An intelligent processing method for UAV remote sensing mapping images

Abstract

The present invention relates to the field of image processing technology, and in particular to an intelligent processing method for unmanned aerial vehicle remote sensing mapping images, comprising: using an unmanned aerial vehicle to collect remote sensing mapping images, and performing preprocessing to generate a target image to be analyzed; generating a target window based on the target image, obtaining edge lines in the target window, obtaining an edge pixel sequence, and performing a primary screening on the edge pixel sequence to generate a first set; performing a secondary screening based on the first set to generate a second set; calculating the integrity of the edge lines in the target window based on the correlation of the second set, and calculating the clarity of the image in the target window by averaging regularity; calculating the clarity of the image in the target window in the same manner according to different contrast limit parameters, and enhancing the target window by analyzing different clarity; moving the target window in the target image, and repeating the above steps until the enhancement of the target image is completed. The present invention can improve the image processing effect with thin clouds and fog.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an intelligent processing method for unmanned aerial vehicle remote sensing mapping images. Background Art

[0002] In recent years, UAV remote sensing technology has developed rapidly and has become an indispensable tool in many fields. Its high efficiency and flexibility have made it widely used in many fields such as agriculture, environmental monitoring, urban planning, archaeological research and disaster management. In the agricultural field, UAVs can monitor crop growth and soil environment in real time, thereby optimizing farmland management. In environmental monitoring, UAVs effectively assess ecosystem changes and pollution. In addition, in the field of urban planning, UAVs provide high-precision topographic mapping data to help infrastructure construction and development strategy formulation. These applications highlight the great potential of UAV remote sensing technology in improving data collection efficiency and assisting decision-making. In scenarios such as monitoring crop growth, evaluating agricultural land use, and land rights confirmation, UAV remote sensing technology is often used for arable land mapping.

[0003] However, when using UAV remote sensing technology to monitor crop growth and map cultivated land, environmental factors such as interference from thin clouds and fog often cause remote sensing images to be blurred, affecting mapping efficiency. To solve this problem, traditional technology usually uses histogram equalization for image enhancement, but due to the uneven distribution of thin cloud and fog concentrations in the image, the global histogram equalization effect is limited, resulting in insignificant image enhancement effects. Summary of the invention

[0004] In order to solve the technical problem that the existing methods are not effective in processing images with thin clouds and fog in agriculture, the purpose of the present invention is to provide an intelligent processing method for unmanned aerial vehicle remote sensing mapping images. The technical solutions adopted are as follows:

[0005] Step S1: using a drone to collect remote sensing images and preprocessing to generate a target image to be analyzed;

[0006] Step S2: generating a target window based on the target image, obtaining edge lines in the target window, obtaining an edge pixel sequence, performing a primary screening on the edge pixel sequence to generate a first set; performing a secondary screening on the first set to generate a second set;

[0007] Step S3: calculating the correlation based on the second set to determine the integrity of the edge line in the target window, and calculating the clarity of the image in the target window by averaging the regularity;

[0008] Step S4: similarly calculating the clarity of the image in the target window according to different contrast limit parameters, and enhancing the target window by analyzing different clarity;

[0009] Step S5: Move the target window in the target image, and repeat steps S2 to S4 until the target image is enhanced.

[0010] Preferably, step S2 includes:

[0011] Step S21: dividing the target image into a number of equilateral windows, wherein four adjacent equilateral windows of any equilateral window form a large window, defining any large window as a target window, performing edge detection on the image in the target window to obtain edge lines, and obtaining edge pixel sequences;

[0012] Step S22: performing feature analysis on the edge pixel sequence and calculating the regularity, and performing a screening based on the regularity to generate a first set;

[0013] Step S23: define any edge pixel sequence in the first set as a target sequence, calculate the probability that the target sequence is parallel to any non-target sequence in the first set, and obtain the probability that the two are collinear. Similarly, obtain the probability that the target sequence is collinear with all remaining non-target sequences, and generate the second set after secondary screening.

[0014] Preferably, step S22 includes:

[0015] All edge pixel sequences in the target window are split into multiple pixel sequences according to the branch points, the original chain code is generated based on the pixel sequence, and the original chain code is split into two segments according to the absolute value of the first-order differential chain code in any original chain code;

[0016] Calculate the length of the pixel sequence and determine whether the pixel sequence is real information. If so, generate the original set;

[0017] The regularity of all pixel sequences in the original set is calculated, and the pixel sequences are screened based on the regularity to generate a first set.

[0018] Preferably, the corresponding calculation formula of regularity is:

[0019] The original chain code representation that defines any pixel sequence in the original set is:

[0020]

[0021] in, Represents a pixel sequence Regularity of Represents a pixel sequence Any two adjacent items of the original chain code; Represents an exponential function with a natural constant as its base.

[0022] Preferably, in step S23, calculating the probability that the target sequence is parallel to any non-target sequence in the first set includes:

[0023] Define any edge pixel sequence in the first set as the target sequence, obtain any non-target sequence in the first set, connect the first and last pixel points of the target sequence and the non-target sequence to obtain pixel line segment 1 and pixel line segment 2, obtain the distance between all pixel points of pixel line segment 2 and the extended pixel line segment 1, calculate the variance based on all distances, and calculate the probability that the target sequence and the non-target sequence are parallel. The corresponding calculation formula is:

[0024]

[0025] in, Represents the target sequence in the first set and non-target sequences Parallel probability; represents variance; Represents an exponential function with a natural constant as its base.

[0026] Preferably, in step S23, the probability of the two being collinear is obtained, and a second set is generated through secondary screening, and the corresponding calculation formula is:

[0027]

[0028] in, Represents the target sequence in the first set and non-target sequences Probability of collinearity; Indicates the pixel line segment 2 The pixel point and the extended pixel line segment are one; Represents the total distance between all pixel points of pixel line segment 2 and the extended pixel line segment 1; Represents the target sequence in the first set and non-target sequences Parallel probability; represents an exponential function with a natural constant as base; represents the maximum and minimum normalized function;

[0029] Similarly, the probability that the target sequence is collinear with all non-target sequences is obtained, and secondary screening is performed based on the probability of collinearity to generate the second set.

[0030] Preferably, in step S3, the correlation is calculated as:

[0031] The second set includes several small sets, and any small set is defined as the target set. The second set is , and the rectangular coordinate representation corresponding to the angular characteristic direction of the second set is obtained as , calculate the correlation as:

[0032]

[0033] in, Represents the target set relevance; represents the total number of small sets in the second set; Represents the target set With any non-target set The probability that the angle characteristics satisfy parallelism; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

[0034] Preferably, in step S3, determining the integrity of the edge line within the target window includes:

[0035] Based on the correlation, the correlation between the target set and all the remaining non-target sets is calculated in the same way. The correlation is used to determine whether all the small sets in the second set are real information. If so, three screenings are performed to generate the third set, where the third set is , the number of pixel sequences contained in each small set is , and the total length of each pixel sequence is obtained as , the calculation formula corresponding to the integrity of the edge line in the target window is:

[0036]

[0037] in, Indicates the target window The integrity of the inner edge line; Represents the third set Small and medium collection The number of pixel sequences included; Represents the third set The total number of small and medium sets; Represents the third set Small and medium collection The total length of the pixel sequence contained; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

[0038] Preferably, calculating the clarity of the image in the target window by averaging the regularity includes:

[0039] Get the average regularity of the pixel sequence contained in each small set in the third set as , calculate the clarity of the image in the target window, the corresponding calculation formula is:

[0040]

[0041] in, Indicates the target window The clarity of the internal image; Indicates the target window The integrity of the inner edge line; Represents the third set Small and medium collection The average regularity of the contained pixel sequences; Represents the third set The total number of small and medium sets; Represents the maximum and minimum normalization function.

[0042] Preferably, in step S4, the value range of the contrast limit parameter is [2, 20], the clarity of the image in the target window is calculated by traversing the contrast limit parameters based on step S2 and step S3, and the images are sorted from high to low, and the target window is enhanced with the highest clarity.

[0043] The present invention has the following beneficial effects:

[0044] The present application defines a histogram equalization window, calculates the clarity within the window, performs feature analysis first, calculates the regularity of the pixel sequence, and then calculates the correlation of each edge pixel sequence set, screens out the pixel sequence that meets the edge line of the farmland area, calculates the integrity of the image in the window based on the length of these pixel sequences and the number of sequences, and finally calculates the clarity of the image based on the regularity, obtains the image clarity under different contrast limit parameters, and takes the enhancement result with the highest clarity as the result of the window, thereby achieving enhancement of the entire image, and can perform intelligent adjustment according to the different blurred areas caused by different concentrations of clouds and fog in the image to achieve a higher enhancement effect, thereby ensuring the accuracy of image quality for analyzing the growth status of crops, realizing real-time monitoring of crop growth, and providing strong technical support for agricultural production. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] 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.

[0046] Figure 1 A schematic diagram of the steps of an intelligent processing method for remote sensing and mapping images of a UAV provided by an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of agricultural land with thin clouds in an intelligent processing method for UAV remote sensing mapping images provided by an embodiment of the present invention;

[0048] Figure 3 An enhanced farmland remote sensing image according to an intelligent processing method for UAV remote sensing mapping images provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] 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 specific implementation method, structure, features and effects of an intelligent processing method for remote sensing and mapping images of unmanned aerial vehicles proposed by the present invention in combination 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.

[0050] 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.

[0051] The specific scheme of the intelligent processing method for UAV remote sensing mapping images provided by the present invention is described in detail below with reference to the accompanying drawings.

[0052] See also Figure 1 , which shows a schematic diagram of the steps of an intelligent processing method for remote sensing and mapping images of a drone provided by an embodiment of the present invention, the method comprising:

[0053] Step S1: using a drone to collect remote sensing images and preprocessing to generate a target image to be analyzed;

[0054] Step S2: generating a target window based on the target image, obtaining edge lines in the target window, obtaining an edge pixel sequence, performing a primary screening on the edge pixel sequence to generate a first set; performing a secondary screening on the first set to generate a second set;

[0055] Step S3: calculating the correlation based on the second set to determine the integrity of the edge line in the target window, and calculating the clarity of the image in the target window by averaging the regularity;

[0056] Step S4: similarly calculating the clarity of the image in the target window according to different contrast limit parameters, and enhancing the target window by analyzing different clarity;

[0057] Step S5: Move the target window in the target image, and repeat steps S2 to S4 until the target image is enhanced.

[0058] To better explain, drone remote sensing mapping refers to the technology of using sensors and camera equipment carried by drones to obtain ground information from the air. It can fly into areas that are difficult to reach with traditional aerial photography, and can fly at a lower altitude to capture more details to obtain high-resolution and high-precision geospatial data. It is widely used in terrain mapping, agricultural monitoring, urban planning, disaster assessment and other fields. In terms of agricultural monitoring, drone remote sensing mapping can help farmers understand the growth status of crops in a timely manner, discover problems such as pests and diseases and lack of water, and take corresponding measures.

[0059] See also Figure 2 , which shows a schematic diagram of agricultural land with thin clouds in a method for intelligent processing of UAV remote sensing images provided by an embodiment of the present invention; specifically, in step S1, a UAV equipped with a high-resolution camera is used to collect images of agricultural land to effectively obtain detailed image data of the farmland; then the collected images are subjected to preprocessing operations such as cropping to ensure the quality and applicability of the images, wherein cropping refers to adjusting the image to a size suitable for subsequent window analysis; in addition, the preprocessing operation also includes grayscale processing to simplify the image analysis data, thereby generating a target image to be analyzed.

[0060] It can be understood that in this embodiment, the main purpose is to solve the problem of poor enhancement effect of global histogram equalization, so local histogram equalization is adopted; and in the window division process, there are mainly two division methods, namely, sub-block overlapping method and sub-block non-overlapping method, among which the sub-block non-overlapping method can reduce the amount of calculation, but may cause block effects, that is, at the junction of adjacent sub-blocks, there may be a large difference due to different grayscale distributions, resulting in a lack of smooth transition, affecting the overall appearance and realism, while the sub-block overlapping method has an unobvious block effect, but it will lead to a large amount of calculation, so in this embodiment, the two are combined and a partial overlapping method is adopted, taking into account both the amount of calculation and the image smoothness.

[0061] It should be noted that histogram equalization is mainly used to improve the contrast of an image by adjusting the histogram distribution of the image so that it is evenly distributed throughout the grayscale range, thereby achieving the effect of enhancing image details and contrast; that is, by remapping the grayscale of the original image, the probability distribution of each grayscale level is made more uniform, thereby optimizing the brightness and contrast of the image.

[0062] Furthermore, step S2 includes:

[0063] Step S21: dividing the target image into a number of equilateral windows, wherein four adjacent equilateral windows of any equilateral window form a large window, defining any large window as a target window, performing edge detection on the image in the target window to obtain edge lines, and obtaining edge pixel sequences;

[0064] Step S22: performing feature analysis on the edge pixel sequence and calculating the regularity, and performing a screening based on the regularity to generate a first set;

[0065] Step S23: define any edge pixel sequence in the first set as a target sequence, calculate the probability that the target sequence is parallel to any non-target sequence in the first set, and obtain the probability that the two are collinear. Similarly, obtain the probability that the target sequence is collinear with all remaining non-target sequences, and generate the second set after secondary screening.

[0066] As an optional implementation, in this embodiment, the side length of the equilateral window is 15.

[0067] Specifically, the target image is divided into A number of equilateral windows, each of which is composed of four adjacent equilateral windows. As the basic window for local histogram equalization, histogram equalization is performed on the image in the large window, that is, any large window is defined as the target window for analysis.

[0068] It is understandable that in remote sensing images of agricultural land, due to the changes in the density of clouds and fog in different areas, the clarity is higher in some areas and lower in some areas; and in remote sensing images of farmland, due to the existence of paths such as roads, canals, and rivers, the farmland is divided into areas with regular shapes and distribution patterns, such as rectangles, polygons, or strips, etc. The more the image in the target window meets these characteristics and the higher the completeness, the higher the clarity; therefore, the image edge in the target window is first obtained for analysis.

[0069] Preferably, the edge detection method adopts Canny edge detection, that is, identifying the boundaries of objects in the image within the target window, and the Canny edge detection algorithm is a multi-stage edge detection algorithm with high edge detection accuracy and low error rate. It can detect most of the important edges in the image while suppressing noise interference. After the edge detection is completed, the edge line that needs to be analyzed can be obtained, and multiple edge pixel sequences are generated based on the edge line.

[0070] To explain, in step S22, regularity refers to the orderliness and consistency of the pixel sequence in image processing and analysis, which describes whether the spatial distribution of pixels shows a certain regularity and to what extent this regularity can be recognized and utilized; for example, in a regular texture image, the regularity of the pixel sequence is higher because the pixel arrangement shows obvious periodicity and repetitiveness; while in random noise or complex scenes, the regularity of the pixel sequence is lower because the pixel arrangement lacks obvious regularity.

[0071] Furthermore, step S22 includes:

[0072] All edge pixel sequences in the target window are split into multiple pixel sequences according to the branch points, the original chain code is generated based on the pixel sequence, and the original chain code is split into two segments according to the absolute value of the first-order differential chain code in any original chain code;

[0073] Calculate the length of the pixel sequence and determine whether the pixel sequence is real information. If so, generate the original set;

[0074] The regularity of all pixel sequences in the original set is calculated, and the pixel sequences are screened based on the regularity to generate a first set.

[0075] To explain, chain code is a specific coding method whose main functions are verification and execution to enhance data processing.

[0076] Specifically, firstly, all edge pixel sequences in the target window are extracted, and split according to each branch point to obtain multiple pixel sequences, wherein when there are more than or equal to 3 pixels in the eight-neighborhood of a certain pixel point, the pixel point is a branch point; that is, all edge pixel sequences in the target window are traversed, and the eight-neighborhood of each pixel point (the eight pixels around the pixel point) is checked, and the number of pixels in the eight-neighborhood of each pixel point is counted. If there are more than or equal to three pixels in the eight-neighborhood of a certain pixel point, it is marked as a branch point, and the edge pixel sequence is split based on all branch points.

[0077] Next, all the split pixel sequences are represented in the 8-chain code to generate the original chain code, such as a pixel sequence in the target window The chain code is , the absolute value of the first-order differential chain code is calculated as , for items greater than or equal to 2, split the pixel sequence according to the subscript, such as Term Difference Chain Code When it is greater than or equal to 2, the original chain code is split into two chain codes, which are recorded as , ,According to this method, the original chain codes of all pixel sequences in the target window are split into two chain codes.

[0078] Secondly, calculate the length of the pixel sequence and set the threshold , preferably, in this embodiment, ; Determine whether the pixel sequence is real information, that is, when the length of the pixel sequence is less than the threshold When the length of the pixel sequence is greater than the threshold, the pixel sequence is considered as non-real information and is excluded. All pixel sequences of constitute the original set, denoted as .

[0079] Then, calculate the original set The regularity of any pixel sequence in , further, the corresponding calculation formula of the regularity is:

[0080] The original chain code representation that defines any pixel sequence in the original set is:

[0081]

[0082] in, Represents a pixel sequence Regularity of Represents a pixel sequence Any two adjacent items of the original chain code; Represents an exponential function with a natural constant as its base.

[0083] It is explained that the smaller the change of the pixel sequence is, the greater the regularity of the pixel sequence is; wherein, express The differential chain code of these two items; Represents the cumulative difference chain code of all adjacent items. The larger the cumulative value, the smaller the pixel sequence. The more complex the change, the more pixel sequence The smaller the regularity.

[0084] Finally, calculate the original set in the same way The regularity of all pixel sequences in is recorded as , set the threshold , preferably, in this embodiment, , when the regularity of the pixel sequence is less than the threshold When , the pixel sequence is irregular and may be a pseudo edge line, and it is excluded; on the contrary, when the regularity of the pixel sequence is greater than the threshold , it may meet the characteristics of effective information such as roads, canals, rivers, etc., and a screening is performed based on this judgment condition to generate the first set, that is, the regularity of the pixel sequence is greater than the threshold All pixel sequences of constitute the first set, denoted as .

[0085] It is understandable that farmland images with clouds and mist appear blurred, and the edges of effective information may appear incoherent due to the blur. Therefore, in the image, when two or more pixel sequences are collinear, these pixel sequences may be the same edge line, so it is necessary to group these collinear edge lines and then determine the correlation between each group of pixel sequences.

[0086] It is explained that when two sets of pixel sequences are approximately parallel or vertical, it is considered that the two sets of pixel sequences have good correlation and meet the farmland edge characteristics.

[0087] Furthermore, in step S23, the probability that the target sequence is parallel to any non-target sequence in the first set is calculated, including:

[0088] Define any edge pixel sequence in the first set as the target sequence, obtain any non-target sequence in the first set, connect the first and last pixel points of the target sequence and the non-target sequence to obtain pixel line segment 1 and pixel line segment 2, obtain the distance between all pixel points of pixel line segment 2 and the extended pixel line segment 1, calculate the variance based on all distances, and calculate the probability that the target sequence and the non-target sequence are parallel. The corresponding calculation formula is:

[0089]

[0090] in, Represents the target sequence in the first set and non-target sequences Parallel probability; represents variance; Represents an exponential function with a natural constant as its base.

[0091] Specifically, the target sequence is , and then obtain any non-target sequence in the first set as , connect the first and last pixel points of the target sequence and the non-target sequence to get a pixel line segment and pixel line segment two , in order to simplify the data and facilitate the analysis of the collective characteristics of the two pixel sequences, the pixel line segments are Extend to a straight line , it can be explained that here it can be extended to the boundary of the target window; record pixel line segment two All pixels on the line The distances are , and calculate the variance based on all distance values, recorded as , and then calculate the first set Target sequence and non-target sequences The probability of parallelism; where the variance The smaller the value, the more pixel segments All pixels on the line The more uniform the distance is, the more parallel the two pixel sequences are. The greater the probability of parallelism.

[0092] Furthermore, in step S23, the probability of the two being collinear is obtained, and a second set is generated through secondary screening. The corresponding calculation formula is:

[0093]

[0094] in, Represents the target sequence in the first set and non-target sequences Probability of collinearity; Indicates the pixel line segment 2 The pixel point and the extended pixel line segment are one; Represents the total distance between all pixel points of pixel line segment 2 and the extended pixel line segment 1; Represents the target sequence in the first set and non-target sequences Parallel probability; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

[0095] To explain, Represents pixel line segment two All pixels on the line The shorter the average distance, the smaller the pixel sequence. The greater the probability of collinearity; Represents a sequence of two pixels The probability of meeting parallelism, pixel sequence The more parallel they are, the more they satisfy the collinearity characteristic.

[0096] Specifically, the probability of the target sequence being collinear with all non-target sequences is obtained in the same way, that is, traversing the first set For all pixel sequences in the image, calculate the probability of collinearity between the target sequence and all other non-target sequences in turn, and set the threshold , preferably, in this embodiment, , the probability value of collinearity is greater than the threshold The pixel sequence is screened twice to exclude it from the first set , and then the probability values ​​of all remaining collinearity are less than the threshold The pixel sequence is divided into a group to generate the second set, denoted as , based on the second set The line segment connecting the first and last pixel points of any pixel sequence in is taken as the angular feature direction of the pixel sequence, and its rectangular coordinate representation is obtained as follows: , the value range is .

[0097] Further, in step S3, the correlation is calculated as:

[0098] The second set includes several small sets, and any small set is defined as the target set. The second set is , and the rectangular coordinate representation corresponding to the angular characteristic direction of the second set is obtained as , calculate the correlation as:

[0099]

[0100] in, Represents the target set relevance; represents the total number of small sets in the second set; Represents the target set With any non-target set The probability that the angle characteristics satisfy parallelism; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

[0101] It can be explained that the second set includes several small sets, that is, the small set refers to a pixel sequence. When the angle features of two small sets in the second set are close to parallel or vertical, it means that the correlation between the two small sets is high. Among them, the target set With the second set The probabilities that the angle features of all non-target sets in are parallel are ; Represents the target set With the second set The probability that the angle features of all small sets in the target set are parallel is greater. The greater the correlation with all other small sets, and the more sets that satisfy the parallel condition, the greater the correlation; Represents the target set With non-target set The difference between the rectangular coordinate representation of angles, It indicates the difference between the two and the right angle. The smaller the difference is, the more the pixel sequences in the two small sets are approximately vertical, and the greater the correlation between the two small sets is. Represents the target set With the second set The probability that the angle features of all small sets in the target set meet the verticality is greater. The greater the correlation with all other non-target sets.

[0102] Furthermore, in step S3, determining the integrity of the edge line within the target window includes:

[0103] Based on the correlation, the correlation between the target set and all the remaining non-target sets is calculated in the same way. The correlation is used to determine whether all the small sets in the second set are real information. If so, three screenings are performed to generate the third set, where the third set is , the number of pixel sequences contained in each small set is , and the total length of each pixel sequence is obtained as , the calculation formula corresponding to the integrity of the edge line in the target window is:

[0104]

[0105] in, Indicates the target window The integrity of the inner edge line; Represents the third set Small and medium collection The number of pixel sequences included; Represents the third set The total number of small and medium sets; Represents the third set Small and medium collection The total length of the pixel sequence contained; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

[0106] It can be explained that the correlation is used to determine whether all small sets in the second set are real information, and the threshold is set , preferably, in this embodiment, , when the correlation exceeds the threshold , the small set currently analyzed is considered to be true information, otherwise the correlation is less than the threshold , indicating that it is not real information, it is excluded, and three screenings are performed to generate the third set .

[0107] To explain, Represents the third set The average number of pixel sequences contained in each small set. The more the average number, the more fragmented the detected edge lines of the farmland area and the worse the integrity. Represents the third set The average length of the pixel sequence contained in each small set. The longer the average length, the more complete the detected edge line of the farmland area.

[0108] Furthermore, the clarity of the image in the target window is calculated by averaging the regularity, including:

[0109] Get the average regularity of the pixel sequence contained in each small set in the third set as , calculate the clarity of the image in the target window, the corresponding calculation formula is:

[0110]

[0111] in, Indicates the target window The clarity of the internal image; Indicates the target window The integrity of the inner edge line; Represents the third set Small and medium collection The average regularity of the contained pixel sequences; Represents the third set The total number of small and medium sets; Represents the maximum and minimum normalization function.

[0112] To illustrate, based on the above-mentioned calculation method of the regularity of the pixel sequence, the average regularity of all pixel sequences in each small set in the third set is obtained as ; Indicates the target window The integrity of the inner edge line. The better the integrity, the higher the clarity. Represents the third set The average regularity of the pixel sequences contained in each small set in the image is the mean value. The higher the mean value, the more regular the edge line of the farmland area detected in the target window is, the greater the degree of real information expression is, and the clearer the image is.

[0113] It can be understood that based on the above data analysis, for the target image, the more complete the edge line of the detected farmland area is, that is, the longer the length, the more regular the edge line is, and the higher the correlation between multiple edge lines, the greater the probability that the edge line is real information in the image, and the clearer the image is.

[0114] Furthermore, in step S4, the value range of the contrast limit parameter is [2, 20], the clarity of the image in the target window is calculated based on the contrast limit parameter traversed in step S2 and step S3, and the images are sorted from high to low, and the target window is enhanced with the highest clarity.

[0115] Specifically, the contrast limit parameter is set to 2, 3, 4 ... up to 20 in turn, and then the image under each parameter value is subjected to histogram equalization. For the image data analysis in the target window, the above steps S2 and S3 are repeated to obtain the image clarity under different contrast limit parameters, and the images are sorted from high to low. The highest clarity is selected to enhance the target window to improve the image enhancement effect of the target window.

[0116] See also Figure 3 , which shows an enhanced farmland remote sensing image of an intelligent processing method for UAV remote sensing mapping images provided by an embodiment of the present invention.

[0117] It can be explained that in step S5, the target window is moved in the target image, and steps S2 to S4 are repeated until the target image is enhanced; specifically, the target window is moved, and the step length of each movement is the same as the side length of the divided equilateral window, and steps S2 to S4 are repeated, wherein, for the enhancement result of the overlapping area, the mean grayscale value of the pixel points after the target window enhancement is taken until the enhancement of the target image is completed.

[0118] It can be understood that the present application defines a histogram equalization window, calculates the clarity within the window, performs feature analysis first, calculates the regularity of the pixel sequence, and then calculates the correlation of each edge pixel sequence set, and screens out the pixel sequence that meets the edge line of the farmland area. The length of these pixel sequences and the number of sequences are used to calculate the integrity of the image in the window, and combined with the regularity, the clarity of the image is finally calculated, and the image clarity under different contrast limit parameters is obtained respectively. The enhancement result with the highest clarity is taken as the result of the window, thereby achieving enhancement of the entire image. Intelligent adjustments can be made according to the different blurred areas caused by different concentrations of clouds and fog in the image to achieve a higher enhancement effect, thereby ensuring the accuracy of image quality for analyzing the growth status of crops, realizing real-time monitoring of crop growth, and providing strong technical support for agricultural production.

[0119] As an optional implementation method, based on the application of this technical solution in real life, first, multi-temporal remote sensing image data is obtained, and the images are geometrically and radiometrically corrected to eliminate the problems of spatial alignment and brightness differences caused by factors such as shooting angles, sensor differences and atmospheric conditions, so as to ensure the consistency of image data in space and brightness; then, this technical solution is used to enhance the impact and remove the impact of clouds and mist; then, the vegetation index is calculated and time series data is constructed to quantify vegetation coverage and health status, track the changes in vegetation over time, and facilitate the understanding of the growth cycle and health status of vegetation; in the change detection stage, statistical methods are used to analyze the time trend of the vegetation index and identify growth cycles and abnormal changes; by dividing the growth stages and establishing a time series model to predict future growth, the results are finally interpreted to provide a scientific basis for agricultural management and land use planning.

[0120] 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.

[0121] 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. An intelligent processing method for UAV remote sensing imaging, characterized in that: The method comprises: Step S1: using a drone to collect remote sensing images and preprocessing to generate a target image to be analyzed; Step S2: generating a target window based on the target image, obtaining edge lines in the target window, obtaining an edge pixel sequence, performing a primary screening on the edge pixel sequence to generate a first set; performing a secondary screening on the first set to generate a second set; Step S3: calculating the correlation based on the second set to determine the integrity of the edge line in the target window, and calculating the clarity of the image in the target window by averaging the regularity; Step S4: similarly calculating the clarity of the image in the target window according to different contrast limit parameters, and enhancing the target window by analyzing different clarity; Step S5: moving the target window in the target image, and repeating steps S2 to S4 until the target image is enhanced; Step S2 includes: Step S21: dividing the target image into a number of equilateral windows, wherein four adjacent equilateral windows of any equilateral window form a large window, defining any large window as a target window, performing edge detection on the image in the target window to obtain edge lines, and obtaining edge pixel sequences; Step S22: performing feature analysis on the edge pixel sequence and calculating the regularity, and performing a screening based on the regularity to generate a first set; Step S23: define any edge pixel sequence in the first set as a target sequence, calculate the probability that the target sequence is parallel to any non-target sequence in the first set, and obtain the probability that the two are collinear, and similarly obtain the probability that the target sequence is collinear with all remaining non-target sequences, and generate the second set after secondary screening; Step S22 includes: All edge pixel sequences in the target window are split into multiple pixel sequences according to the branch points, the original chain code is generated based on the pixel sequence, and the original chain code is split into two segments according to the absolute value of the first-order differential chain code in any original chain code; Calculate the length of the pixel sequence and determine whether the pixel sequence is real information. If so, generate the original set; Calculate the regularity of all pixel sequences in the original set, and screen the pixel sequences based on the regularity to generate a first set; The corresponding calculation formula for regularity is: The original chain code representation that defines any pixel sequence in the original set is: in, Represents a pixel sequence Regularity of Represents a pixel sequence Any two adjacent items of the original chain code; represents an exponential function with a natural constant as base, Represents a pixel sequence The total number of original chain codes, express The total quantity.

2. The method for intelligent processing of UAV remote sensing imaging images as claimed in claim 1, characterized in that: In step S23, the probability that the target sequence is parallel to any non-target sequence in the first set is calculated, including: Define any edge pixel sequence in the first set as the target sequence, obtain any non-target sequence in the first set, connect the first and last pixel points of the target sequence and the non-target sequence to obtain pixel line segment 1 and pixel line segment 2, obtain the distance between all pixel points of pixel line segment 2 and the extended pixel line segment 1, calculate the variance based on all distances, and calculate the probability that the target sequence and the non-target sequence are parallel. The corresponding calculation formula is: in, Represents the target sequence in the first set and non-target sequences Parallel probability; represents variance; Represents an exponential function with a natural constant as its base.

3. The intelligent processing method for UAV remote sensing imaging as claimed in claim 2, characterized in that: In step S23, the probability of the two being collinear is obtained, and the second set is generated through secondary screening. The corresponding calculation formula is: in, Represents the target sequence in the first set and non-target sequences Probability of collinearity; Represents the pixel line segment on the second The pixel point and the extended pixel line segment are one; Represents the total distance between all pixel points of pixel line segment 2 and the extended pixel line segment 1; Represents the target sequence in the first set and non-target sequences Parallel probability; represents an exponential function with a natural constant as base; represents the maximum and minimum normalized function; Similarly, the probability that the target sequence is collinear with all non-target sequences is obtained, and secondary screening is performed based on the probability of collinearity to generate the second set.

4. The method for intelligent processing of UAV remote sensing imaging images as claimed in claim 3, characterized in that: In step S3, the correlation is calculated as: The second set includes several small sets, and any small set is defined as the target set. The second set is , and the rectangular coordinate representation corresponding to the angular characteristic direction of the second set is obtained as , calculate the correlation as: in, Represents the target set relevance; represents the total number of small sets in the second set; Represents the target set With any non-target set The probability that the angle characteristics satisfy parallelism; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

5. The method for intelligent processing of UAV remote sensing imaging images as claimed in claim 4, characterized in that: In step S3, the integrity of the edge line in the target window is determined, including: Based on the correlation, the correlation between the target set and all the remaining non-target sets is calculated in the same way. The correlation is used to determine whether all the small sets in the second set are real information. If so, three screenings are performed to generate the third set, where the third set is , the number of pixel sequences contained in each small set is , and the total length of each pixel sequence is obtained as , the calculation formula corresponding to the integrity of the edge line in the target window is: in, Indicates the target window The integrity of the inner edge line; Represents the third set Small and medium collection The number of pixel sequences included; Represents the third set The total number of small and medium sets; Represents the third set Small and medium collection The total length of the pixel sequence contained; represents an exponential function with a natural constant as base; Represents the maximum and minimum normalization function.

6. The method for intelligent processing of UAV remote sensing imaging images as claimed in claim 5, characterized in that: The clarity of the image in the target window is calculated by averaging the regularity, including: Get the average regularity of the pixel sequence contained in each small set in the third set as , calculate the clarity of the image in the target window, the corresponding calculation formula is: in, Indicates the target window The clarity of the internal image; Indicates the target window The integrity of the inner edge line; Represents the third set Small and medium collection The average regularity of the contained pixel sequences; Represents the third set The total number of small and medium sets; Represents the maximum and minimum normalization function.

7. The method for intelligent processing of UAV remote sensing imaging images as claimed in claim 1, characterized in that: Specifically, in step S4, the value range of the contrast limit parameter is [2, 20], the clarity of the image in the target window is calculated based on the contrast limit parameters traversed in step S2 and step S3, and the images are sorted from high to low, and the target window is enhanced with the highest clarity.

Images