obtained, and further analysis is performed on whether the different divided areas need to be segmented. In an embodiment of the present invention, it includes:

[0082] For any mergable connected domain, the longest edge in the mergable connected domain is obtained as the suspected edge of the mergable connected domain, and the vertical direction of the mean value of the gradient direction of all pixels on the suspected edge is used as the extension direction. Due to the insignificance of the suspected edge, the connected domain cannot usually be completely divided, so it needs to be extended. It should be noted that the acquisition of pixel gradients is a technical means well known to those skilled in the art and will not be elaborated here.

[0083] Further on the suspected edge, the endpoints that do not intersect with the boundary of the mergeable connected domain are extended along the extension direction until they intersect with the boundary of the mergeable connected domain, so as to obtain the partitioning edge of the mergeable connected domain. There are two intersection points between the partitioning edge and the boundary of the mergeable connected domain, that is, after the suspected edge is extended, the connected domain can be divided into independent areas, and the two areas divided by the partitioning edge of the mergeable connected domain are regarded as different partitioning areas.

[0084] The differences between different divided areas are analyzed to determine the degree of segmentation required. In actual crop planting scenarios, the sizes of the planting area blocks are similar, so the closer the areas are, the more likely they are to correspond to two planting areas. The degree of segmentation required is analyzed in combination with the differences in crop characteristic indicators retrieved from the regions.

[0085] Moreover, at a certain moment, even if two regions appear to be connected in space, their vegetation growth conditions may be affected by a variety of factors, resulting in similar performance at that moment, making it difficult to accurately determine whether they are essentially the same vegetation type or have the same growth characteristics. Therefore, it is too one-sided to determine the division of connected domains based on remote sensing images of a certain period. At this time, it is considered that the NDVI value has different changes in different cycles of crop growth. For example, at the beginning, the NDVI values ​​between regions always change synchronously, while in the middle of crop growth, the NDVI values ​​between regions may have a gap. Only then can the independence between the two regions be revealed. Please refer to Figure 4 , which shows a schematic diagram of the NDVI value change curve of different divided areas at historical sampling times provided by an embodiment of the present invention, the horizontal axis represents the time series, and the vertical axis represents the NDVI value. If there are large differences in different divided areas in the middle and late stages, the degree of segmentation required is higher.

[0086] Therefore, the necessary degree of segmentation is analyzed by combining the crop index and area of ​​different divided areas and the normalized vegetation index of the area at different historical sampling moments, and the connected domain is re-divided. Preferably, in an embodiment of the present invention, according to the deviation of the crop characteristic index and area of ​​different divided areas and the difference degree of the normalized vegetation index between different historical sampling moments, a new mergeable connected domain is determined, including:

[0087] First, for any mergeable connected domain, the crop characteristic index of each divided area in the mergeable connected domain and the normalized vegetation index on the remote sensing grayscale image at different historical sampling times are obtained, and the crop characteristic index of each area is re-analyzed through step S2, and the NDVI value is obtained based on the remote sensing images of the same location with different historical sampling times.

[0088] Furthermore, the product of the negative correlation mapping value of the area difference between different divided regions and the difference in crop characteristic indicators is used as the regional independence indicator of the mergeable connected domain. When the area difference is smaller, the regional areas are more similar, and the difference in crop characteristic indicators is larger, the deviation in characteristics is more significant, which means that different regions are more likely to correspond to different crop areas, the regional independence is higher, and the division demand is greater.

[0089] The normalized vegetation index of each divided area in the mergeable connected domain is further arranged in time series to obtain the NDVI sequence, which reflects the change of NDVI value in time series. After calculating the difference of the normalized vegetation index of the same sequence number in the NDVI sequence between the two divided areas, the variance of all differences is calculated to obtain the vegetation difference index of the mergeable connected domain, that is, after calculating the difference of the NDVI values ​​corresponding to the same sampling time, the sum of the differences is calculated to reflect the degree of difference in the amplitude of vegetation growth. The larger the vegetation difference index, the more obvious the difference in crop characteristics between regions, and the greater the need for division.

[0090] Finally, the product of the regional independence index and the vegetation difference index of the mergeable connected domain is normalized to obtain the degree of segmentation necessity of the mergeable connected domain. The area and crop characteristics between regions, as well as the changes in NDVI values ​​over time, are combined to obtain the degree of segmentation necessity. The greater the degree of segmentation necessity, the more the connected domain needs to be divided.

[0091] When the segmentation necessity of the mergeable connected domain is greater than the preset segmentation threshold, the corresponding mergeable connected domain is segmented into different segmentation areas as new mergeable connected domains. That is, the mergeable connected domain is updated, and the original mergeable connected domain is updated to the mergeable connected domains corresponding to the two segmentation areas for subsequent merge analysis, thereby improving the reliability of subsequent merge association analysis. In the embodiment of the present invention, the preset segmentation threshold is set to 0.7, and the specific numerical value can be adjusted by the implementer according to the specific real-time scenario, which will not be described in detail here.

[0092] S4: Through each pixel point on the overlapping edge between each adjacent mergeable connected domain, the longest line segment intersecting the edge of the adjacent mergeable connected domain is used as each reference line segment; according to the closeness of the distribution of pixel values ​​of each adjacent mergeable connected domain on both sides of each reference line segment, the feature correlation of each reference line segment is obtained; based on the consistent distribution of the feature correlation of each adjacent mergeable connected domain on all reference line segments and the approximate distribution of crop characteristic indicators, they are merged to obtain a merged connected domain.

[0093] Since crop characteristic indicators are relatively general when they only reflect the overall correlation of connected domains based on the overall pixel points inside the connected domain, they may weaken the manifestation of some characteristics. Therefore, when analyzing the correlation of adjacent connected domains, several line segments can be selected for segmentation analysis to reflect the correlation through multiple overall analyses. For adjacent connected domains, segmented line segments can be constructed by overlapping edges. Therefore, for each pixel point on the overlapping edge between adjacent mergeable connected domains, the longest line segment intersecting the edge of the adjacent mergeable connected domain is used as each reference line segment, including:

[0094] For any two adjacent mergable connected domains, the pixels on the overlapping edges between the adjacent mergable connected domains are used as reference points in turn, and the pixels on each edge are analyzed. All straight lines passing through the reference points are compared to the line segments on the boundaries of the adjacent mergable connected domains as optional line segments of the reference points. See Figure 5 , which shows a schematic diagram of an optional line segment of a reference point provided by an embodiment of the present invention, where P represents the reference point.

[0095] The longest optional line segment of the reference point is further used as a reference line segment, and all edge points correspond to a reference line segment. Subsequently, the deviation on both sides can be analyzed based on the reference line segment, and the overall consistency can be analyzed by the deviation after multiple segmentations. Therefore, the correlation degree of each reference line segment after differentiation can be obtained by the distribution on both sides of the reference line segment. Preferably, in an embodiment of the present invention, the method for obtaining feature correlation includes:

[0096] First, for any two adjacent mergable connected domains, the area after the adjacent mergable connected domains are merged is used as the reference area, and the area is merged for overall analysis. Each reference line segment in the reference area is used as the target line segment in turn, and the reference area is divided into two associated areas by the target line segment. The associated area is also the part on both sides of the target line segment.

[0097] The difference in the mean pixel values ​​between the two associated regions is calculated as the mean deviation of the target segment. The difference in the variance of the pixel values ​​between the two associated regions is calculated as the variance deviation of the target segment. The two distribution characteristics of pixel value size and disorder are used to reflect the degree of similar association. The product of the mean deviation and variance deviation of the target segment is negatively correlated and normalized to obtain the feature correlation of the target segment. The smaller the deviation, the higher the consistency between the regions and the greater the correlation. As an example, the expression of feature correlation is:

[0098] ; In the formula, Indicates can be merged with its first The first of the adjacent mergeable connected domains The feature correlation of the reference line segments; Indicated in The mean deviation of the reference line segments; Respectively expressed in The variance deviation of the reference line segments. Expressed as an exponential function with a natural constant as base.

[0099] When two adjacent mergable connected domains have slightly different crop feature levels, but are highly consistent overall and clearly belong to the same feature region, in order to obtain accurate merging results, the merging threshold should be relatively increased to ensure that these regions can be correctly merged. Conversely, when adjacent mergable connected domains show significant differences in local features, and this difference is sufficient to indicate that they belong to different feature regions, in order to avoid incorrect merging, the merging threshold should be appropriately lowered. This can ensure accurate connected domain division, and ultimately obtain more accurate seed points to accurately segment crop regions and obtain clear boundaries.

[0100] Therefore, finally, through the feature correlation analysis of all reference line segments and the adjustment threshold of the distribution of crop characteristic indicators, the connected domains are merged to obtain the merged connected domain. Preferably, in an embodiment of the present invention, the method for obtaining the merged connected domain includes:

[0101] First, for any two adjacent mergeable connected domains, the mean of the crop characteristic indexes of the adjacent mergeable connected domains is used as the characteristic reference index of the adjacent mergeable connected domains. The larger the overall crop characteristic reference index is, the higher the characteristic significance is, and the higher the possibility of being the same crop is, so the more reliable the merger is.

[0102] Furthermore, the product of the normalized value of the length of each reference line segment in the adjacent mergeable connected domain and the feature correlation is used as the weighted correlation of each reference line segment, and the weighting is performed by the line segment length. The longer the reference segment is, the more reliable the correlation analysis is, so the feature correlation is re-weighted and distributed by the length. Then, the sum of the weighted correlations of all reference line segments is used as the correlation feedback index of the adjacent mergeable connected domain, reflecting the correlation degree of all reference line segment analyses. The larger the correlation feedback index, the higher the credibility of the merger.

[0103] Furthermore, after calculating the feature correlation difference between every two reference line segments in the adjacent mergeable connected domain, the means of all feature correlation differences are negatively correlated to obtain the correlation consistency index of the adjacent mergeable connected domain. The more consistent the correlation of the reference line segment analysis, the more accurate the correlation analysis. Therefore, the correlation consistency index is obtained by negatively correlating the means of the feature correlation differences between the two reference line segments. The smaller the overall difference, that is, the larger the correlation consistency index, the more credible the correlation degree analysis of the comprehensive reference line segments is.

[0104] Finally, the product of the feature reference index, association feedback index and association consistency index of the adjacent mergeable connected domain is used as the merging credibility of the adjacent mergeable connected domain. The larger the feature reference index, association feedback index and association consistency index are, the higher the consistency between the connected domains, the higher the merging credibility, and the larger the threshold retention is to increase the possibility of retention.

[0105] Then, the product of the merging credibility of the adjacent merging connected domain and the preset maximum merging threshold is used as the merging threshold of the adjacent merging connected domain. The smaller the merging credibility, the lower the retention degree of the merging threshold, so as to reduce the possibility of merging untrustworthy merging connected domains. In the embodiment of the present invention, the preset maximum merging threshold is set to 0.5, and the specific value can be adjusted by the implementer, and is not limited here.

[0106] Therefore, when the difference in crop characteristic indexes between adjacent mergable connected domains is less than the merging threshold, it means that the difference is small and can meet the characteristics of the feature consistent area, and the adjacent mergable connected domains are merged to obtain a merged connected domain. On the contrary, when the mergable connected domains cannot be merged, they can be used alone as a merged connected domain to directly participate in the subsequent seed point selection.

[0107] S5: In each non-mergeable connected domain and merged connected domain, the seed point is determined according to the pixel value and gradient deviation of each pixel point and the position distribution of the pixel points; regional growth is performed based on the seed points in the current remote sensing grayscale image to obtain the crop area.

[0108] In each non-mergeable connected domain and merged connected domain, that is, in all connected domains in the merged remote sensing grayscale image, in order to perform region growing more accurately and efficiently, each pixel point is analyzed by position, pixel value and gradient, and the most representative pixel point is used as the seed point to facilitate rapid region growing.

[0109] Preferably, in an embodiment of the present invention, the method for obtaining a seed point includes:

[0110] First, in each non-mergeable connected domain and merged connected domain, for any connected domain, the mean pixel value of all pixels in the connected domain is taken as the pixel mean, and the mean gradient value of all pixels in the connected domain is taken as the gradient mean. The closer the pixel value and gradient value of the pixel are to the mean, the higher the preference is.

[0111] Therefore, each pixel point of the connected domain is taken as an analysis point in turn, and each pixel point is analyzed. The difference between the pixel value of the analysis point and the pixel mean is normalized to obtain the regional pixel deviation of the analysis point, and the difference between the gradient value of the analysis point and the gradient mean is normalized to obtain the regional gradient deviation of the analysis point. When the deviation is smaller, it means that the preference is greater. Therefore, the deviation is first combined, and the sum of the regional pixel deviation and the regional gradient deviation of the analysis point is used as the first characterization of the analysis point.

[0112] Through further position analysis, the distance between the analysis point and every other pixel point in the connected domain is calculated, and then the sum of all distances is calculated and normalized to obtain the second characterization degree of the analysis point. The smaller the overall distance between the pixel point and all other pixel points, the greater the preference. Therefore, the second characterization degree is obtained by combining the distance between the analysis point and all other pixel points.

[0113] Finally, the product of the first characterization degree and the second characterization degree of the analysis point is negatively correlated and mapped to obtain the seed preference of the analysis point. The smaller the first characterization degree and the second characterization degree are, the smaller the deviation degree and the overall distance are, and the higher the preference is. Therefore, the greater the seed priority is, the higher the possibility of being used as a seed point is.

[0114] Finally, the pixel point with the maximum seed preference in the connected domain is used as the seed point. At this time, the seed point is the optimal seed point position. Based on the seed point in the current remote sensing grayscale image, regional growth is performed to obtain a more accurate crop area. In an embodiment of the present invention, the seed point is used as an initial point to more accurately guide the segmentation process of the remote sensing image. Through the regional growing algorithm, starting from each seed point, the region is gradually expanded according to the similarity and continuity between pixels until the stop condition is met. Then, the crop planting area is determined more accurately. After obtaining the accurate segmentation result of the crop planting area, the management personnel can analyze the specific growth status of the crop based on the performance characteristics of the crop planting area in the remote sensing image. These characteristics may include the color, texture, shape, and changes in the growth cycle of the crop. By comparing the performance of crops in different areas, management personnel can promptly discover problems such as abnormal growth or pests and diseases, and take corresponding management measures.

[0115] In summary, the present invention analyzes the crop characteristics of each connected domain, preliminarily determines the connected domain that can be merged, considers the edge annexation caused by the blurred boundary of the crop area, and further analyzes the division of the connected domain in the mergeable connected domain. Through the different division areas generated by the edge points of the mergeable connected domain, through the differences in area and crop characteristics between the divided areas, and combined with the differences between the historical normalized vegetation index, the possibility of different planting areas is analyzed, and the mergeable connected domain is updated to reduce the possibility of analyzing the same area due to blurred boundaries even though there are large differences in vegetation and crops. Further, each adjacent mergeable connected domain is divided by overlapping edge points to make reference line segments, and the difference analysis is performed on both sides of the division of each reference line segment to obtain feature correlation, analyze the correlation between connected domains as a whole, and combine crop feature indicators to adjust and merge to obtain merged connected domains, thereby improving the accuracy of merging similar crop areas. Finally, in order to obtain more accurate division results, the selectivity of the seed points is further analyzed, the positions, pixel values ​​and gradients of the pixel points in the remaining connected domains are analyzed, and the most preferred seed points are determined for regional growth to obtain crop areas. The present invention divides and merges crop feature analysis between preliminary connected domains in remote sensing images, and analyzes the optimal seed point position, so as to make the selection of seed points in the preliminary crop area better and improve the accuracy and efficiency of regional growth segmentation.

[0116] The present invention also provides a crop planting area segmentation system based on remote sensing image information enhancement, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the above methods when executing the computer program.

[0117] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0118] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A crop planting area segmentation method based on remote sensing image information enhancement, characterized in that: The method comprises: Obtain remote sensing grayscale images of the same geographical area at different historical sampling times; obtain the connected domain of the remote sensing grayscale image and the normalized vegetation index of each connected domain; According to the pixel value distribution and texture distribution consistency between the pixels in each connected domain, the crop characteristic index of each connected domain is obtained; based on the approximate crop characteristic index between adjacent connected domains, the connected domains that can be merged are determined; According to the distribution of edge points in the mergeable connected domain, different division areas of the mergeable connected domain are obtained; according to the deviation of crop characteristic indicators and areas in different division areas, as well as the difference degree of normalized vegetation index between different historical sampling moments, new mergeable connected domains are determined; Through each pixel point on the coincident edge between each adjacent mergible connected domain, the longest line segment intersecting the edge of the adjacent mergible connected domain is used as each reference line segment; according to the closeness of the distribution of pixel values ​​of each adjacent mergible connected domain on both sides of each reference line segment, the feature correlation of each reference line segment is obtained; based on the consistent distribution of the feature correlation of each adjacent mergible connected domain on all reference line segments and the approximate distribution of crop feature indicators, the adjacent mergible connected domain is merged to obtain a merged connected domain; In each non-mergeable connected domain and merged connected domain, the seed point is determined according to the pixel value and gradient deviation of each pixel point, combined with the position distribution of the pixel points; regional growth is performed based on the seed points in the current remote sensing grayscale image to obtain the crop area; The method for obtaining the feature association includes: For any two adjacent mergable connected domains, the region after the adjacent mergable connected domains are merged is used as a reference region; each reference line segment in the reference region is used as a target line segment in turn, and the reference region is divided into two associated regions by the target line segments; The difference in the mean pixel values ​​between the two associated regions is calculated as the mean deviation of the target line segment; the difference in the variance of the pixel values ​​between the two associated regions is calculated as the variance deviation of the target line segment; the product of the mean deviation and the variance deviation of the target line segment is negatively correlated and normalized to obtain the feature correlation of the target line segment.

2. According to claim 1, a crop planting area segmentation method based on remote sensing image information enhancement is characterized in that: The method for obtaining the crop characteristic index comprises: For any connected domain, the mode of the pixel values ​​in the connected domain is normalized to obtain the maximum pixel distribution degree of the connected domain; the intersection size between the distribution range of the pixel values ​​in the connected domain and the preset crop pixel value range is used as the pixel distribution characteristic degree of the connected domain; the product of the maximum pixel distribution degree of the connected domain and the pixel distribution characteristic degree is used as the crop pixel characteristic index of the connected domain; The variance of the pixel values ​​of all the pixels in the connected domain is used as the pixel chaos distribution degree of the connected domain; the variance of the gradient values ​​of all the pixels in the connected domain is calculated to obtain the texture chaos distribution degree of the connected domain; the product of the pixel chaos distribution degree and the texture chaos distribution degree of the connected domain is negatively correlated and mapped to obtain the crop uniformity characteristic index of the connected domain; The crop pixel feature index and the crop uniform feature index of the connected domain are combined to obtain the crop feature index of the connected domain.

3. The crop planting area segmentation method based on remote sensing image information enhancement according to claim 1, characterized in that: The method for obtaining the mergeable connected domains includes: Half of the crop characteristic index of all connected domains is used as the initial difference threshold; When the difference between the crop characteristic indicators between adjacent connected domains is less than the initial difference threshold, the corresponding adjacent connected domains are regarded as mergeable connected domains.

4. According to claim 1, a crop planting area segmentation method based on remote sensing image information enhancement is characterized in that: The method of dividing the connected domain according to the distribution of edge points in the connected domain to obtain different divided areas of the connected domain includes: For any mergible connected domain, the longest edge in the mergible connected domain is obtained as the suspected edge of the mergible connected domain; the vertical direction of the mean value of the gradient direction of all pixels on the suspected edge is used as the extension direction; On the suspected edge, the endpoints that do not intersect with the boundary of the mergeable connected domain are extended along the extension direction until they intersect with the boundary of the mergeable connected domain, thereby obtaining the partition edge of the mergeable connected domain; there are two intersection points between the partition edge and the boundary of the mergeable connected domain; The two regions divided by the partition edge of the mergeable connected domain are regarded as different partition regions.

5. The crop planting area segmentation method based on remote sensing image information enhancement according to claim 1, characterized in that: The new mergeable connected domains are determined based on the deviations of crop characteristic indicators and areas in different divided areas and the differences in normalized vegetation indexes between different historical sampling moments, including: For any mergable connected domain, the crop characteristic index of each divided area in the mergable connected domain and the normalized vegetation index on the remote sensing grayscale image at different historical sampling times are obtained; The product of the negative correlation mapping value of the area difference between different divided regions and the difference of crop characteristic index is used as the regional independent index of the mergeable connected domain; The normalized vegetation index of each divided area in the mergeable connected domain is arranged in time series to obtain an NDVI sequence; after calculating the difference of the normalized vegetation index of the same sequence number in the NDVI sequence between two divided areas, the variance of all differences is calculated to obtain the vegetation difference index of the mergeable connected domain; Normalizing the product of the regional independence index and the vegetation difference index of the mergible connected domain to obtain the segmentation necessity degree of the mergible connected domain; When the segmentation necessity of the mergible connected domain is greater than a preset segmentation threshold, the corresponding mergible connected domain is segmented into different segmentation areas as new mergible connected domains.

6. The crop planting area segmentation method based on remote sensing image information enhancement according to claim 1, characterized in that: The method for obtaining the reference line segment includes: For any two adjacent mergible connected domains, the pixel points on the overlapping edges between the adjacent mergible connected domains are used as reference points in turn; All straight lines passing through the reference point are compared with the line segments of the boundary of the adjacent mergeable connected domain as the optional line segments of the reference point; and the longest optional line segment of the reference point is used as a reference line segment.

7. The crop planting area segmentation method based on remote sensing image information enhancement according to claim 1, characterized in that: The method for obtaining the merged connected domain includes: For any two adjacent mergible connected domains, the mean of the crop characteristic indexes of the adjacent mergible connected domains is used as the characteristic reference index of the adjacent mergible connected domains; The product of the normalized value of the length of each reference line segment in the adjacent mergable connected domain and the feature correlation is used as the weighted correlation degree of each reference line segment; the sum of the weighted correlation degrees of all reference line segments is used as the correlation feedback index of the adjacent mergable connected domain; After calculating the feature correlation difference between every two reference line segments in the adjacent mergable connected domain, negative correlation mapping is performed on the mean values ​​of all feature correlation differences to obtain the correlation consistency index of the adjacent mergable connected domain; The product of the feature reference index, the associated feedback index and the associated consistency index of the adjacent mergable connected domain is used as the merging credibility of the adjacent mergable connected domain; The product of the merging credibility of the adjacent merging connected domain and the preset maximum merging threshold is used as the merging threshold of the adjacent merging connected domain; When the difference in crop characteristic indicators between the adjacent mergable connected domains is less than the merging threshold, the adjacent mergable connected domains are merged to obtain a merged connected domain.

8. The crop planting area segmentation method based on remote sensing image information enhancement according to claim 1, characterized in that: The method for obtaining the seed point includes: In each non-mergeable connected domain and merged connected domain, for any connected domain, the mean pixel value of all pixels in the connected domain is taken as the pixel mean; the mean gradient value of all pixels in the connected domain is taken as the gradient mean; Each pixel point of the connected domain is taken as an analysis point in turn; the difference between the pixel value of the analysis point and the pixel mean is normalized to obtain the regional pixel deviation of the analysis point; the difference between the gradient value of the analysis point and the gradient mean is normalized to obtain the regional gradient deviation of the analysis point; the sum of the regional pixel deviation and the regional gradient deviation of the analysis point is taken as the first representation of the analysis point; After calculating the distance between the analysis point and every other pixel in the connected domain, the sum of all distances is calculated and normalized to obtain the second representation of the analysis point; Perform negative correlation mapping on the product of the first characterization degree and the second characterization degree of the analysis point to obtain the seed preference degree of the analysis point; The pixel point with the maximum seed preference in the connected domain is used as the seed point.

9. A crop planting area segmentation system based on remote sensing image information enhancement, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of a crop planting area segmentation method based on remote sensing image information enhancement as described in any one of claims 1 to 8 are implemented.

Images