Unmanned aerial vehicle monitoring method based on cloud control platform

Abstract

The present application relates to the technical field of unmanned aerial vehicle, and particularly relates to a kind of unmanned aerial vehicle monitoring method based on cloud control platform, comprising: obtaining the first clipping image of the flight area where the unmanned aerial vehicle flight route is located in remote sensing image;The first clipping image is grid divided, and the grid area image of multiple grid areas is obtained;Obtain the ground feature complexity index of each grid area image;Based on the ground feature complexity index of each grid area image, obtain the initial speed of the unmanned aerial vehicle when flying in each grid area;Wherein, initial speed is inversely proportional to ground feature complexity index;Control unmanned aerial vehicle flies along unmanned aerial vehicle flight route to obtain aerial image.The present application can effectively avoid the image blur or shaking problem caused by too fast or too slow speed by actively adjusting the flight speed of unmanned aerial vehicle, ensure that the image collected by unmanned aerial vehicle has higher stability and clarity, help to improve the defect recognition rate and monitoring imaging effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicles, and particularly relates to an unmanned aerial vehicle monitoring method based on a cloud control platform. BACKGROUND

[0002] With the development of unmanned aerial vehicle technology, unmanned aerial vehicles can obtain high-quality images through aerial photography, which can be used in industrial inspection, urban management and emergency rescue scenes, and the quality of images taken by unmanned aerial vehicles is largely dependent on the stability of flight speed.

[0003] At present, the unmanned aerial vehicle speed adjustment method based on sensor feedback and artificial intelligence (AI) control in the related technology is mostly to passively adjust the speed of the unmanned aerial vehicle after finding image blur or after shaking occurs, which directly affects the quality of images taken by the unmanned aerial vehicle. Therefore, the unmanned aerial vehicle adjustment method in the related technology has the problem of response lag. SUMMARY

[0004] The present application provides an unmanned aerial vehicle monitoring method based on a cloud control platform to solve the existing problems.

[0005] The unmanned aerial vehicle monitoring method based on a cloud control platform provided by the present application adopts the following technical scheme:

[0006] One embodiment of the present application provides an unmanned aerial vehicle monitoring method based on a cloud control platform, which comprises:

[0007] acquiring a first cropped image of a flight area where an unmanned aerial vehicle flight route is located in a remote sensing image;

[0008] dividing the first cropped image into grids to acquire grid area images of multiple grid areas;

[0009] acquiring a ground feature complexity index of each grid area image; wherein the ground feature complexity index is used to represent the texture richness of the grid area image;

[0010] acquiring an initial speed of the unmanned aerial vehicle when flying in each grid area based on the ground feature complexity index of each grid area image; wherein the initial speed is inversely proportional to the ground feature complexity index;

[0011] controlling the unmanned aerial vehicle to fly along the unmanned aerial vehicle flight route to acquire aerial images.

[0012] Further, acquiring an initial speed of the unmanned aerial vehicle when flying in each grid area based on the ground feature complexity index of each grid area image comprises:

[0013] performing threshold segmentation on the ground feature complexity index of multiple grid area images to acquire multiple segmentation sections;

[0014] obtaining a reference segmentation section; wherein the reference segmentation section is the segmentation section corresponding to the grid area image with the largest number among all the segmentation sections;

[0015] obtaining an initial speed of the UAV when flying in the grid area corresponding to the reference segmentation section;

[0016] obtaining the initial speed of the UAV when flying in the grid area corresponding to other segmentation sections based on the ratio between the representative value of the feature complexity index of other segmentation sections and the representative value of the feature complexity index of the reference segmentation section.

[0017] Further, the grid division of the first cropped image to obtain the grid area images of multiple grid areas comprises:

[0018] performing superpixel segmentation on the first cropped image to obtain multiple superpixel blocks;

[0019] obtaining the circumscribed rectangle of the superpixel block with the smallest area among the multiple superpixel blocks;

[0020] taking the size of the circumscribed rectangle as a standard grid size, performing grid division on the first cropped image to obtain the grid area images of multiple grid areas.

[0021] Further, the method further comprises:

[0022] obtaining the coordinate position of the UAV when controlling the UAV to fly along the UAV flight route;

[0023] obtaining the current grid area where the UAV is located based on the coordinate position;

[0024] stitching the aerial image of the UAV in the current grid area to obtain a stitched image;

[0025] obtaining a second cropped image corresponding to the stitched image in the remote sensing image;

[0026] resampling the stitched image to obtain a resampled image with the same resolution as the second cropped image;

[0027] obtaining a difference index of the resampled image and the second cropped image; wherein the difference index is used to represent the difference between the resampled image and the second cropped image;

[0028] when the difference index is greater than a preset difference index threshold, correcting the initial speed corresponding to the current grid area to obtain a first corrected speed;

[0029] controlling the UAV to fly along the UAV flight route in the current grid region at the first modified speed.

[0030] Further, the method further comprises:

[0031] obtaining a ground feature complexity index of the spliced image and a ground feature complexity index of the second cropped image;

[0032] modifying the initial speed corresponding to the current grid region based on a ratio of the ground feature complexity index of the spliced image to the ground feature complexity index of the second cropped image to obtain a first modified speed.

[0033] Further, the method further comprises:

[0034] before the UAV enters a next grid region, predicting the ground feature complexity index of the next grid region based on historical remote sensing images to obtain an adjustment probability that a ground feature complexity prediction index is greater than the ground feature complexity index of the next grid region;

[0035] modifying the initial speed corresponding to the next grid region based on the adjustment probability to obtain a second modified speed;

[0036] after the UAV enters the next grid region, controlling the UAV to fly along the UAV flight route in the next grid region at the second modified speed.

[0037] Further, the method further comprises:

[0038] obtaining a third cropped image of a flight region where the UAV flight route is located in historical remote sensing images of different historical periods;

[0039] dividing the third cropped image into a plurality of historical grid regions by using a standard grid size used when dividing the first cropped image into grid regions to obtain historical grid region images of the plurality of historical grid regions;

[0040] obtaining historical ground feature complexity indexes of the historical grid region images;

[0041] performing threshold segmentation on the ground feature complexity indexes of the plurality of grid region images to obtain a plurality of segmented sections;

[0042] The historical feature complexity indexes of a plurality of the historical grid area images are threshold segmented to obtain a plurality of historical segmentation sections;

[0043] The historical segmentation section corresponding to each of the historical grid area images is obtained.

[0044] The historical segmentation sections corresponding to each of the historical grid area images of all the historical periods are counted to obtain a probability table; wherein the probability table comprises probabilities of the historical grid area adjacent to the historical grid area corresponding to different historical segmentation sections when the historical grid area is in different historical segmentation sections.

[0045] Based on the relative position between the next grid area and the current grid area and the feature complexity index of the current grid area, probabilities of the next grid area corresponding to different historical segmentation sections are obtained in the probability table.

[0046] An adjustment probability is obtained; wherein the adjustment probability is used to represent a sum of probabilities of all historical segmentation sections whose historical feature complexity index representative value of the historical segmentation section corresponding to the next grid area is greater than the feature complexity index representative value of the next grid area.

[0047] Further, the initial speed corresponding to the next grid area is corrected based on the adjustment probability to obtain a second correction speed, comprising:

[0048] Based on the adjustment probability, a correction factor is obtained; wherein the correction factor is negatively correlated with the adjustment probability.

[0049] Based on the correction factor, the initial speed corresponding to the next grid area is corrected to obtain a second correction speed.

[0050] Further, the feature complexity index of each of the grid area images is obtained, comprising:

[0051] An entropy value of a gray level co-occurrence matrix of each of the grid area images is obtained; wherein the entropy value is positively correlated with the feature complexity.

[0052] Further, the first clipping image of the flight area where the unmanned aerial vehicle flight route is located in the remote sensing image is obtained, comprising:

[0053] An outer rectangle of the unmanned aerial vehicle flight route on the remote sensing image is obtained.

[0054] The area corresponding to the outer rectangle is taken as the first clipping image of the flight area where the unmanned aerial vehicle flight route is located.

[0055] The technical scheme of the present application has the following advantages:

[0056] In the embodiment of the present application, a first cropped image of a flight area where a flight route of a UAV is located is acquired in a remote sensing image; the first cropped image is grid divided to acquire grid area images of a plurality of grid areas; a ground feature complexity index of each grid area image is acquired; an initial speed of the UAV when flying in each grid area is acquired based on the ground feature complexity index of each grid area image; wherein the initial speed is inversely proportional to the ground feature complexity index; the UAV is controlled to fly along the flight route of the UAV to acquire aerial images. Thus, the present application can effectively avoid the image blurring or shaking problem caused by too fast or too slow speed by actively adjusting the flight speed of the UAV, ensuring that the images collected by the UAV have higher stability and clarity, which helps to improve the defect recognition rate and monitoring imaging effect. BRIEF DESCRIPTION OF DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0058] Figure 1 The flowchart of the UAV monitoring method based on the cloud control platform provided by the embodiment of the present application is shown in the figure;

[0059] Figure 2 The relationship between the flight route of the UAV, the remote sensing image and the first cropped image provided by the embodiment of the present application is shown in the figure;

[0060] Figure 3 The flight area after grid division of the flight area of the UAV provided by the embodiment of the present application is shown in the figure;

[0061] Figure 4 The flight area historical remote sensing image from January to April of a certain year provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0062] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined object, the specific implementation, structure, features and effects of the UAV monitoring method based on the cloud control platform according to the present application are described in detail as follows by combining with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0063] Unless otherwise defined, 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 application belongs.

[0064] The specific scheme of the unmanned aerial vehicle monitoring method based on the cloud control platform provided by the application will be described in detail below with reference to the drawings.

[0065] Please refer to Figure 1 which shows a kind of unmanned aerial vehicle monitoring method based on cloud control platform provided by an embodiment of the application, the method comprises:

[0066] Step S110: obtain the first cropped image of the flight area where the unmanned aerial vehicle flight route is located in the remote sensing image.

[0067] The remote sensing image refers to the image data of the earth's surface obtained by satellite or other remote sensing technology, for example, satellite images collected by Gaofen-1 satellite. The satellite image can meet the data requirement of pixel range from 50 meters to 2 meters, and can also provide high-resolution and large-range high-precision remote sensing data, which can be applied to various application scenarios such as mineral resource investigation and monitoring, urban planning, geological disaster monitoring, land cover analysis, land use change analysis, and land surface temperature analysis. The remote sensing image can be used to analyze the geographical density and ground object complexity of the flight area where the unmanned aerial vehicle flight route is located. Step S110 can select the latest remote sensing image at the current time to obtain the first cropped image, thereby reducing the deviation between the flight area ground object complexity index calculation value and the actual ground object complexity index value.

[0068] For example, the unmanned aerial vehicle flight route refers to the pre-obtained unmanned aerial vehicle flight route, which can be a flight route planned manually by an operator, or a flight route automatically generated by using a flight planning algorithm. The embodiments of the present application do not limit the acquisition method of the unmanned aerial vehicle flight route.

[0069] The cloud control platform refers to the cloud control platform of the unmanned aerial vehicle, which is a comprehensive management platform designed to realize the automatic management of unmanned aerial vehicle flight tasks.

[0070] Preferably, in an embodiment of the present application, step S110 can comprise: obtaining the circumscribed rectangle of the unmanned aerial vehicle flight route on the remote sensing image; and taking the region corresponding to the circumscribed rectangle as the first cropped image of the flight area where the unmanned aerial vehicle flight route is located. For example, please refer to Figure 2 First, the flight route of the unmanned aerial vehicle flight route on the remote sensing image is obtained by coordinate conversion and other methods. Then, the circumscribed rectangle of the unmanned aerial vehicle flight route is obtained, and the remote sensing image of the circumscribed rectangle region is determined as the first cropped image.

[0071] Step S120: grid division is performed on the first cropped image to obtain grid region images of a plurality of grid regions.

[0072] Preferably, in one embodiment of the present application, the step S120 comprises: performing superpixel segmentation on the first cropped image to obtain a plurality of superpixel blocks; obtaining a circumscribed rectangle of a superpixel block with the smallest area among the plurality of superpixel blocks; and performing grid division on the first cropped image with the size of the circumscribed rectangle as a standard grid size to obtain the grid region images of the plurality of grid regions.

[0073] The superpixel segmentation is an image processing technique for segmenting the first cropped image into a plurality of regions with similar textures or features, also known as superpixel blocks. The superpixel segmentation can decompose an image into a series of relatively uniform regions, each of which has pixels with similar color, texture, or brightness, thereby providing a more concise representation for subsequent image analysis. In the embodiments of the present application, the functions of the superpixel segmentation include: (1) image simplification: dividing the first cropped image into a plurality of superpixel blocks to simplify the image representation and reduce the computational complexity of subsequent processing; (2) texture analysis: using the uniformity of superpixel blocks to facilitate the calculation of the feature complexity index of each grid region, which is used to estimate the complexity of the geographic environment; (3) providing a basis for grid division: determining the standard grid size through superpixel segmentation as a basis for subsequent grid division, which can realize the similarity of geographic scenery in the same grid.

[0074] For example, the superpixel segmentation can be implemented by using the SLIC (Simple Linear Iterative Clustering) algorithm. SLIC is a clustering-based superpixel segmentation method that can quickly segment an image into superpixel blocks with approximately equal size and clear boundaries. The grid regions obtained by the SLIC algorithm are shown in FIG. 2. Figure 3

[0075] ​The minimum area superpixel block obtained by the superpixel segmentation determines the grid size, on the one hand, the superpixel segmentation can divide the image into a plurality of superpixel blocks according to the image content, the pixels in each superpixel block have similarity in color, texture and the like, by selecting the minimum area superpixel block as the standard grid size, the image content can be flexibly adapted, so that the grid division is more in line with the actual distribution of the ground object, and the rationality of the initial speed distribution is improved; on the other hand, the standard grid size determined by the superpixel segmentation makes the texture and color feature difference between adjacent grid regions smaller, which is convenient for subsequent ground object complexity analysis and flight speed control; on the other hand, the superpixel segmentation divides the image into a plurality of superpixel blocks, reduces the data amount of subsequent processing, and the grid division based on the superpixel block can reduce the calculation complexity of the grid division and subsequent processing steps, and improve the overall processing efficiency of the above-mentioned unmanned aerial vehicle monitoring method based on the cloud control platform.

[0076] Step S130: obtaining a ground object complexity index of each grid region image; wherein the ground object complexity index is used to represent the texture richness of the grid region image.

[0077] Preferably, the step S130 comprises: obtaining an entropy value of a gray level co-occurrence matrix of each grid region image; wherein the entropy value is proportional to the ground object complexity.

[0078] The above scheme takes the gray level co-occurrence matrix entropy value of the image as the ground object complexity index, the gray level co-occurrence matrix (GLCM) is a tool for analyzing the texture features of the image. It describes the texture characteristics of the image by statistically analyzing the spatial relationship of the pixel gray values in the image. Specifically, the gray level co-occurrence matrix records the occurrence probability of pixel pairs with specific gray values in a specific direction and distance. The entropy value of the gray level co-occurrence matrix is mainly used to describe the complexity of image information, and the greater the entropy value, the richer the texture of the image, that is, the higher the ground object complexity.

[0079] The above scheme takes the gray level co-occurrence matrix entropy value of the grid region image as the ground object complexity index, on the one hand, the entropy value of the gray level co-occurrence matrix is sensitive to the complexity of the image texture, and can finely reflect the texture change of the ground object, thereby providing more accurate basis for the unmanned aerial vehicle speed control and improving the rationality of the initial speed distribution; on the other hand, the calculation method of the gray level co-occurrence matrix entropy value is simple and efficient, and the gray level co-occurrence matrix entropy value can be calculated in a short time, thereby improving the response speed of the above-mentioned unmanned aerial vehicle monitoring method based on the cloud control platform.

[0080] It can be understood that, in addition to the entropy value of the gray level co-occurrence matrix, other indicators can also be used to represent the complexity of the features in the region, or the richness of the features in the region (i.e. the complexity, randomness, variability or information amount of the texture, etc.), for example:

[0081] (1) Other indicators based on the gray level co-occurrence matrix, including:

[0082] Contrast, mainly used to reflect the size or definition of local gray level change in the image. The more points with large pixel value difference, the greater the contrast. Generally, the higher the feature complexity, the richer the texture, the more details the image contains, and the higher the contrast, that is, the contrast is positively correlated with the feature complexity.

[0083] Energy, mainly used to reflect the uniformity of the image gray level distribution or the degree of repetition of the pixel gray level value. The higher the energy value, the more uniform the texture in the image and the smaller the change; the lower the energy, the greater the texture change and the more dispersed the gray level distribution. That is, the energy value of the image is negatively correlated with the feature complexity. In addition, energy can be used as a complementary indicator of the above-mentioned entropy value.

[0084] Homogeneity, mainly used to measure the uniformity of the local region of the image. The higher the homogeneity, the more uniform the local region and the more gentle the gray level change; the lower the homogeneity, the more intense the local gray level change and the more rough or complex the texture. That is, the homogeneity value is negatively correlated with the feature complexity. In addition, homogeneity can also be used as a complementary indicator of the above-mentioned entropy value.

[0085] (2) Indicators based on image statistical histogram, including:

[0086] Gray-Level Variance, the greater the variance, the higher the dispersion of the pixel gray level value relative to the mean value, the more intense the gray level change, and generally means that the texture is richer and the details are more.

[0087] Gray-Level Standard Deviation, the greater the gray level standard deviation, generally means that the texture is richer and the details are more.

[0088] For example, the above-mentioned multiple indicators can be combined for use, for example, the sum or weighted sum of the entropy value of the gray level co-occurrence matrix, the image contrast and the energy value is used as the feature complexity indicator of the image. Of course, the above-mentioned multiple indicators can also be used alone, for example, the above-mentioned scheme mentions using the entropy value of the gray level co-occurrence matrix as the feature complexity indicator of the image.

[0089] Step S140: obtaining an initial speed of the UAV when flying in each grid region based on the ground feature complexity index of the image of each grid region; wherein the initial speed is inversely proportional to the ground feature complexity index.

[0090] Preferably, in an embodiment of the present application, the step S140 comprises: threshold segmentation is performed on the ground feature complexity indexes of the plurality of grid region images to obtain a plurality of segmentation sections; a reference segmentation section is obtained; wherein the reference segmentation section is the segmentation section with the largest number of grid region images; an initial speed of the UAV when flying in the grid region corresponding to the reference segmentation section is obtained; and an initial speed of the UAV when flying in the grid region corresponding to other segmentation sections is obtained based on the ratio between the representative value of the ground feature complexity index of the other segmentation sections and the representative value of the ground feature complexity index of the reference segmentation section. For example, when the ground feature complexity index is the gray level co-occurrence matrix entropy value, the entropy values are arranged in ascending order to obtain an ascending sequence, and the Otsu multi-threshold segmentation method is used to obtain a plurality of segmentation sections from the ascending sequence. The entropy values in the same segmentation section are close to each other, i.e., the ground feature complexity of the grid region corresponding to different entropy values in the same segmentation section is close. The number of elements corresponding to each segmentation section, i.e., the number of grid regions, is counted, and the segmentation section with the largest number of corresponding elements is taken as the reference segmentation section. An initial speed can be set for the reference segmentation section according to the experience of experts in the field. The initial speed of the other segmentation sections is obtained by the ratio between the representative value of the entropy value of the reference segmentation section and the representative value of the entropy value of the other segmentation sections, and the calculation method can be:

[0091] The initial speed of the other segmentation sections = the representative value of the entropy value of the reference segmentation section / the representative value of the entropy value of the other segmentation sections * the initial speed of the reference segmentation section.

[0092] It can be understood that the greater the representative value of the entropy value of the other segmentation section, the richer the texture of the grid region image corresponding to the segmentation section, and the lower the flight speed of the UAV needs to be set, and therefore, the lower the initial speed allocated to the segmentation section.

[0093] It should be noted that the Otsu multi-threshold segmentation described above is a relatively mature threshold segmentation method in the field, and its specific implementation paradigm can be referred to related technologies, and will not be described herein.

[0094] For example, the representative value of the entropy value of the segmentation section can be the median value of the range of the entropy value corresponding to the segmentation section.

[0095] The above scheme can more finely divide different ground feature complexity regions by threshold segmentation of the ground feature complexity indicators of the grid region image, and based on the ground feature complexity characteristics of each segmentation segment, a more accurate initial speed is assigned to the unmanned aerial vehicle, so as to reduce the speed in a complex region to obtain a clear image, and to increase the speed in a simple region to save time, thereby improving the unmanned aerial vehicle speed control effect of the above cloud control platform-based unmanned aerial vehicle monitoring method. On the other hand, the determination of the reference segmentation segment is based on the principle of the largest number of grid region images, so that the initial speed as a reference value is reasonable and suitable for most regions, and the speed of other segmentation segments is calculated according to the ratio of the ground feature complexity indicators of the reference segmentation segment, so as to realize reasonable allocation of flight speed and improve the overall flight efficiency of the unmanned aerial vehicle.

[0096] Step S150: controlling the unmanned aerial vehicle to fly along the unmanned aerial vehicle flight route to obtain aerial images.

[0097] For example, the coordinate position of the unmanned aerial vehicle can be obtained in real time by the GPS sensor carried by the unmanned aerial vehicle to determine the grid region where the unmanned aerial vehicle is located. After the unmanned aerial vehicle enters the current grid region, the unmanned aerial vehicle is controlled to fly in the current grid region at the initial speed corresponding to the current grid region to obtain aerial images of the current grid region.

[0098] It can be understood that the initial speed assigned to the unmanned aerial vehicle by the above scheme is not only the flight speed at the moment when the unmanned aerial vehicle enters the corresponding grid region, but also the flight speed after entering the corresponding grid region. In other words, before the initial speed is corrected by the speed correction means (for example, the two speed correction means provided in the subsequent content), the unmanned aerial vehicle will always fly in the corresponding grid region at the initial speed. After the initial speed is corrected by the speed correction means and the corrected speed is obtained, the unmanned aerial vehicle will continue to fly in the current grid region at the corrected speed.

[0099] The above scheme sets the initial speed of the unmanned aerial vehicle when flying in different grid regions based on the ground feature complexity indicators of the flight region. On the one hand, compared with the passive adjustment of the unmanned aerial vehicle speed adjustment mode, the above scheme can pre-set the initial flight speed of the unmanned aerial vehicle when flying in the corresponding region based on the ground feature complexity, which is conducive to improving the overall quality of the aerial images. On the other hand, regions with high ground feature complexity usually contain more details and textures. When the unmanned aerial vehicle flies at a lower speed for aerial photography, it can capture clearer images. By assigning a lower initial speed to these regions, image blurring and detail loss caused by high speed can be reduced, which is conducive to further improving the quality of the aerial images. On the other hand, in regions with low ground feature complexity, the unmanned aerial vehicle can appropriately increase the flight speed, thereby ensuring image quality while more efficiently completing the photography task in the corresponding region and improving overall work efficiency.

[0100] The initial speed is obtained based on the latest remote sensing image. Since the remote sensing image has a certain timeliness, it is usually updated once in a long time, and the actual geographical situation of the UAV flight area may have changed. For example, a piece of farmland may have planted new crops after the remote sensing image is obtained, resulting in an increase in the complexity of the ground objects. Based on this, the embodiments of the present application provide the following scheme:

[0101] Preferably, in an embodiment of the present application, the UAV monitoring method based on the cloud control platform further comprises: obtaining the coordinate position of the UAV when controlling the UAV to fly along the UAV flight route; obtaining the current grid area where the UAV is located based on the coordinate position; splicing the aerial image of the UAV in the current grid area to obtain a spliced image; obtaining a second cropped image corresponding to the spliced image in the remote sensing image; resampling the spliced image to obtain a resampled image having the same resolution as the second cropped image; obtaining a difference index of the resampled image and the second cropped image; wherein the difference index is used to represent the difference degree between the resampled image and the second cropped image; when the difference index is greater than a preset difference index threshold, correcting the initial speed corresponding to the current grid area to obtain a first corrected speed; controlling the UAV to fly along the UAV flight route in the current grid area at the first corrected speed. For example, the current coordinate position of the UAV is obtained by the GPS sensor carried on the UAV, and then the current grid area where the UAV is located can be obtained; the aerial image of the current grid area obtained by the UAV is spliced to obtain a spliced image; the spliced image is used as a mask to obtain a second cropped image corresponding to the spliced image in the remote sensing image; the spliced image is resampled by the mean value to obtain a resampled image having the same resolution as the second cropped image; the MSE (Mean Squared Error) value of the resampled image and the second cropped image is obtained and normalized; if the normalized MSE value is greater than 0.3, it means that the actual ground objects in the current grid area where the UAV is located are significantly different from the ground objects in the remote sensing image, and the initial speed needs to be corrected in combination with the ground object change.

[0102] The gray level co-occurrence matrix entropy value of the spliced image is calculated, denoted as a;

[0103] The gray level co-occurrence matrix entropy value of the second cropped image (the gray level co-occurrence matrix entropy value of the second cropped image is the gray level co-occurrence matrix entropy value of the current grid area) is calculated, denoted as b;

[0104] If a≤b, the initial speed is not changed, and the UAV continues to fly in the current grid area at the initial speed corresponding to the current grid area; if a>b, it indicates that the ground object in the current grid area has a large change, and the complexity of the current ground object exceeds the complexity of the ground object on the remote sensing image, so the flight speed of the UAV can be appropriately reduced on the basis of the initial speed allocated in advance for the current grid area.

[0105] The mean resampling is an image processing technology for adjusting images of different resolutions to the same resolution. In addition to the mean resampling, the scheme can also use a nearest neighbor interpolation method, a bilinear interpolation method, or the like to resample the spliced image.

[0106] For example, the difference index of the resampled image and the second cropped image can use a mean square error, a root mean square error, a square absolute error, a correlation coefficient, or the like, and can measure the difference between the resampled image and the second cropped image. The calculation method of the difference index is described in related technologies, and will not be described herein.

[0107] The scheme can timely find the difference between the actual aerial image and the remote sensing image by comparing the difference index of the spliced image and the remote sensing image. If the difference index exceeds a threshold, it indicates that the ground object in the current grid area can have changed. At this time, the flight speed of the UAV is adjusted by the difference index to adapt to the complexity of the ground object in the current grid area, which is beneficial to improve the control effect of the UAV monitoring method based on the cloud control platform.

[0108] The following describes several optional embodiments of the first modified speed acquisition scheme:

[0109] In a first embodiment, the initial speed is modified by using the ratio of the ground object complexity index of the spliced image to the ground object complexity index of the second cropped image to obtain the first modified speed.

[0110] Preferably, in an embodiment of the present application, the initial speed corresponding to the current grid area is modified to obtain the first modified speed, including: obtaining the ground object complexity index of the spliced image and the ground object complexity index of the second cropped image; modifying the initial speed corresponding to the current grid area based on the ratio of the ground object complexity index of the spliced image to the ground object complexity index of the second cropped image to obtain the first modified speed.

[0111] Taking the ground object complexity index as an example of the entropy value of the gray level co-occurrence matrix, the entropy value of the gray level co-occurrence matrix of the spliced image is a, and the entropy value of the gray level co-occurrence matrix of the second cropped image is b. The calculation method of the first modified speed can be:

[0112]

[0113] wherein, is the first correction speed corresponding to the current grid region after correction, is the initial speed of the current grid region. The first correction speed is inversely proportional to the gray level co-occurrence matrix entropy value a of the spliced image, that is, the greater the gray level co-occurrence matrix entropy value a, the smaller the first correction speed , which also conforms to the speed adjustment principle that the more complex the ground object, the lower the required unmanned aerial vehicle flight speed.

[0114] The above scheme directly corrects the initial speed using the ratio of the ground object complexity indicators of the spliced image and the second cropped image. The ratio can directly reflect the complexity difference between the actual ground object and the preset remote sensing image ground object, can more accurately judge the deviation of the actual ground object complexity from the preset value, and is beneficial to improving the control effect of the above-mentioned unmanned aerial vehicle monitoring method based on the cloud control platform.

[0115] Second implementation: correction based on the difference between the ground object complexity indicators of the resampled image and the second cropped image, specifically:

[0116] A plurality of threshold ranges of the difference between the ground object complexity indicators of the resampled image and the second cropped image can be set in advance, and corresponding adjustment coefficients are respectively assigned to the plurality of threshold ranges. After obtaining the difference between the ground object complexity indicators of the resampled image and the second cropped image, the difference value threshold range in which the difference value is located can be directly determined, so as to obtain the corresponding adjustment coefficient. Then, the product of the adjustment coefficient and the initial speed can obtain the first correction speed.

[0117] Third implementation: correction using a neural network model;

[0118] A neural network model can be pre-trained to learn the relationship between the complexity of ground objects and the flight speed of the unmanned aerial vehicle. When the difference indicator of the resampled image and the second cropped image is greater than a preset difference indicator threshold, the flight speed suitable for the complexity of the ground object of the current grid region is directly obtained using the neural network model, so as to correct the initial speed.

[0119] It can be understood that the above-mentioned three implementations can be combined and optimized according to actual application scenarios and needs, so that the flight speed of the unmanned aerial vehicle in different environments can be dynamically adjusted, thereby obtaining high-quality aerial images.

[0120] During the flight of the unmanned aerial vehicle, due to the dynamic changes of the environment and the timeliness of the remote sensing image, the actual ground object complexity may not be consistent with the predicted ground object complexity of the remote sensing image. Based on this, the embodiments of the present application provide the following scheme:

[0121] Preferably, in one embodiment of the present application, the unmanned aerial vehicle monitoring method based on the cloud control platform further comprises: before the unmanned aerial vehicle enters the next grid area, predicting the ground feature complexity index of the next grid area based on historical remote sensing images, obtaining an adjustment probability that the ground feature complexity prediction index is greater than the ground feature complexity index of the next grid area; correcting the initial speed corresponding to the next grid area based on the adjustment probability to obtain a second corrected speed; after the unmanned aerial vehicle enters the next grid area, controlling the unmanned aerial vehicle to fly along the flight route of the unmanned aerial vehicle in the next grid area at the second corrected speed.

[0122] The adjustment probability can be the probability that the ground feature complexity index of the next grid area exceeds the remote sensing image prediction index, which is calculated based on historical remote sensing images and Markov prediction ideas. Specifically, the adjustment probability reflects the uncertainty of the possible increase in the ground feature complexity of the next grid area, and is used to determine whether the flight speed of the unmanned aerial vehicle needs to be adjusted in advance.

[0123] The above scheme predicts the ground feature complexity of the next grid area, so that the unmanned aerial vehicle can adjust the flight speed in advance before entering the area, thereby ensuring that the images taken are clearer and more stable, reducing blurring or information omission caused by excessive speed, and improving the aerial photography effect. On the other hand, by using historical remote sensing images and Markov prediction ideas, the unmanned aerial vehicle monitoring method based on the cloud control platform can better adapt to changes in the environment. Even if there is a difference between the actual ground features and the remote sensing images, the unmanned aerial vehicle monitoring method based on the cloud control platform can still ensure that the flight strategy of the unmanned aerial vehicle matches the actual environment through prediction and adjustment, thereby improving the adaptability and robustness of the unmanned aerial vehicle monitoring method based on the cloud control platform.

[0124] Preferably, in one embodiment of the present application, the above-mentioned prediction of the ground feature complexity index of the next grid area based on historical remote sensing images, obtaining the adjustment probability that the predicted ground feature complexity index is greater than the ground feature complexity index of the next grid area, comprises: obtaining a third cropped image of the flight area where the unmanned aerial vehicle flight route is located in the historical remote sensing images of multiple different historical periods; dividing the third cropped image into a plurality of historical grid area images by using the standard grid size used for dividing the first cropped image into a plurality of grid areas; obtaining the historical ground feature complexity index of each historical grid area image; threshold segmenting the ground feature complexity index of each historical grid area image to obtain a plurality of segmented sections; obtaining the historical ground feature complexity index of each historical grid area image; obtaining the historical ground feature complexity index of each historical grid area image; obtaining the historical ground feature complexity index of each historical grid area image; obtaining the probability table by statistically analyzing the historical segmented sections corresponding to each historical grid area image of all historical periods; wherein the probability table comprises the probability that the historical grid area adjacent to the historical grid area corresponding to the different historical segmented sections is in different historical segmented sections; obtaining the probability that the next grid area corresponding to the different historical segmented sections in the probability table based on the relative position between the next grid area and the current grid area and the ground feature complexity index of the current grid area; obtaining the adjustment probability; wherein the adjustment probability represents the sum of the probabilities of all historical segmented sections whose historical ground feature complexity index representative value is greater than the ground feature complexity index representative value of the next grid area. For example:

[0125] When the unmanned aerial vehicle is about to leave the current grid area and enter the next grid area, the ground feature complexity of the next grid area is predicted, and the specific process is as follows:

[0126] A plurality of historical remote sensing images of different periods are obtained by using Gaofen-1 (here, Gaofen-1 satellite is taken as an example, and relevant remote sensing data can be obtained according to actual needs); for example, all remote sensing images in the past three years;

[0127] A third cropped image of the flight area where the unmanned aerial vehicle flight route is located in the historical remote sensing images of multiple different historical periods is obtained; the third cropped image is divided into a plurality of historical grid area images by using the standard grid size used for dividing the first cropped image into a plurality of grid areas; the threshold segmenting of the gray level co-occurrence matrix entropy value of each historical grid area image is performed to obtain a plurality of historical segmented sections; and the historical segmented section corresponding to each historical grid area image is obtained.

[0128] The probability table is formed by the probability of each direction of the four directions of the grid in the different entropy value ranges when each grid region in all periods is in different entropy value ranges, and the entropy value range of the adjacent network can be obtained by the entropy value range of the target grid and the relative direction of the adjacent grid to the target grid.

[0129] Please refer to Figure 4 , Figure 4 Four remote sensing images of the UAV flight area from January to April of a certain year are shown, wherein the light-colored squares are grid regions, and the dark-colored squares are the current grid region of the UAV. The UAV flight route starts from the upper left of the route image and ends at the lower right of the image.

[0130] The different entropy value ranges of the current grid region can be obtained through the above-mentioned probability table, and the probability of the adjacent grid region (for example, the four adjacent grid regions in the up, down, left and right directions) of the current grid region being in different entropy value ranges when the current grid region is in each entropy value range can be obtained.

[0131] The next grid region that the UAV is about to enter can be obtained in combination with the UAV flight route. The aforementioned content obtains the gray level co-occurrence matrix entropy value a of the current grid region by calculation, and in combination with the a value and the relative position of the next grid region to the current grid region, the different probability entropy value ranges of the next grid region can be obtained.

[0132] A representative value is calculated for each entropy value range of the next grid region. The median value of each entropy value range (for example, when the entropy value range is [5, 10], 15 / 2 can be taken as the representative value of the entropy value range [5, 10]) is taken as the representative value, which is denoted as the predicted representative value. The median value of the entropy value range corresponding to the next grid region on the latest remote sensing image is taken as the representative value, which is denoted as the latest representative value.

[0133] The predicted representative values greater than the latest representative value are screened out, and the probabilities of the screened representative values are added to obtain an adjustment probability. If the adjustment probability is greater than 60%, it means that the complexity of the ground objects in the next grid region has a high probability of increasing, and the initial speed set for the grid region needs to be reduced to shoot high-quality aerial images.

[0134] By analyzing the historical remote sensing image data, a probability table is constructed, which can predict the complexity index of the next grid region in advance. According to the predicted complexity index of the ground objects, the flight speed of the UAV can be adjusted in advance to avoid the lag of the UAV speed adjustment caused by the sudden change of the complexity of the ground objects, which is conducive to improving the clarity and usability of the aerial images.

[0135] The above scheme can predict the ground object complexity index of the next grid area in advance by analyzing historical remote sensing image data and constructing a probability table. According to the predicted ground object complexity index, the flight speed of the unmanned aerial vehicle can be adjusted in advance to avoid the lag of the unmanned aerial vehicle speed adjustment caused by sudden changes in the ground object complexity, thereby improving the clarity and usability of the aerial image. On the other hand, by flexibly adjusting the flight speed of the unmanned aerial vehicle, the unmanned aerial vehicle monitoring method based on the cloud control platform can flexibly cope with changes in different scenes and ground object types, thereby improving the adaptability of the unmanned aerial vehicle monitoring method based on the cloud control platform.

[0136] Preferably, in an embodiment of the present application, the initial speed corresponding to the next grid area is corrected based on the adjustment probability to obtain a second correction speed, including: based on the adjustment probability, a correction factor is obtained; wherein the correction factor is negatively correlated with the adjustment probability; based on the correction factor, the initial speed corresponding to the next grid area is corrected to obtain the second correction speed. For example, in the above embodiment, the adjustment probability is calculated based on the initial speed corresponding to the next grid area and the ground object complexity index of the next grid area, and the correction factor is calculated based on the adjustment probability. The second correction speed is calculated based on the correction factor.

[0137]

[0138]

[0139] By introducing the correction factor and making it negatively correlated with the adjustment probability, the flight speed of the unmanned aerial vehicle is accurately and dynamically adjusted. When the adjustment probability is high, the correction factor is small, thereby reducing the flight speed and ensuring the image acquisition quality. When the adjustment probability is low, the correction factor is large, and the flight speed is relatively fast, thereby improving the flight efficiency and improving the aerial photography effect.

[0140] The application process of the unmanned aerial vehicle monitoring method based on the cloud control platform in the above application scenario is provided below. In this application scenario, the entropy value of the gray level co-occurrence matrix is used as the ground object complexity index. The unmanned aerial vehicle monitoring method based on the cloud control platform mainly includes:

[0141] Step one, data preparation;

[0142] ​​​​​​​​​​First, obtain remote sensing satellite images of the flight area where the UAV flight route is located, including all images within three years. The remote sensing image closest to the current time can be used to assign an initial speed to the UAV, and other remote sensing images can be used to obtain a probability table for predicting adjustment probabilities.

[0143] Second, obtain aerial images obtained by the UAV when flying along the UAV flight route in the flight area.

[0144] Step two, set an initial speed for the UAV based on the latest remote sensing image, mainly including:

[0145] First, obtain the latest high-resolution remote sensing image corresponding to the flight area where the UAV flight route is located, for example: high-resolution satellite images can meet the data requirements of pixel ranges from 50m to 2m. High-resolution and large-scale high-precision remote sensing data can be provided, suitable for various application scenarios such as mineral resource investigation and monitoring, urban planning, geological disaster monitoring, land cover analysis, land use change analysis, and land surface temperature analysis.

[0146] The UAV flight area is usually only a part of the larger range of remote sensing images, so the flight area image can be cropped on the remote sensing image. Specifically: first, obtain the flight route of the UAV flight route on the remote sensing image through coordinate conversion and other methods, then obtain the circumscribed rectangle of the UAV flight route, and record the circumscribed rectangle area as the first cropped image. Then analyze the UAV flight route based on the first cropped image.

[0147] Obtain multiple superpixel blocks by SLIC superpixel segmentation method on the first cropped image. Obtain the circumscribed rectangle of the smallest superpixel block, and use the size of the rectangle as the standard grid size based on the retained area. Divide the first cropped image into multiple grid regions by the standard grid size.

[0148] Grid division is on one hand to facilitate the subsequent prediction of the possible terrain complexity index of the next flight area of the UAV by Markov prediction idea. On the other hand, using the circumscribed rectangle of the smallest superpixel block as the standard grid size is to make the geographical scenery in the same grid as possible, and the same grid region is the same type of terrain.

[0149] For each grid region, calculate the entropy value of the gray level co-occurrence matrix in the region. The greater the entropy value, the richer the texture of the grid region, i.e. the more complex the terrain, and the slower the speed of the aircraft needs to be to obtain high-quality images for subsequent analysis.

[0150] arranging the entropy values in ascending order to obtain an ascending sequence, and obtaining multiple segmentation sections by an Otsu multi-threshold segmentation method on the ascending sequence, the entropy values in the same segmentation section being similar, i.e., the different entropy values in the same segmentation section corresponding to the complex degrees of the ground objects in the grid regions being similar. The grid region with the most frequent complex degree of ground objects is set with a flight speed standard value, and the initial speeds of other segmentation sections can be adjusted by reference to the standard value. The adjustment manner can be that the initial speed of other segmentation sections is obtained by calculating the ratio of the representative values of the entropy values of the reference segmentation section and other segmentation sections (the median value of the entropy value range can be selected as the representative value of the entropy value range), and the product of the ratio and the initial speed corresponding to the reference segmentation section is taken as the initial speed of the grid region corresponding to other segmentation sections.

[0151] It should be noted that the above Otsu multi-threshold segmentation is a relatively mature threshold segmentation method in the art, and the specific implementation paradigm can be referred to related technologies, and the embodiments of the present application will not be repeated.

[0152] Step three, correcting the speed of the unmanned aerial vehicle based on the actual aerial image of the unmanned aerial vehicle, mainly including:

[0153] When the unmanned aerial vehicle flies to a certain grid region, if the actual scene of the region is quite different from the remote sensing image of the grid region, the initial speed of the grid region is less suitable. The greater the complex degree of the actual region compared with the grid region image, the more the speed needs to be slowed down, and more information needs to be collected for the region. After that, clearer or more unmanned aerial vehicle image data can be obtained for subsequent analysis, whether it is surface cover analysis or disaster monitoring.

[0154] The current coordinate position of the unmanned aerial vehicle is obtained through the GPS sensor, and then the grid region where the unmanned aerial vehicle is located can be obtained. The multiple aerial images of the current grid region obtained by the unmanned aerial vehicle are obtained by image stitching to obtain a stitched image. The stitched image is used as a mask to obtain the image of the corresponding region on the collected image, which is recorded as a second cropped image.

[0155] The stitched image is obtained by the mean resampling method to obtain an image with the same resolution as the second cropped image, which is recorded as a resampled image. The mean square error (MSE) value of the resampled image and the second cropped image is calculated and normalized. If the normalized MSE value is greater than 0.3, the actual ground object in the flight region of the unmanned aerial vehicle is quite different from the ground object on the second cropped image. At this time, the initial speed needs to be adjusted in combination with the ground object change. Specifically:

[0156] ​Calculate the gray level co-occurrence matrix entropy value of the spliced image, denoted as a; calculate the gray level co-occurrence matrix entropy value of the second cropped image, denoted as b;

[0157] If a≤b, the speed does not change, and the unmanned aerial vehicle continues to fly in the grid area at the initial speed;

[0158] If a>b, it indicates that the ground features in the current grid area have changed greatly, and the complexity of the current ground features exceeds the complexity of the ground features on the remote sensing image. On the basis of the pre-set initial speed, the speed needs to be reduced. The product of the ratio of b to a and the initial speed of the grid area is taken as the corrected speed of the unmanned aerial vehicle;

[0159] Then the unmanned aerial vehicle continues to fly in the current grid area at the corrected speed.

[0160] Step four, correcting the initial speed of the next grid area based on the Markov prediction idea, mainly including:

[0161] In order to avoid the problem of not timely adjustment, resulting in the problem that the unmanned aerial vehicle image is difficult to support subsequent analysis, for example: subsequent detailed analysis of land changes is required, but the unmanned aerial vehicle speed is too fast, resulting in a small number of images taken, the Markov prediction idea can be combined to obtain the possible actual geographical density of the unmanned aerial vehicle in the next grid area in advance, and the speed of the unmanned aerial vehicle is adjusted in advance to avoid the problem that the unmanned aerial vehicle image taken due to the too fast speed cannot accurately monitor the changes of the land, affecting the subsequent analysis of the unmanned aerial vehicle image.

[0162] In addition, the ground features in each square are relatively single, and the ground features in different squares may be the same or different. If the ground features of two adjacent squares are the same, for example: both are farmland, the ground complexity of different growth stages is different, but the change with time is similar, that is, the probability of the ground complexity of adjacent land types is similar, and then the ground complexity of adjacent land can be predicted through the surrounding land by combining historical data.

[0163] Specifically: when the unmanned aerial vehicle is about to leave the current grid area and enter the next grid area, the ground complexity of the next grid area is predicted. Obtain the grid area of multiple different periods of cropped images through Gaofen 1 (Gaofen 1 satellite is taken as an example, and relevant remote sensing data can be obtained according to actual needs). Each grid area in each period corresponds to an entropy value range (here, the entropy value range refers to the entropy value range of the segmentation section in the above scheme). By statistically analyzing the data of each grid area in all periods, the probability of each direction of the grid located in different entropy value ranges in the four directions of each grid area in different entropy value ranges can be obtained, forming a probability table. Through the entropy value range of the target grid and the relative direction of the adjacent grid to the target grid, the entropy value range of the adjacent network can be obtained.

[0164] The current grid area where the UAV is located is obtained, and the next grid area can be obtained in combination with the UAV flight route. The entropy value a of the current grid area is obtained by calculation, and in combination with the a value and the relative position of the next grid area and the current grid area, the entropy value range of different probabilities of the next grid area can be obtained. Then, a representative value is calculated for each entropy value range of the next grid area, and the median value of each entropy value range is taken as the representative value, which is denoted as a predicted representative value. The median value of the entropy value range corresponding to the next grid area on the latest image is obtained as a representative value, which is denoted as a latest representative value. The predicted representative values greater than the latest representative value are screened out, and the probabilities of the screened representative values are added together as an adjustment probability. If the adjustment probability is greater than 60%, it indicates that the next grid area has a greater probability of increasing the complexity of the ground object, and the initial speed of the grid area set before is not suitable, and needs to be slowed down. The initial speed of the next area is adjusted to the product of the initial speed of the next grid area and the adjustment probability p.

[0165] Step five, controlling the UAV to fly along the UAV flight route to obtain aerial images.

[0166] Through the above method, the UAV can obtain high-quality UAV images, which is convenient for subsequent analysis and processing. Whether it is geological disaster monitoring, land cover analysis, land use change analysis, or as a data set for scientific research, high-quality UAV aerial images with rich quantities of different land types can be obtained.

[0167] Thus, the present application is completed.

[0168] To sum up, in the embodiment of the present application, the first cropped image of the flight area where the UAV flight route is located is obtained in the remote sensing image; the grid area image of a plurality of grid areas is obtained by grid division on the first cropped image; the ground object complexity index of each grid area image is obtained; the initial speed of the UAV when flying in each grid area is obtained based on the ground object complexity index of each grid area image; wherein the initial speed is inversely proportional to the ground object complexity index; the UAV is controlled to fly along the UAV flight route to obtain aerial images. The present application actively adjusts the flight speed of the UAV, which can effectively avoid the image blur or shaking problem caused by too fast or too slow speed, ensures that the images collected by the UAV have higher stability and clarity, and helps to improve the defect recognition rate and monitoring imaging effect.

[0169] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.​

Claims

1. A method for monitoring unmanned aerial vehicles (UAVs) based on a cloud control platform, characterized in that, The method comprises: acquiring a first cropped image of a flight area where a flight route of a UAV is located in a remote sensing image; performing grid division on the first cropped image to acquire grid area images of a plurality of grid areas; acquiring a ground feature complexity index of each of the grid area images; wherein the ground feature complexity index is used to represent texture richness of the grid area image; acquiring an initial speed of the UAV when flying in each of the grid areas based on the ground feature complexity index of each of the grid area images; wherein the initial speed is inversely proportional to the ground feature complexity index; controlling the UAV to fly along the flight route of the UAV to acquire aerial images; The method further comprises: acquiring a coordinate position of the UAV when controlling the UAV to fly along the flight route of the UAV; acquiring a current grid area where the UAV is located based on the coordinate position; splicing aerial images of the UAV in the current grid area to acquire a spliced image; acquiring a second cropped image corresponding to the spliced image in the remote sensing image; resampling the spliced image to acquire a resampled image having the same resolution as the second cropped image; acquiring a difference index of the resampled image and the second cropped image; wherein the difference index is used to represent a difference degree between the resampled image and the second cropped image; when the difference index is greater than a preset difference index threshold, correcting the initial speed corresponding to the current grid area to acquire a first corrected speed; controlling the UAV to fly along the flight route of the UAV in the current grid area at the first corrected speed. 2.The cloud-control-platform-based UAV monitoring method of claim 1, wherein, The acquiring of the initial speed of the UAV when flying in each of the grid areas based on the ground feature complexity index of each of the grid area images comprises: performing threshold segmentation on the ground feature complexity index of a plurality of the grid area images to acquire a plurality of segmented sections; acquiring a reference segmented section; wherein the reference segmented section is the segmented section corresponding to the grid area image with the largest number among all the segmented sections; acquiring an initial speed of the UAV when flying in a grid area corresponding to the reference segmented section; acquiring the initial speed of the UAV when flying in the grid area corresponding to other segmented sections based on a ratio between a representative value of the ground feature complexity index of other segmented sections and a representative value of the ground feature complexity index of the reference segmented section. 3.The cloud-control-platform-based UAV monitoring method of claim 1, wherein, The performing of the grid division on the first cropped image to acquire the grid area images of a plurality of grid areas comprises: performing superpixel segmentation on the first cropped image to acquire a plurality of superpixel blocks; acquiring a circumscribed rectangle of the superpixel block with the smallest area among the plurality of superpixel blocks; performing grid division on the first cropped image to acquire the grid area images of a plurality of grid areas by taking the size of the circumscribed rectangle as a standard grid size. 4.The cloud-control-platform-based UAV monitoring method of claim 1, wherein, The correcting of the initial speed corresponding to the current grid area to acquire a first corrected speed comprises: acquiring a ground feature complexity index of the spliced image and a ground feature complexity index of the second cropped image; Correct the initial speed corresponding to the current grid region based on a ratio of the ground feature complexity index of the spliced image to the ground feature complexity index of the second cropped image, to obtain a first corrected speed. 5.The cloud-control-platform-based UAV monitoring method of claim 1, wherein, The method further comprises: Before the UAV enters a next grid region, predicting the ground feature complexity index of the next grid region based on historical remote sensing images to obtain an adjustment probability that a ground feature complexity prediction index is greater than the ground feature complexity index of the next grid region; Correcting the initial speed corresponding to the next grid region based on the adjustment probability to obtain a second corrected speed; After the UAV enters the next grid region, controlling the UAV to fly along the UAV flight route in the next grid region at the second corrected speed. 6.The cloud-control-platform-based UAV monitoring method of claim 5, wherein, The prediction of the ground feature complexity index of the next grid region based on historical remote sensing images to obtain an adjustment probability that a ground feature complexity prediction index is greater than the ground feature complexity index of the next grid region comprises: Obtaining third cropped images of a flight region where the UAV flight route is located in historical remote sensing images of different historical periods; Dividing the third cropped images into grid regions with a standard grid size used when dividing the first cropped image into grid regions to obtain historical grid region images of a plurality of historical grid regions; Obtaining historical ground feature complexity indexes of the historical grid region images; Threshold segmenting the ground feature complexity indexes of the grid region images to obtain a plurality of segmented sections; Threshold segmenting the historical ground feature complexity indexes of the historical grid region images to obtain a plurality of historical segmented sections; Obtaining the historical segmented sections corresponding to each of the historical grid region images; Statistically obtaining a probability table from the historical segmented sections corresponding to each of the historical grid region images of all the historical periods; wherein the probability table comprises probabilities that the historical grid region is in different historical segmented sections and the historical grid region adjacent to the historical grid region is in different historical segmented sections; Based on the relative position between the next grid region and the current grid region and the ground feature complexity index of the current grid region, obtaining probabilities that the next grid region is in different historical segmented sections in the probability table; Obtaining an adjustment probability; wherein the adjustment probability represents a probability sum of all historical segmented sections in which a representative value of the historical ground feature complexity index corresponding to the next grid region is greater than a representative value of the ground feature complexity index of the next grid region. 7.The cloud-control-platform-based UAV monitoring method according to claim 5, characterized in that, The correction of the initial speed corresponding to the next grid region based on the adjustment probability to obtain a second corrected speed comprises: Based on the adjustment probability, obtaining a correction factor; wherein the correction factor is negatively correlated with the adjustment probability; Based on the correction factor, correcting the initial speed corresponding to the next grid region to obtain a second corrected speed. 8.The cloud-control-platform-based UAV monitoring method of claim 1, wherein, The obtaining of the ground feature complexity index of each of the grid region images comprises: An entropy value of a gray level co-occurrence matrix of each of the grid region images is obtained, wherein the entropy value is directly proportional to the complexity of the ground object. 9.The cloud-control-platform-based UAV monitoring method of claim 1, wherein, The first clipping image of the flight region where the flight route of the unmanned aerial vehicle is located in the remote sensing image comprises: An outer rectangle of the flight route of the unmanned aerial vehicle on the remote sensing image is obtained. The region corresponding to the outer rectangle is taken as the first clipping image of the flight region where the flight route of the unmanned aerial vehicle is located.

Images