Farmland road detection method, equipment and system based on remote sensing satellite images

Abstract

The present invention relates to the field of image analysis technology, and in particular to a method, device and system for detecting cultivated land roads based on remote sensing satellite images. Remote sensing satellite images of cultivated land in the same cultivated land area at different times are obtained, and then the straight line extraction results at multiple times are integrated using the Hough line detection technology to obtain suspected cultivated land road lines. Subsequently, the position distribution relationship between these straight lines is analyzed, the road topological relationship value of each suspected cultivated land road line is calculated, and the influence of seasonal changes of crops on the stability of the topological relationship is considered, and the real cultivated land road factor is obtained based on the fluctuation of the road topological relationship value. Finally, the voting value of the straight line detection is combined with the real cultivated land road factor to evaluate and screen out the real cultivated land roads, which significantly improves the accuracy and reliability of cultivated land road detection. This method effectively integrates multi-temporal data and spatial topological analysis, providing strong support for precision agriculture and land resource management.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a method, device and system for detecting cultivated land roads based on remote sensing satellite images. Background Art

[0002] As an important means of protecting and utilizing cultivated land, cultivated land road detection is of great significance for understanding the current status of cultivated land utilization, verifying land statistical data, supervising land utilization, and protecting cultivated land quality. Therefore, accurate identification of cultivated land roads is the basis for achieving precision agricultural management.

[0003] With the rapid development of remote sensing technology, especially the widespread application of remote sensing satellite images, a new technical means has been provided for the detection of cultivated land roads. Since farmland in large areas is distributed in blocks, and cultivated land roads often appear in the form of straight lines, the existing technology usually uses Hough line detection to process remote sensing satellite images to identify and extract cultivated land roads. However, since the arrangement of crops in farmland often presents linear features similar to the shape of roads, only using Hough line detection on the image will cause the linear features of the arrangement of crops in the image to be misjudged as cultivated land roads, resulting in low accuracy of the final cultivated land road detection results. Summary of the invention

[0004] In order to solve the technical problem that only using Hough line detection on an image will cause the linear features of the arrangement of crops in the image to be misjudged as farmland roads, thereby causing the final farmland road detection result to have low accuracy, the purpose of the present invention is to provide a farmland road detection method, device and system based on remote sensing satellite images, and the technical solutions adopted are as follows:

[0005] Obtain remote sensing satellite images of cultivated land in the same cultivated land area at different times;

[0006] Based on the Hough linear transformation algorithm, the cultivated land remote sensing satellite images at each time are detected, and the voting value of each parameter combination is obtained in the Hough space; the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is integrated to screen out suspected cultivated land road lines;

[0007] In the cultivated land remote sensing satellite images at each time, the position distribution relationship between the suspected cultivated land road lines is analyzed to determine the road topological relationship value of each suspected cultivated land road line; in the cultivated land remote sensing satellite images at all times, the real cultivated land road factor is determined based on the fluctuation between the road topological relationship values ​​of each suspected cultivated land road line;

[0008] All voting values ​​of the parameter combination corresponding to each suspected cultivated land road straight line and the real cultivated land road factor are integrated and analyzed to obtain the real cultivated land road evaluation value; based on the real cultivated land road evaluation value, the real cultivated land road is screened out from all suspected cultivated land road straight lines.

[0009] Furthermore, the method for obtaining the suspected farmland road straight line includes:

[0010] The normalized value of the mean of the voting values ​​of each parameter combination in the cultivated land remote sensing satellite images at all times is used as the score rate of the straight line corresponding to each parameter combination;

[0011] Among the straight lines corresponding to all parameter combinations, the straight line corresponding to the score rate greater than the preset threshold is taken as the suspected farmland road straight line.

[0012] Furthermore, the method for obtaining the road topology relationship value includes:

[0013] In the cultivated land remote sensing satellite image at each time, the intersection points between all the suspected cultivated land road lines are obtained;

[0014] For any intersection, calculate the differences between all angles formed by the intersection to obtain the road topology factor corresponding to each intersection;

[0015] For any suspected cultivated land road line in the cultivated land remote sensing satellite image at each time, the road topology factors corresponding to all intersection points on the suspected cultivated land road line are averaged to obtain the road topology relationship value of the suspected cultivated land road line.

[0016] Furthermore, the method for obtaining the road topology factor includes:

[0017] For any intersection point, among all the angles formed by the intersection point, combine any two angles to obtain all non-repeating angle combinations;

[0018] Calculate the absolute value of the difference between the two angles in each angle combination as the angle deviation value;

[0019] Based on the angle deviation values ​​of all angle combinations corresponding to the intersection, a road topology factor corresponding to the intersection is determined, and the road topology factor is negatively correlated with the angle deviation value.

[0020] Furthermore, the method for obtaining the real farmland road factor includes:

[0021] When a suspected cultivated land road line does not exist in the cultivated land remote sensing satellite image at a certain time, the road topological relationship value of the suspected cultivated land road line in the cultivated land remote sensing satellite image at that time is recorded as 0;

[0022] The mean of the road topological relationship values ​​of each suspected farmland road line in the farmland remote sensing satellite images at all times is taken as the mean eigenvalue;

[0023] The absolute value of the difference between the road topological relationship value of each suspected cultivated land road line in the cultivated land remote sensing satellite image at each time and the mean characteristic value is used as a deviation factor;

[0024] The mean values ​​of all deviation factors of each suspected farmland road straight line are negatively correlated and normalized to obtain a value which is used as the true farmland road factor.

[0025] Furthermore, the method for obtaining the real farmland road evaluation value includes:

[0026] The product of the real cultivated land road factor of each suspected cultivated land road line and the preset weight value is used as the first real cultivated land road assessment factor;

[0027] The product of the value after negative correlation mapping of the preset weight value and the score rate of each suspected farmland road straight line is used as the second real farmland road evaluation factor;

[0028] The normalized value of the sum of the first real farmland road assessment factor and the second real farmland road assessment factor of each suspected farmland road straight line is used as the real farmland road assessment value of each suspected farmland road straight line.

[0029] Furthermore, the method for obtaining the real farmland road includes:

[0030] Among all the suspected farmland road straight lines, the suspected farmland road straight lines whose real farmland road evaluation values ​​are greater than a preset road threshold are taken as real farmland road straight lines, so as to obtain the real farmland road corresponding to the real farmland road straight lines.

[0031] Furthermore, the method for obtaining the voting value includes:

[0032] When detecting the cultivated land remote sensing satellite image at each time based on the Hough linear transformation algorithm, in the Hough space, whenever the edge pixel point meets a certain parameter combination, the corresponding value of the parameter combination in the accumulator array is added by 1, thereby obtaining the corresponding value of each parameter combination in the accumulator array;

[0033] The corresponding values ​​of each parameter combination in the accumulator array are normalized and used as the voting value of each parameter combination in the cultivated land remote sensing satellite image at each time.

[0034] A farmland road detection system based on remote sensing satellite images, the system comprising:

[0035] An image data acquisition module is used to acquire remote sensing satellite images of cultivated land in the same cultivated land area at different times;

[0036] The suspected cultivated land road screening module is used to detect the cultivated land remote sensing satellite images at each time based on the Hough linear transformation algorithm, and obtain the voting value of each parameter combination in the Hough space; the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is integrated to screen out the suspected cultivated land road lines;

[0037] The cultivated land road topology analysis module is used to analyze the position distribution relationship between the suspected cultivated land road lines in the cultivated land remote sensing satellite images at each time, and determine the road topology relationship value of each suspected cultivated land road line; in the cultivated land remote sensing satellite images at all times, based on the fluctuation between the road topology relationship values ​​of each suspected cultivated land road line, determine the real cultivated land road factor;

[0038] The real cultivated land road screening module is used to integrate and analyze all voting values ​​of the parameter combination corresponding to each suspected cultivated land road line and the real cultivated land road factor to obtain the real cultivated land road evaluation value; based on the real cultivated land road evaluation value, the real cultivated land road line is screened out from all suspected cultivated land road lines.

[0039] A farmland road detection device based on remote sensing satellite images comprises a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and when the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor, any one of the steps of a farmland road detection method based on remote sensing satellite images is implemented.

[0040] The present invention has the following beneficial effects:

[0041] When detecting cultivated land roads in remote sensing satellite images, the present invention further analyzes the difference between the linear characteristics of cultivated land roads and the linear characteristics of farmland crop arrangement based on the Hough line detection processing of the image, thereby helping to obtain more accurate cultivated land road detection results. In view of the seasonal characteristics of farmland crops, cultivated land remote sensing satellite images of the same cultivated land area at different times are obtained as the basis for the subsequent analysis process. Then, based on the linear characteristics of cultivated land roads, Hough line detection is performed on the cultivated land remote sensing satellite images at each time, and the voting value of each parameter combination is obtained in the Hough space, and each parameter combination represents each straight line; and the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is combined to preliminarily screen out suspected cultivated land road straight lines, and the suspected cultivated land road straight lines at this time include the straight lines corresponding to the cultivated land roads and the straight lines corresponding to the farmland crops, which need to be further distinguished. Because the distribution characteristics of cultivated land roads usually follow a certain topological relationship and are interspersed in the cultivated land farmland in a square shape, they often present a grid-like topological feature. Therefore, the position distribution relationship between the suspected cultivated land road straight lines is analyzed, and the road topological relationship value of each suspected cultivated land road straight line is determined. Further, since the crops in the farmland have seasonal characteristics, and the cultivated land roads are relatively fixed, the topological relationship of the cultivated land roads will not change with the change of time and season. Therefore, at all times, the fluctuation of the road topological relationship value of the suspected cultivated land road straight line is analyzed to determine the real cultivated land road factor. At this time, the real cultivated land road factor can be used as a basis for evaluating whether the suspected cultivated land road straight line represents the real cultivated land road. Finally, all the voting values ​​of the parameter combination corresponding to each suspected cultivated land road straight line and the real cultivated land road factor are fused to obtain the real cultivated land road evaluation value, thereby screening out the real cultivated land road. In summary, the present invention combines multi-phase characteristics and the topological relationship of cultivated land roads, and effectively improves the accuracy of cultivated land road detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] 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.

[0043] Figure 1 A method flow chart of a method for detecting cultivated land roads based on remote sensing satellite images provided by one embodiment of the present invention;

[0044] Figure 2 A schematic diagram of an enhanced image corresponding to a remote sensing satellite image of cultivated land provided by an embodiment of the present invention;

[0045] Figure 3 A method flow chart of a method for obtaining a road topology relationship value provided by an embodiment of the present invention;

[0046] Figure 4 A system block diagram of a farmland road detection system based on remote sensing satellite images provided by one embodiment of the present invention;

[0047] Figure 5 A schematic structural diagram of a farmland road detection device based on remote sensing satellite images provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] 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 a method, device and system for detecting cultivated land roads based on remote sensing satellite images 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.

[0049] 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.

[0050] The following is a detailed description of a method, device and system for detecting cultivated land roads based on remote sensing satellite images provided by the present invention in conjunction with the accompanying drawings.

[0051] See also Figure 1 , which shows a method flow chart of a method for detecting cultivated land roads based on remote sensing satellite images provided by an embodiment of the present invention, the method comprising the following steps:

[0052] Step S1: Acquire remote sensing satellite images of cultivated land in the same cultivated land area at different times.

[0053] With the rapid development of remote sensing technology, especially the widespread application of remote sensing satellite images, a new technical means has been provided for the detection of cultivated land roads. Since farmland in large areas is distributed in blocks, roads often appear in the form of straight lines. Therefore, existing technologies usually use Hough line detection to process remote sensing satellite images to identify cultivated land roads. However, when identifying the straight lines corresponding to cultivated land roads based on Hough line transformation, the arrangement of crops in the farmland will also show linear features similar to the road shape, and will be mistakenly identified as the straight lines corresponding to cultivated land roads.

[0054] Because the crops in the cultivated land have the characteristic of changing with the seasons, in order to be able to further distinguish the straight lines of crop arrangement and the straight lines of cultivated land roads based on this characteristic in the subsequent process, the embodiments of the present invention obtain cultivated land remote sensing satellite images of the same cultivated land area at different times. Specifically, the total time span includes spring, summer, autumn and winter, and the acquisition time interval of cultivated land remote sensing satellite images can be set to once a week, and the cultivated land remote sensing satellite images can be obtained using an unmanned aerial vehicle remote sensing platform; when determining the same cultivated land area, the geographic coordinate system can be used for auxiliary confirmation to ensure that all cultivated land remote sensing satellite images belong to the same cultivated land area, that is, the pixel points in the cultivated land remote sensing satellite images at different times have the characteristics of spatial correspondence.

[0055] It should be noted that the acquisition equipment and collection time interval of the cultivated land remote sensing satellite images can be adjusted according to the implementation scenario, and are not limited here. The cultivated land remote sensing satellite images can be subjected to radiation correction, geometric correction, and atmospheric correction, etc. to eliminate some systematic errors and interferences in the image acquisition process. The specific correction process is a well-known technology and will not be elaborated here.

[0056] Step S2: Based on the Hough linear transformation algorithm, the cultivated land remote sensing satellite images at each time are detected, and the voting value of each parameter combination is obtained in the Hough space; the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is integrated to screen out suspected cultivated land road lines.

[0057] In the remote sensing satellite images of cultivated land, cultivated land roads usually appear as straight lines or approximate straight lines. Therefore, the Hough linear transformation algorithm is used to detect them, and straight lines that can represent cultivated land roads can be obtained. In the Hough space, each possible parameter combination will obtain a voting value, which reflects the number of pixels in the image that support the parameter combination, and can also represent the strength of the straight line feature. Then, the voting values ​​of the parameter combinations in the remote sensing satellite images of cultivated land at all times can be combined to screen out suspected cultivated land road straight lines from the image.

[0058] Hough line detection is a well-known technology. Here we briefly describe its process: (1) edge detection, using edge detection algorithms (Canny or Sobel) to identify edge pixels in an image; (2) parameter space mapping, mapping the rectangular coordinate space to the parameter space ( ), any straight line can be expressed using polar coordinates as ,in, It represents the vertical distance from the origin to the straight line in the rectangular coordinate system. represents the angle between the vertical line and the x-axis, x represents the horizontal coordinate of the edge pixel in the rectangular coordinate system, and y represents the vertical coordinate of the edge pixel in the rectangular coordinate system; (3) accumulator array, in Hough space, whenever the edge pixel meets the parameter combination When a parameter combination in the accumulator array is found, the parameter combination at the corresponding position in the accumulator array will receive a vote; (4) Peak detection, after iterating all edge pixels, analyze the accumulator array to find the parameter combination with the highest vote count (peak); (5) Line parameter extraction, using the numerical value of the parameter combination at the peak to reconstruct the straight line equation and obtain the straight line. It should be noted that in the embodiment of the present invention, The value of is between 0 and the length of the image diagonal. The discretization step size can be set to 1 degree.

[0059] In the embodiment of the present invention, before obtaining edge pixels, the cultivated land remote sensing satellite image at each time can be subjected to histogram equalization to obtain an enhanced image, which can enhance the contrast of the image and make the details in the image clearer. Figure 2 , which shows a schematic diagram of an enhanced image corresponding to a remote sensing satellite image of cultivated land in one embodiment of the present invention.

[0060] Since there are remote sensing satellite images of cultivated land at multiple times, the calculation of the voting value in the above process is modified. Preferably, in one embodiment of the present invention, the method for obtaining the voting value includes:

[0061] When detecting the cultivated land remote sensing satellite image at each time based on the Hough linear transformation algorithm, in the Hough space, whenever the edge pixel point meets a certain parameter combination, the corresponding value of the parameter combination in the accumulator array is increased by 1 (the initial value is 0). After traversing all the edge pixels, the corresponding value of each parameter combination in the accumulator array can be obtained. At this time, each corresponding value in the accumulator array represents the number of edge pixels passing through the straight line represented by the parameter combination.

[0062] Then, the corresponding values ​​of each parameter combination in the accumulator array are normalized and used as the voting value of each parameter combination in the cultivated land remote sensing satellite image at each time. That is, the proportion of the corresponding values ​​of each parameter combination in the accumulator array is analyzed. At this time, the larger the voting value, the more edge pixels support the parameter combination in the cultivated land remote sensing satellite image, and the straight line corresponding to the parameter combination is regarded as the most likely straight line to appear in the cultivated land remote sensing satellite image.

[0063] After obtaining the voting value corresponding to each parameter combination in a single phase, that is, the cultivated land remote sensing satellite image at each time, the voting values ​​of the parameter combination in the cultivated land remote sensing satellite images at all times can be integrated to screen out the straight lines that are supported by more edge pixels in the cultivated land remote sensing satellite as suspected cultivated land road lines.

[0064] Preferably, in one embodiment of the present invention, the method for obtaining the suspected farmland road straight line includes:

[0065] In order to avoid the influence of noise or environmental changes in a single phase, the voting results of the same parameter combination in all cultivated land remote sensing satellite images are combined: the average of the voting values ​​of each parameter combination in cultivated land remote sensing satellite images at all times is normalized as the score rate of the line corresponding to each parameter combination. The higher the score rate corresponding to the parameter combination, the more edge pixels in the image exist on the line corresponding to the parameter combination. It should be noted that the normalization here can be used function.

[0066] Therefore, among the straight lines corresponding to all parameter combinations, the straight line corresponding to the score rate greater than the preset threshold is regarded as the suspected farmland road straight line. It should be noted that the preset threshold is set to 0.6, and the specific value can be adjusted according to the implementation scenario and is not limited here.

[0067] At this point, the straight lines in the cultivated land remote sensing satellite images can be preliminarily screened out, which correspond to cultivated land roads or crops arranged regularly in cultivated land farmlands, so they are named suspected cultivated land road straight lines.

[0068] Step S3: In the remote sensing satellite images of cultivated land at each time, the position distribution relationship between the suspected cultivated land road lines is analyzed to determine the road topology relationship value of each suspected cultivated land road line; in the remote sensing satellite images of cultivated land at all times, the real cultivated land road factor is determined based on the fluctuation between the road topology relationship values ​​of each suspected cultivated land road line.

[0069] Since farmland is distributed in a square shape in a large regional cultivated land area, generally in a rectangular or regular rectangle, cultivated land roads have relatively regular geometric characteristics and topological structures, and their position distribution characteristics follow a grid-like topological relationship. This feature can be used as a topological basis for judging whether a suspected road straight line represents a real cultivated land road; and for the straight lines corresponding to crops that are closely and regularly arranged in cultivated land, because crops are seasonal, cultivated land will show obvious vegetation changes with seasonal changes, so in multi-phase, the straight lines corresponding to crops are relatively less stable in structure, while the structure of cultivated land roads is more stable, making the position relationship of the straight lines corresponding to cultivated land roads relatively consistent in the multi-phase cultivated land remote sensing satellite images. Based on the above two features, the possibility of each suspected cultivated land road straight line being a real cultivated land road straight line can be further measured.

[0070] First, in the remote sensing satellite images of cultivated land at each time, the topological structure of the suspected cultivated land road straight lines is determined as a basis for distinguishing the real cultivated land road straight lines from the straight lines corresponding to the crop arrangement.

[0071] The position distribution relationship between the suspected farmland road straight lines is analyzed to determine the road topological relationship value of each suspected farmland road straight line.

[0072] Preferably, in one embodiment of the present invention, the method for obtaining the road topology relationship value includes:

[0073] See also Figure 3 , which shows a method flow chart of a method for obtaining a road topology relationship value in one embodiment of the present invention, the method comprising the following steps:

[0074] Step S301: Obtain the intersection points between all suspected farmland road lines in the farmland remote sensing satellite image at each time.

[0075] After obtaining each suspected farmland road straight line, the expression of each suspected farmland road straight line is also obtained, so the intersection points between the suspected farmland road straight lines can be obtained by mathematical calculation.

[0076] It should be noted that the embodiment of the present invention is analyzed in the remote sensing satellite image of cultivated land, so each suspected cultivated land road straight line is not extended infinitely, but the definition domain and value range of each suspected cultivated land road straight line have a value range. Therefore, if the coordinates of the calculated intersection point are not within the value range of the definition domain or value range of the suspected cultivated land road straight line, it will not participate in the calculation of subsequent steps.

[0077] In other embodiments of the present invention, since each suspected farmland road straight line is marked in the farmland remote sensing satellite image, the intersection point can also be directly obtained through the marking.

[0078] Step S302: For any intersection, calculate the differences between all angles formed by the intersection to obtain a road topology factor corresponding to each intersection.

[0079] For the intersection formed by the intersection of two straight lines, four angles will be formed. For farmland roads with a grid arrangement, each of the four angles should be close to 90 degrees, that is, the sizes of the four angles are relatively consistent, so for any intersection in step S301, among all the angles formed by the intersection, any two angles are combined to obtain all non-repeating angle combinations. For example: the four angles are recorded as angle 1, angle 2, angle 3, and angle 4, respectively, then all non-repeating angle combinations are (angle 1, angle 2), (angle 1, angle 3), (angle 1, angle 4), (angle 2, angle 3), (angle 2, angle 4), (angle 3, angle 4).

[0080] It should be noted that the angle can be calculated using the vector method, and then the cosine formula of the vector angle can be used to obtain the cosine value of the angle between the two lines, and the cosine value can be converted into an angle value through the arccos function. The angle value obtained by this method may be in the range of [0, ], but because of the four angles formed by the two straight lines, each set of vertical angles is equal and the two sets of vertical angles are complementary, so one angle value can be used to get all four angle values.

[0081] Then, the absolute value of the difference between the two angles in each angle combination is calculated as the angle deviation value. The larger the angle deviation value, the greater the difference between the two angles. It is considered that the two suspected farmland road straight lines corresponding to the intersection point do not conform to the characteristics of farmland roads arranged in a grid.

[0082] Finally, based on the angle deviation values ​​of all angle combinations corresponding to the intersection, the road topology factor corresponding to the intersection is determined. The formula model of the road topology factor includes:

[0083]

[0084] in, Indicates the road topology factor corresponding to each intersection; Indicates the total number of angle combinations; Represents the value of angle 1 in the i-th angle combination; Represents the value of angle 2 in the i-th angle combination; Represents an exponential function with the natural constant e as the base.

[0085] In the formula model of the road topology factor, when the angle deviation between two angles in the angle combination of a certain intersection When the value is larger, it means that the grid shape of the two suspected farmland road lines corresponding to the intersection is less obvious, and the possibility that the two suspected farmland road lines are farmland roads is lower, so the road topology degree of the intersection is lower. Therefore, the angle deviation value is negatively correlated and normalized to achieve logical relationship correction, and the result is , the larger the value is, the higher the road topology of the intersection is. The road topology factor corresponding to the intersection is obtained by averaging. The larger the road topology factor is, the more obvious the grid shape of the two suspected farmland road straight lines forming the intersection is, and the more likely it is to represent the real farmland road.

[0086] Step S303: The road topology factors corresponding to the intersection points on each suspected farmland road straight line are integrated to obtain the road topology relationship value of each suspected farmland road straight line.

[0087] Based on the analysis in step S302, it can be known that the larger the road topology factor of the intersection point is, the more likely the suspected farmland road line on which it is located is to represent the real farmland road. Therefore, for any suspected farmland road line in the farmland remote sensing satellite image at each time, the road topology factors corresponding to all intersections on the suspected farmland road line are averaged to obtain the road topology relationship value of the suspected farmland road line. The larger the road topology relationship value is, the more likely the suspected farmland road line is to represent the real farmland road.

[0088] Based on the above steps, the road topological relationship value of each suspected cultivated land road line in the cultivated land remote sensing satellite image at each time can be obtained. The road topological relationship value preliminarily represents the possibility that each suspected cultivated land road line represents the real cultivated land road. Because cultivated farmland will show obvious vegetation changes with the change of seasons, and the structure of cultivated land roads is relatively stable in time, the topological structure of the straight line corresponding to the cultivated land road will be relatively consistent in the cultivated land remote sensing satellite image at all times, that is, the fluctuation of the topological relationship value will be small, so the change fluctuation between the road topological relationship values ​​of each suspected cultivated land road line in the cultivated land remote sensing satellite image at all times can be combined to determine the real cultivated land road factor.

[0089] Preferably, in one embodiment of the present invention, the method for obtaining the real farmland road factor includes:

[0090] Because farmland vegetation changes with the seasons, the straight line corresponding to the crop arrangement in the summer time period may not exist in the cultivated land remote sensing satellite image in the winter time period. In this case, the road topology relationship value of the straight line in the cultivated land remote sensing satellite image in the winter time period is recorded as 0. In other words, when a suspected cultivated land road straight line does not exist in the cultivated land remote sensing satellite image at a certain time, the road topology relationship value of the suspected cultivated land road straight line in the cultivated land remote sensing satellite image at that time is recorded as 0. At this time, each suspected cultivated land road straight line has a road topology relationship value in the cultivated land remote sensing satellite image at each time.

[0091] Then, the mean of the road topological relationship values ​​of each suspected cultivated land road line in the cultivated land remote sensing satellite images at all times is taken as the mean eigenvalue. The mean eigenvalue reflects the average performance of the topological structure of the suspected cultivated land road line in the cultivated land remote sensing satellite images at all times, which is helpful for the subsequent analysis of the stability of the suspected cultivated land road line at different times.

[0092] Then, the absolute value of the difference between the road topological relationship value in the cultivated land remote sensing satellite image at each time and the corresponding mean eigenvalue of each suspected cultivated land road line is taken as the deviation factor. The larger the deviation factor, the greater the deviation, indicating that there is a greater deviation between the topological structure of the suspected cultivated land road line in the cultivated land remote sensing satellite image at each time and the average level of the topological structure in the cultivated land remote sensing satellite image at all times. In this case, the topological structure of the suspected cultivated land road line has undergone a large fluctuation over time, and the structural stability is poor, and the possibility of the line corresponding to the real cultivated land road is lower.

[0093] Finally, the deviation factors of each suspected farmland road line at all times are combined to provide an overall evaluation index. The mean of all deviation factors of each suspected farmland road line is negatively correlated and normalized to achieve logical relationship correction. The corrected value is used as the real farmland road factor. At this time, when the real farmland road factor is larger, the topological structure of the suspected farmland road line at all times is more consistent, that is, it is less affected by seasonal changes, and it is more likely to represent a real farmland road. It should be noted that the normalization method here can be used Normalization function, negative correlation mapping and normalization can be used function, represents an exponential function with the natural constant e as the base, Represents the independent variable.

[0094] At this point, by analyzing the topological relationship between the suspected cultivated road lines in the cultivated land remote sensing satellite images, the real cultivated road factor of each suspected cultivated road line can be obtained, which is used to measure the possibility that each suspected cultivated road line represents the real cultivated road.

[0095] Step S4: Fusion analysis of all voting values ​​of the parameter combination corresponding to each suspected cultivated land road straight line and the real cultivated land road factor to obtain the real cultivated land road evaluation value; based on the real cultivated land road evaluation value, the real cultivated land road is screened out from all suspected cultivated land road straight lines.

[0096] The voting value of each parameter combination obtained based on step S2 can preliminarily represent the probability of it being a straight line in the remote sensing satellite image of cultivated land. The real cultivated land road factor obtained based on step S3 measures the probability of each straight line representing a real cultivated land road. Therefore, step S4 can integrate these two indicators to determine a more comprehensive and accurate indicator for screening out straight lines that can represent real cultivated land roads from all suspected cultivated land road straight lines.

[0097] First, all voting values ​​of the parameter combination corresponding to each suspected farmland road line and the real farmland road factor are integrated and analyzed to obtain the real farmland road evaluation value. Preferably, in one embodiment of the present invention, the method for obtaining the real farmland road evaluation value includes:

[0098] When the voting value and the actual cultivated land road factor are combined, a weight value can be preset to balance the importance between the two.

[0099] The product of the real cultivated land road factor of each suspected cultivated land road line and the preset weight value is used as the first real cultivated land road assessment factor. The larger the real cultivated land road factor, the more likely the suspected cultivated land road line is to represent the real cultivated land road. Therefore, the larger the first real cultivated land road assessment factor, the higher the possibility that the suspected cultivated land road line represents the real cultivated land road.

[0100] In the process of screening suspected cultivated land road lines in step S2, the voting values ​​of the parameter combination in the cultivated land remote sensing satellite images at all times are integrated to obtain the score rate. Therefore, the product of the value after negative correlation mapping of the preset weight value and the score rate of each suspected cultivated land road line is used as the second real cultivated land road evaluation factor. The higher the score rate, the more likely it is that the straight line will appear in the cultivated land remote sensing satellite image, and the higher the possibility that it represents a real cultivated land road.

[0101] Finally, the sum of the first real farmland road assessment factor and the second real farmland road assessment factor of each suspected farmland road line is normalized to the value obtained as the real farmland road assessment value of each suspected farmland road line. The larger the real farmland road assessment value, the more the suspected farmland road line can represent the real farmland road. The normalization method here can be used Normalization function.

[0102] It should be noted that in the embodiment of the present invention, since the structure of the cultivated land road is relatively stable over time, the weight of the real cultivated land road factor calculated based on the topological structure is increased, so the preset weight value is 0.6. In other embodiments of the present invention, the numerical value of the weight value can be adjusted, but it should be ensured to be no less than 0.5.

[0103] After obtaining the real farmland road evaluation value of each suspected farmland road straight line, the real farmland roads can be screened out from all the suspected farmland road straight lines based on the index.

[0104] Preferably, in one embodiment of the present invention, the method for obtaining the real farmland road includes:

[0105] Based on the above analysis, it can be seen that the larger the real cultivated land road assessment value is, the higher the possibility that the suspected cultivated land road line represents the real cultivated land road. Therefore, among all suspected cultivated land road lines, the suspected cultivated land road line with a real cultivated land road assessment value greater than the preset road threshold is taken as the real cultivated land road line, so as to obtain the corresponding real cultivated land road according to the position of the real cultivated land road line in the cultivated land remote sensing satellite image. It should be noted that the preset road threshold is set to 0.7, and the specific value can be adjusted according to the implementation scenario, which is not limited here.

[0106] After obtaining the real cultivated land roads, we can analyze the proportion of cultivated land roads in the cultivated land area to understand the current status of cultivated land utilization.

[0107] In summary, when detecting cultivated land roads in remote sensing satellite images, the embodiment of the present invention further analyzes the difference between the linear characteristics of cultivated land roads and the linear characteristics of farmland crop arrangement based on Hough line detection, thereby helping to obtain more accurate cultivated land road detection results. In view of the seasonal characteristics of farmland crops, cultivated land remote sensing satellite images of the same cultivated land area at different times are obtained as the basis for the subsequent analysis process. Then, based on the linear characteristics of cultivated land roads, Hough line detection is performed on the cultivated land remote sensing satellite images at each time, and the voting value of each parameter combination is obtained in the Hough space, and each parameter combination represents each straight line; and the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is combined to preliminarily screen out suspected cultivated land road straight lines. At this time, the suspected cultivated land road straight lines include the straight lines corresponding to the cultivated land roads and the straight lines corresponding to the farmland crops, which need to be further distinguished. Because the distribution characteristics of cultivated land roads usually follow a certain topological relationship and are interspersed in the square-shaped cultivated land farmland, they often present a grid-like topological feature. Therefore, the position distribution relationship between the suspected cultivated land road straight lines is analyzed, and the road topological relationship value of each suspected cultivated land road straight line is determined. Further, since the crops in the farmland have seasonal characteristics, and the cultivated land roads are relatively fixed, the topological relationship of the cultivated land roads will not change with the change of time and season. Therefore, at all times, the fluctuation of the road topological relationship value of the suspected cultivated land road straight line is analyzed to determine the real cultivated land road factor. At this time, the real cultivated land road factor can be used as a basis for evaluating whether the suspected cultivated land road straight line represents the real cultivated land road. Finally, all the voting values ​​of the parameter combination corresponding to each suspected cultivated land road straight line and the real cultivated land road factor are fused to obtain the real cultivated land road evaluation value, thereby screening out the real cultivated land road. In summary, the embodiment of the present invention combines multi-phase characteristics and the topological relationship of cultivated land roads, effectively improving the accuracy of cultivated land road detection.

[0108] The present invention also provides a farmland road detection system based on remote sensing satellite images. Figure 4 , the system includes:

[0109] Image data acquisition module 401, used to acquire cultivated land remote sensing satellite images of the same cultivated land area at different times;

[0110] The suspected cultivated land road screening module 402 is used to detect the cultivated land remote sensing satellite image at each time based on the Hough linear transformation algorithm, and obtain the voting value of each parameter combination in the Hough space; the voting values ​​of the parameter combinations in the cultivated land remote sensing satellite images at all times are integrated to screen out the suspected cultivated land road lines;

[0111] The cultivated land road topology analysis module 403 is used to analyze the position distribution relationship between the suspected cultivated land road lines in the cultivated land remote sensing satellite images at each time, and determine the road topology relationship value of each suspected cultivated land road line; in the cultivated land remote sensing satellite images at all times, based on the fluctuation between the road topology relationship values ​​of each suspected cultivated land road line, determine the real cultivated land road factor;

[0112] The real cultivated land road screening module 404 is used to integrate and analyze the voting value of the parameter combination corresponding to each suspected cultivated land road line and the real cultivated land road factor to obtain the real cultivated land road evaluation value; based on the real cultivated land road evaluation value, the real cultivated land road line is screened out from all suspected cultivated land road lines.

[0113] It should be noted that: the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the farmland road detection system based on remote sensing satellite images and the farmland road detection method based on remote sensing satellite images provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0114] See also Figure 5 , which shows a schematic structural diagram of a farmland road detection device based on remote sensing satellite images provided by an embodiment of the present invention, including a processor 500, a memory 501, a bus 502 and a communication interface 503, wherein the processor 500, the communication interface 503 and the memory 501 are connected via the bus 502; wherein the memory 501 may include a high-speed random access memory, the bus 502 may be an ISA bus, a PCI bus or an EISA bus, etc., and the processor 500 may be an integrated circuit chip with signal processing capabilities; the memory 501 stores at least one instruction, at least one program, a code set or an instruction set, and when the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor, the steps in a farmland road detection method based on remote sensing satellite images are implemented.

[0115] 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.

[0116] 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 method for detecting cultivated land roads based on remote sensing satellite images, characterized in that: The method comprises: Obtain remote sensing satellite images of cultivated land in the same cultivated land area at different times; Based on the Hough linear transformation algorithm, the cultivated land remote sensing satellite images at each time are detected, and the voting value of each parameter combination is obtained in the Hough space; the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is integrated to screen out suspected cultivated land road lines; In the cultivated land remote sensing satellite images at each time, the position distribution relationship between the suspected cultivated land road lines is analyzed to determine the road topological relationship value of each suspected cultivated land road line; in the cultivated land remote sensing satellite images at all times, the real cultivated land road factor is determined based on the fluctuation between the road topological relationship values ​​of each suspected cultivated land road line; All voting values ​​of the parameter combination corresponding to each suspected cultivated land road straight line and the real cultivated land road factor are integrated and analyzed to obtain the real cultivated land road evaluation value; based on the real cultivated land road evaluation value, the real cultivated land road is screened out from all suspected cultivated land road straight lines; The method for obtaining the road topology relationship value comprises: In the cultivated land remote sensing satellite image at each time, the intersection points between all the suspected cultivated land road lines are obtained; For any intersection, calculate the differences between all angles formed by the intersection to obtain the road topology factor corresponding to each intersection; For any suspected cultivated land road straight line in the cultivated land remote sensing satellite image at each time, the road topology factors corresponding to all intersection points on the suspected cultivated land road straight line are averaged to obtain the road topology relationship value of the suspected cultivated land road straight line; The method for obtaining the road topology factor includes: For any intersection point, among all the angles formed by the intersection point, combine any two angles to obtain all non-repeating angle combinations; Calculate the absolute value of the difference between the two angles in each angle combination as the angle deviation value; Based on the angle deviation values ​​of all angle combinations corresponding to the intersection, a road topology factor corresponding to the intersection is determined, and the road topology factor is negatively correlated with the angle deviation value; The method for obtaining the real cultivated land road factor includes: When a suspected cultivated land road line does not exist in the cultivated land remote sensing satellite image at a certain time, the road topological relationship value of the suspected cultivated land road line in the cultivated land remote sensing satellite image at that time is recorded as 0; The mean of the road topological relationship values ​​of each suspected farmland road line in the farmland remote sensing satellite images at all times is taken as the mean eigenvalue; The absolute value of the difference between the road topological relationship value of each suspected cultivated land road line in the cultivated land remote sensing satellite image at each time and the mean characteristic value is used as a deviation factor; The mean values ​​of all deviation factors of each suspected farmland road straight line are negatively correlated and normalized to obtain a value which is used as the true farmland road factor.

2. The method for detecting cultivated land roads based on remote sensing satellite images according to claim 1, characterized in that: The method for obtaining the suspected farmland road straight line includes: The normalized value of the mean of the voting values ​​of each parameter combination in the cultivated land remote sensing satellite images at all times is used as the score rate of the straight line corresponding to each parameter combination; Among the straight lines corresponding to all parameter combinations, the straight line corresponding to the score rate greater than the preset threshold is taken as the suspected farmland road straight line.

3. The method for detecting cultivated land roads based on remote sensing satellite images according to claim 2, characterized in that: The method for obtaining the real farmland road evaluation value includes: The product of the real cultivated land road factor of each suspected cultivated land road line and the preset weight value is used as the first real cultivated land road assessment factor; The product of the value after negative correlation mapping of the preset weight value and the score rate of each suspected farmland road straight line is used as the second real farmland road evaluation factor; The normalized value of the sum of the first real farmland road assessment factor and the second real farmland road assessment factor of each suspected farmland road straight line is used as the real farmland road assessment value of each suspected farmland road straight line.

4. The method for detecting cultivated land roads based on remote sensing satellite images according to claim 1, characterized in that: The method for obtaining the real farmland road comprises: Among all the suspected farmland road straight lines, the suspected farmland road straight lines whose real farmland road evaluation values ​​are greater than a preset road threshold are taken as real farmland road straight lines, so as to obtain the real farmland road corresponding to the real farmland road straight lines.

5. The method for detecting cultivated land roads based on remote sensing satellite images according to claim 1, characterized in that: The method for obtaining the voting value includes: When detecting the cultivated land remote sensing satellite image at each time based on the Hough linear transformation algorithm, in the Hough space, whenever the edge pixel point meets a certain parameter combination, the corresponding value of the parameter combination in the accumulator array is added by 1, thereby obtaining the corresponding value of each parameter combination in the accumulator array; The corresponding values ​​of each parameter combination in the accumulator array are normalized and used as the voting value of each parameter combination in the cultivated land remote sensing satellite image at each time.

6. A farmland road detection system based on remote sensing satellite images, characterized in that: The system comprises: An image data acquisition module is used to acquire remote sensing satellite images of cultivated land in the same cultivated land area at different times; The suspected cultivated land road screening module is used to detect the cultivated land remote sensing satellite images at each time based on the Hough linear transformation algorithm, and obtain the voting value of each parameter combination in the Hough space; the voting value of each parameter combination in the cultivated land remote sensing satellite images at all times is integrated to screen out the suspected cultivated land road lines; The cultivated land road topology analysis module is used to analyze the position distribution relationship between the suspected cultivated land road lines in the cultivated land remote sensing satellite images at each time, and determine the road topology relationship value of each suspected cultivated land road line; in the cultivated land remote sensing satellite images at all times, based on the fluctuation between the road topology relationship values ​​of each suspected cultivated land road line, determine the real cultivated land road factor; A real cultivated land road screening module is used to integrate and analyze all voting values ​​of the parameter combination corresponding to each suspected cultivated land road line and the real cultivated land road factor to obtain a real cultivated land road evaluation value; based on the real cultivated land road evaluation value, a real cultivated land road line is screened out from all suspected cultivated land road lines; The method for obtaining the road topology relationship value comprises: In the cultivated land remote sensing satellite image at each time, the intersection points between all the suspected cultivated land road lines are obtained; For any intersection, calculate the differences between all angles formed by the intersection to obtain the road topology factor corresponding to each intersection; For any suspected cultivated land road straight line in the cultivated land remote sensing satellite image at each time, the road topology factors corresponding to all intersection points on the suspected cultivated land road straight line are averaged to obtain the road topology relationship value of the suspected cultivated land road straight line; The method for obtaining the road topology factor includes: For any intersection point, among all the angles formed by the intersection point, combine any two angles to obtain all non-repeating angle combinations; Calculate the absolute value of the difference between the two angles in each angle combination as the angle deviation value; Based on the angle deviation values ​​of all angle combinations corresponding to the intersection, a road topology factor corresponding to the intersection is determined, and the road topology factor is negatively correlated with the angle deviation value; The method for obtaining the real cultivated land road factor includes: When a suspected cultivated land road line does not exist in the cultivated land remote sensing satellite image at a certain time, the road topological relationship value of the suspected cultivated land road line in the cultivated land remote sensing satellite image at that time is recorded as 0; The mean of the road topological relationship values ​​of each suspected farmland road line in the farmland remote sensing satellite images at all times is taken as the mean eigenvalue; The absolute value of the difference between the road topological relationship value of each suspected cultivated land road line in the cultivated land remote sensing satellite image at each time and the mean eigenvalue is used as a deviation factor; The mean values ​​of all deviation factors of each suspected farmland road straight line are negatively correlated and normalized to obtain a value which is used as the true farmland road factor.

7. A farmland road detection device based on remote sensing satellite images, characterized in that: It includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and when the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor, the steps of a method for detecting cultivated land roads based on remote sensing satellite images as described in any one of claims 1 to 5 are implemented.

Images