A dynamic fuzzy target recognition method and system for unmanned aerial vehicles

Abstract

The present application relates to the field of image processing, in particular to a kind of unmanned aerial vehicle dynamic fuzzy target identification method and system, comprising: obtaining final area and contrast edge in final area;The chain code of different contrast edge in each final area is compared, and the fuzzy performance of each final area is obtained;According to the gradient amplitude of each pixel point on each contrast edge in each final area and the distribution of gradient amplitude, and the fuzzy performance of each final area, the final blurring degree of each final area is obtained;According to the final blurring degree of each final area, the blur kernel of each final area is obtained, and the deblurring processing of image is completed.The present application aims to solve the problem that when the current image is deblurred, the same degree of deblurring intensity is used for different parts of the image, so that the deblurring effect of the image is relatively poor.

Description

Technical Field

[0001] This invention relates to the field of image processing, and specifically to a method and system for identifying dynamically blurred targets from unmanned aerial vehicles (UAVs). Background Technology

[0002] With the rapid development of drone technology, drones have been widely used in aerial surveying, environmental monitoring, security patrols, and agricultural plant protection. In typical application scenarios, drones equipped with imaging devices capture images of target areas at low or high altitudes to obtain regional image information for subsequent analysis and decision-making. However, during actual drone flight and filming, the relative motion between objects within the target area and the drone creates complex and spatially non-uniform dynamic changes in the filmed scene, resulting in some areas of an image being blurry. Therefore, image deblurring processing is necessary.

[0003] Existing deblurring methods apply the same level of deblurring to different regions of an image. However, due to the typical spatial variability of blur in UAV aerial images, a globally uniform deblurring strategy leads to unsatisfactory results. Specifically, for highly blurred regions, the deblurring amplitude may be insufficient, failing to effectively restore detailed textures; while for originally clear or slightly blurred regions, excessive deblurring can introduce ringing artifacts, noise amplification, or edge distortion, resulting in image quality degradation. Therefore, there is an urgent need to propose an adaptive deblurring method that can address the differences in blur levels across different regions of UAV aerial images, overcoming the problem of relatively poor deblurring results from existing globally uniform deblurring methods. Summary of the Invention

[0004] This invention provides a method for identifying dynamically blurred targets from unmanned aerial vehicles (UAVs) to solve the problem that the deblurring results are not ideal when performing deblurring operations on images acquired by UAVs.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Use an image acquisition device mounted on the drone to acquire images captured by the drone;

[0007] Edge detection algorithms are used to detect edges in images captured by drones. Based on these edges, final regions and the edges within those regions are identified. The intersection points of the edges within each final region are used as segmentation points to segment the edges, resulting in contrasting edges within each final region. An 8-connected chain code is used to obtain the chain code for each contrasting edge within each final region. The chain codes of different contrasting edges within each final region are compared to determine the morphological differences between each contrasting edge and other contrasting edges within each final region. Based on these morphological differences, the blur representation of each final region is determined.

[0008] Based on the gradient magnitude and distribution of pixels on each contrast edge in each final region, and the blurring behavior of each final region, the final blur level of each final region is obtained; based on the final blur level of each final region, the length and width of the blur kernel of each final region are obtained; based on the length and width of the blur kernel of each final region, the blur kernel of each final region is obtained; based on the blur kernel of each final region in the image, the image deblurring process is completed.

[0009] Furthermore, the specific calculation steps for obtaining the final region based on the edges in the image, and the edges located within the final region, are as follows:

[0010] The closed region enclosed by the edge in the image taken by the drone is recorded as the initial region, and all the initial regions in the image taken by the drone are obtained.

[0011] If the image captured by the drone contains the first If an initial region is not part of any other initial region in the image, it is recorded as the final region.

[0012] Will be located in the The edges of the final region and the surrounding area The edge of each final region is denoted as the first... The edge of the final region.

[0013] Furthermore, the specific calculation steps for comparing the chain codes of different contrast edges within each final region to obtain the morphological differences between each contrast edge and other contrast edges within each final region are as follows:

[0014] The first The first in the final region The contrast edge and the first By combining the contrasting edges, we obtain the first... The first in the final region The first line of contrasting edges A number of combinations;

[0015] If the first The first in the final region The first line of contrasting edges If the chain code lengths of the two edges in the combined pair are the same, then the first pair will be... The first in the final region The contrast edge and the first The contrast edges are respectively denoted as the first. The first in the final region The first line of contrasting edges The control edge and the final reference edge in each combination pair;

[0016] If the first The first in the final region The first line of contrasting edges If the chain code lengths of the two contrasting edges in the first pair are different, then the first pair will be... The first in the final region The contrast edge and the first The longest and shortest chain code lengths of the edges in the comparison edges are denoted as the first and second edges, respectively. The first in the final region The first line of contrasting edges The long edge and the control edge in each combination pair;

[0017] For the first The first in the final region The first line of contrasting edges By comparing and analyzing the chain codes of the two contrasting edges in the first combination pair, we obtain the first... The first in the final region The first line of contrasting edges The final reference edge in each combination pair;

[0018] The first The first in the final region The first line of contrasting edges The chain codes of the corresponding edges in each combination pair are compared one by one with the final reference edges to obtain the first... The first in the final region The contrast edge and the first Edge similarity of contrasting edges;

[0019] According to the The edge similarity of different contrasting edges within each final region is used to obtain the blurred representation of each final region.

[0020] Furthermore, the aforementioned [context missing] The first in the final region The first line of contrasting edges By comparing and analyzing the chain codes of the two contrasting edges in the first combination pair, we obtain the first... The first in the final region The first line of contrasting edges The specific steps for calculating the final reference edge in each combination pair are as follows:

[0021] With the first The first in the final region The first line of contrasting edges The chain code length of the control edge in the combined pair is the window scale, with a step size of 1, starting from the . The first in the final region The first line of contrasting edges Slide the chain code on the long edge of each combination pair from beginning to end, and denote the resulting equal-length chain code segments as the th... The first in the final region The first line of contrasting edges Suspected reference edges in a combination pair;

[0022] Get the The first in the final region The first line of contrasting edges The first of the combinations The specific formula for calculating the morphological difference between a suspected reference edge and a control edge is as follows:

[0023]

[0024]

[0025]

[0026] In the formula, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The morphological differences between a suspected reference edge and a control edge. Indicates the first The first in the final region The first line of contrasting edges The edge morphology of the contrast edge in each combination pair Indicates the first The first in the final region The first line of contrasting edges The first of the combinations An edge shape that is suspected to be a reference edge. Indicates the ordinal value. Indicates the first The first in the final region The first line of contrasting edges The number of digits contained in the chain code of the contrast edge in each combination pair. Indicates the first The first in the final region The first line of contrasting edges The numbers on the chain code at the contrast edge of each combination pair Quantity, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations Numbers on a chaincode that appears to be at a reference edge Quantity, Represents the absolute value function; where the first... The first in the final region The first line of contrasting edges The first of the combinations The morphological differences between the suspected reference edge and the control edge are as follows: The first in the final region The first line of contrasting edges The first of the combinations A suspected reference edge, with the first The first in the final region The first line of contrasting edges Morphological differences in the control edges of each combination pair;

[0027] According to the The first in the final region The first line of contrasting edges The morphological differences between each suspected reference edge and control edge in each combination pair were analyzed using the Otsu method. The first in the final region The first line of contrasting edges Clustering the suspected reference edges in each combination pair yields two clusters;

[0028] Connect all suspected reference edges in the two clusters with the first... The first in the final region The first line of contrasting edges The cluster containing the lowest mean morphological difference in the control edges among the 10 pairs of combinations is denoted as the 1st cluster. The first in the final region The first line of contrasting edges The final reference edge in each combination pair;

[0029] If the first The first in the final region The first line of contrasting edges If the number of suspected reference edges in the combined pairs is 2, then the th pair will be... The first in the final region The first line of contrasting edges The suspected reference edges in each of the combined pairs are denoted as the i-th. The first in the final region The first line of contrasting edges The final reference edge in each combination pair.

[0030] Furthermore, the first The first in the final region The first line of contrasting edges The chain codes of the corresponding edges in each combination pair are compared one by one with the final reference edges to obtain the first... The first in the final region The contrast edge and the first The specific steps for calculating the edge similarity of the contrasting edges are as follows:

[0031] If the first The first in the final region The first line of contrasting edges The contrast edge in the first combination pair and the first The first in the chaincode of the final reference edge If the values ​​of the nth numbers are the same, then the nth number... The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The difference is 0 if the values ​​are different; if the values ​​are different, the difference is 1.

[0032] By using the first chain code in each edge The method of generating a window containing five numbers centered on the nth number yields the nth... The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The comparison window for the number, and the number The first in the final region The first line of contrasting edges The first of the combinations The first in the chain code of the first contrast edge A comparison window for each number;

[0033] Get the The first in the final region The first line of contrasting edges The first of the combinations The specific formula for calculating the edge similarity between the final reference edge and the control edge is as follows:

[0034]

[0035]

[0036] In the formula, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations Edge similarity between the final reference edge and the control edge. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The final difference, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge One difference, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The number of digits contained in the chaincode of the final reference edge. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The number of digits contained in the comparison window. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The degree of difference; if the first... The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The number, with the first The first in the final region The first line of contrasting edges The first chain code of the corresponding edge in the first combination pair The first number in the comparison window If the values ​​of the numbers are the same, then ,otherwise ;

[0037] The first The first in the final region The first line of contrasting edges The maximum value among the edge similarities of all final reference edges and control edges in the combined pairs is denoted as the i-th... The first in the final region The contrast edge and the first Edge similarity of the contrasting edges.

[0038] Furthermore, the statement based on the first The specific calculation steps for obtaining the blurred representation of each final region by calculating the edge similarity of different contrasting edges within each final region are as follows:

[0039] Using Otsu's method to... The first in the final region Clustering is performed on the edge similarity of the single contrasting edge and all contrasting edges to obtain two clusters;

[0040] Let the edge similarities within the cluster with the largest mean edge similarity between two clusters be denoted as the i-th. The first in the final region The final edge similarity of the contrasting edges;

[0041] The maximum value of the edge similarity among the clusters with the smallest mean edge similarity is denoted as the i-th cluster. The first final region One difference similarity;

[0042] If the first The first in the final region The similarity of each final edge of the contrasting edges is greater than that of the first edge. For each difference similarity in the final region, the first... The first in the final region The final edge similarity of the contrasting edges is denoted as the i-th edge. The first in the final region The final edge similarity representation of the contrasting edges, and the first edge The first in the final region A single contrasting edge is denoted as a blurred representation edge;

[0043] The first The first in the final region The sum of the final edge similarity representations of all contrasting edges, denoted as the i-th. The first in the final region The degree of fuzziness in contrasting edges;

[0044] The first The mean of the fuzziness of all contrast edges within the final region is denoted as the i-th. The final region is a blurry representation.

[0045] Furthermore, the specific steps for calculating the final blur level of each final region based on the gradient magnitude and distribution of pixels on each contrast edge within each final region, and the blurring behavior of each final region, are as follows:

[0046] Get the The first in the final region The specific formula for calculating the motion performance of contrasting edges is as follows:

[0047]

[0048] In the formula, Indicates the first The first in the final region The motion performance of the contrasting edges Indicates the first The first in the final region The variance of the gradient magnitude of all pixels on the contrast edge. Indicates the first The first in the final region The average gradient magnitude of all pixels on the contrast edge;

[0049] The first The first in the final region The average gradient magnitude of all pixels on the edge of the comparison is denoted as the i-th. The first in the final region The gradient mean of the contrasting edges;

[0050] Get the The specific formula for calculating the final ambiguity of each final region is as follows:

[0051]

[0052] In the formula, Indicates the first The final ambiguity of the final region, Indicates the first The mean of the motion performance of all contrasting edges within the final region. Indicates the first The average of the gradients of all contrasting edges within the final region. Indicates the first The blurry representation of the final region This represents the sigmoid function.

[0053] Furthermore, the specific calculation steps for obtaining the length and width of the blur kernel of each final region based on the final blur degree of each final region are as follows:

[0054] Get the The specific formula for calculating the length of the fuzzy kernel of each final region is as follows:

[0055]

[0056] In the formula, Indicates the first The length of the fuzzy kernel of the final region, This represents the minimum value of the final ambiguity among all final regions. This represents the floor function; and the length of the blur kernel for each final region is less than or equal to 31; where, the first... The length and width of the fuzzy kernel of each final region are the same.

[0057] A second aspect of the present invention provides a dynamic blurred target recognition system for unmanned aerial vehicles (UAVs), the system comprising an image acquisition module, an initial blur determination module, and a deblurring operation module, wherein:

[0058] Image acquisition module: Uses an image acquisition device mounted on the drone to acquire images captured by the drone;

[0059] Initial Blur Determination Module: This module uses an edge detection algorithm to detect edges in images captured by the drone, obtaining the edges within the images. Based on these edges, it identifies the final region and the edges located within it. The module then segments the edges within each final region using the intersection points of the edges as segmentation points, obtaining the contrast edges within each final region. An 8-connected chain code is used to obtain the chain code for each contrast edge within each final region. The chain codes of different contrast edges within each final region are compared to determine the morphological differences between each contrast edge and other contrast edges within each final region. Based on these morphological differences, the module determines the blur representation of each final region.

[0060] Deblurring module: Based on the gradient magnitude and distribution of pixels on each contrast edge in each final region, and the blurring performance of each final region, the final blur level of each final region is obtained; based on the final blur level of each final region, the length and width of the blur kernel of each final region are obtained; based on the length and width of the blur kernel of each final region, the blur kernel of each final region is obtained; based on the blur kernel of each final region in the image, the image deblurring process is completed.

[0061] The beneficial effects of the technical solution of the present invention are:

[0062] 1. This invention, when deblurring images, fully considers the typical spatial differences in blurriness in UAV aerial images. During the deblurring process, it first performs edge detection and divides the image into final regions. For each final region, its blur representation and final blur degree are calculated separately, and then a suitable blur kernel is generated for each region. This allows the deblurring intensity to adaptively adjust according to the actual blur degree of each region of the image, effectively overcoming the problem of unsatisfactory processing results caused by globally uniform deblurring.

[0063] 2. In the process of estimating the blur kernel length, this invention utilizes motion blur to degenerate sharp edges into gentle transition zones, resulting in relatively small gradient magnitudes for edge pixels within the blurred region. It integrates multi-dimensional information such as the morphological similarity of edges within the region (i.e., blur performance), the gradient magnitude of edge pixels, and the dispersion of gradient magnitude distribution. This allows for a quantitative evaluation of the blur degree in each region of the image, making the estimated blur kernel length more closely match the actual blur level of each region, thus improving the accuracy of blur kernel estimation. This significantly improves the overall sharpness and detail fidelity of the image obtained after processing, providing higher-quality input data for subsequent advanced vision tasks such as target detection, recognition, and tracking. Compared to traditional global deblurring methods, this invention better preserves the original texture information and edge structure of the image, avoiding information loss due to over-processing. This improves the target recognition accuracy and the reliability of analysis and decision-making in applications such as aerial surveying, environmental monitoring, and security patrol using UAVs. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1This is a flowchart illustrating the steps of a method for identifying dynamically blurred targets using a drone, according to an embodiment of the present invention. Detailed Implementation

[0066] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a UAV dynamic fuzzy target recognition method and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0067] 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 invention pertains.

[0068] Example 1:

[0069] This invention provides a method for identifying dynamically blurred targets in unmanned aerial vehicles (UAVs), specifically as follows: Figure 1 As shown, it includes:

[0070] Step S001: Use the image acquisition device mounted on the drone to acquire images captured by the drone.

[0071] Specifically, an image acquisition device mounted on a drone is used to photograph the area under test, resulting in images captured by the drone. The use of a drone to photograph the target area is a well-known technique and will not be elaborated upon in this embodiment.

[0072] At this point, the images taken by the drone were obtained.

[0073] Step S002: Use an edge detection algorithm to perform edge detection on the image captured by the drone to obtain the edges in the image; based on the edges in the image, obtain the final region and the edges located within the final region; use the intersection points of the edges in each final region as segmentation points to segment the edges and obtain the contrast edges in each final region; use 8-connected chain codes to obtain the chain code of each contrast edge in each final region; compare the chain codes of different contrast edges in each final region to obtain the morphological differences between each contrast edge in each final region and other contrast edges; based on the morphological differences between each contrast edge in each final region and other contrast edges, obtain the blur representation of each final region.

[0074] It's important to note that when taking photos with a camera, an exposure time needs to be set. The camera's exposure time refers to the duration from when the shutter opens to when it closes; it can be understood as the length of time the camera records light information. During this time, light passes through the lens and illuminates the image sensor, such as a CMOS or CCD, where it is converted into electrical signals, ultimately forming a digital image. If the object moves relative to the camera during the exposure time, the moving object in the image will appear blurry, requiring deblurring processing.

[0075] It should be further noted that during the actual flight and shooting of the drone, different objects in the target area may have relative motion with the drone, resulting in complex and spatially non-uniform dynamic changes in the scene being filmed. This can cause some areas of an image to be blurry. Therefore, the images taken by the drone are deblurred.

[0076] It should be further noted that existing deblurring methods, such as non-blind deconvolution algorithms based on prior knowledge, end-to-end deblurring networks based on deep learning, and methods using convolutional kernels, generally assume that the blur kernels in the image have global consistency, or use a globally uniform deblurring intensity to process the entire image. In other words, when deblurring images using existing techniques, the same magnitude of deblurring is applied to different regions. However, due to the typical spatial variability of blur in UAV aerial images, using a globally uniform deblurring strategy will lead to unsatisfactory results. Specifically, for areas with high blur levels, the deblurring amplitude may be insufficient, failing to effectively restore detailed textures; while for areas that were originally clear or slightly blurred, excessive deblurring can introduce ringing artifacts, noise amplification, or edge distortion, causing image quality degradation. Ultimately, this results in a weak deblurring effect when using existing techniques to deblur images. Therefore, this paper proposes an adaptive deblurring method that can address the differences in blur levels in different regions of UAV aerial images, thereby improving the image deblurring effect.

[0077] It should be further explained that, since different regions of the image need to be deblurred with different intensities, and these regions differ from each other, edge detection is first performed on the image to divide it into several regions before deblurring.

[0078] It should be further explained that, because the grayscale values ​​of pixels differ in different regions of an image, edge detection is performed to obtain the edges in the image. Then, the image is divided into multiple regions by the areas enclosed by these edges. Since subsequent analysis requires analyzing the edges within a single region, a relatively small threshold is used during edge detection to detect weak edges in the image.

[0079] It's important to further explain that during the camera's exposure time, when an object is moving, it will create artifacts distributed along its trajectory in the captured image. Because objects may deform during movement, but their overall shape usually remains stable for a short period, the image area corresponding to the moving object will contain several similar or repetitive edge structures. It's noteworthy that the more identical edges a region contains, the stronger the motion of the object corresponding to that region, because rapid movement can cause the edges of the same object to overlap or trail multiple times during the exposure.

[0080] It should be further noted that, due to the influence of object motion on the edges within a region, the edges formed by object motion are not sharp boundaries. Therefore, when performing edge detection on an image, a relatively small threshold is used to obtain all edges in the image. Then, different edges within a region are compared to obtain the similarity between different edges within each region and other edges. Based on the similarity of different edges within each region, the blur representation of each region is obtained.

[0081] It should be further explained that when judging the similarity between two edges, since chain code can represent the positional relationship between adjacent pixels on the edge, the chain code of each edge is compared with the chain code of other edges to obtain the similarity between the two edges.

[0082] It should be further explained that when comparing the chain codes of two edges to determine their similarity, since the chain code lengths of the two edges may be different, the longer chain code is first segmented into several equal-length short chain code segments. Directly comparing each segment with the short chain code would be computationally intensive. Therefore, a subset of chain code segments is selected as representative for subsequent similarity calculations. The selection criteria are: when the shapes and trends of two edges are similar, the distribution of numbers in their chain codes is also consistent; representative chain code segments are selected based on this. Then, the similarity of the two edges is obtained by comparing the chain code used for similarity calculation with the shorter chain code.

[0083] It should be further explained that when comparing each chain code used for similarity calculation with a shorter chain code to obtain the similarity of two edges, both are first divided into several equal-length sub-segments. Then, based on the distribution relationship of the numbers on the corresponding sub-segments, the similarity of the two chain code segments is calculated, thereby obtaining the overall similarity of the two edges.

[0084] Specifically, using a low threshold The Canny edge detection algorithm is used to detect edges in images captured by a drone, obtaining edge pixels in the images. In this embodiment, a preset low threshold is used. This is because the low threshold value of the Canny edge detection algorithm currently used when weak edges need to be detected is in the range of 30-50, and other values ​​can be set in other embodiments.

[0085] Furthermore, the edge pixels in the image captured by the drone are connected to obtain all the edges in the image. The closed regions enclosed by these edges in the image are designated as initial regions, resulting in all initial regions in the image. Note that different initial regions in the image may overlap, and one initial region may be part of another.

[0086] Furthermore, if the image captured by the drone contains the first... If the first initial region is not part of any other initial region in the image, then it is considered the final region. An initial region is a final region if it is not contained within any other initial region in the image. Conversely, if an initial region is part of another initial region, it is not a final region. A region is considered a final region if and only if the first initial region is not contained within any other initial region. The initial region contains the first When the initial region is b, that is, all pixels in the b-th initial region are in the b-th initial region. Within the initial region, the first The initial region is the first Part of an initial region.

[0087] Furthermore, the first image captured by the drone... The intersection points of the edges within the final region are used as segmentation points to segment the edges. Each segment is recorded as a contrast edge, resulting in the first segment in the image captured by the drone. The contrast edges within the final region. Specifically, the edges enclosing the first region in the drone-captured image. The edge of the final region is also the first The edges within the final region. Where, if the image captured by the drone contains the... If there is only one edge in the final region, then the region has no final blur, no subsequent operations are performed on the region, and no deblurring operation is performed on the region.

[0088] This yields the contrast edges within each final region of the image captured by the drone.

[0089] Furthermore, using 8-connected chain code, the first... The first in the final region The chain code for comparing edges is used. The chain code represents the positional relationship between two adjacent pixels, and the values ​​of the 8-connected chain code are 0, 1, 2, 3, 4, 5, 6, and 7. The method for obtaining the chain code of an edge in an image is a well-known existing technique and will not be described in detail in this embodiment. If there is only one edge in a final region, no deblurring operation is performed on that final region; that is, the final region has no final blur.

[0090] Furthermore, the first The first in the final region The contrast edge and the first By combining the contrasting edges, we obtain the first... The first in the final region The first line of contrasting edges A combination of pairs.

[0091] Furthermore, if the first The first in the final region The first line of contrasting edges If the chain code lengths of the two edges in the combined pair are the same, then the first pair will be... The first in the final region The contrast edge is denoted as the first. The first in the final region The first line of contrasting edges The contrast edge in the first combination pair will be the first The first in the final region The contrast edge is denoted as the first. The first in the final region The first line of contrasting edges The final reference edge in each combination pair. Then perform operation b.

[0092] Furthermore, if the first The first in the final region The chain code at the edge of the comparison is compared with the first If the chain code lengths of the contrasting edges are different, then the first... The first in the final region The contrast edge and the first The edge with the longest chain code among the contrasting edges is denoted as the first edge. The first in the final region The first line of contrasting edges The long edge in the combined pair. The first... The first in the final region The contrast edge and the first The edge with the shortest chain code length among the contrasting edges is denoted as the first edge. The first in the final region The first line of contrasting edges The contrast edges in each combination pair. Then, the operations are performed sequentially. With operation .

[0093] The operation as follows:

[0094] Furthermore, with the first The first in the final region The first line of contrasting edges The chain code length of the corresponding edge in each combination pair is .by Let the window size be 1, and the step size be 1, starting from the first... The first in the final region The first line of contrasting edges Slide the chain code on the long edge of each combination pair from beginning to end, and denote the resulting equal-length chain code segments as the th... The first in the final region The first line of contrasting edges Suspected reference edges in a combination pair.

[0095] Furthermore, obtain the first The first in the final region The first line of contrasting edges The first of the combinations The specific formula for calculating the morphological difference between a suspected reference edge and a control edge is as follows:

[0096]

[0097]

[0098]

[0099] In the formula, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The morphological differences between a suspected reference edge and a control edge. Indicates the first The first in the final region The first line of contrasting edges The edge morphology of the contrast edge in each combination pair Indicates the first The first in the final region The first line of contrasting edges The first of the combinations An edge shape that is suspected to be a reference edge. Indicates the ordinal value. Indicates the first The first in the final region The first line of contrasting edges The number of digits contained in the chain code of the contrast edge in each combination pair. Indicates the first The first in the final region The first line of contrasting edges The numbers on the chain code at the contrast edge of each combination pair Quantity, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations Numbers on a chaincode that appears to be at a reference edge Quantity, This represents the absolute value function. Due to mathematical rules, the independent variable in a logarithmic function cannot be zero; therefore, when... season ,when season .

[0100] It should be noted that, This reflects information entropy, and in this embodiment, it is used to quantify the first... The first in the final region The first line of contrasting edges The distribution of chain code values ​​on the chain code at the control edge of each combination pair.

[0101] It should be further clarified that, since information entropy is used to quantify the uncertainty, amount of information, or degree of disorder of a random variable or a system, when using information entropy to quantify the distribution of chain code values, if one chain code has all 0s and another chain code has all 1s, then the information entropy of the two chain codes is the same. and When multiplying, consider the distribution of numerical values ​​on the chain code.

[0102] It should be further explained that, The smaller the value, the more... The first in the final region The first line of contrasting edges The first of the combinations The distribution of numerical values ​​on the chain codes of the suspected reference edge and the control edge is quite similar, further suggesting that the shapes of these two edges may be more similar.

[0103] Furthermore, according to the first The first in the final region The first line of contrasting edges The morphological differences between each suspected reference edge and control edge in each combination pair were analyzed using the Otsu method. The first in the final region The first line of contrasting edges Clustering the suspected reference edges in the first pair of combinations yields two clusters. All suspected reference edges in the two clusters are then compared with the first... The first in the final region The first line of contrasting edges The cluster containing the lowest mean morphological difference in the control edges among the combined pairs is denoted as the i-th cluster. The first in the final region The first line of contrasting edges The final reference edge in each of the combined pairs. Among them, the first... The first in the final region The first line of contrasting edges The first of the combinations The morphological difference between the suspected reference edge and the control edge is the first. The first in the final region The first line of contrasting edges The first of the combinations A suspected reference edge, with the first The first in the final region The first line of contrasting edges The morphological differences between the reference and control edges in each combination pair are analyzed. Otsu's method, also known as the Otsu's method, is used for data clustering, a well-known technique that will not be elaborated upon in this embodiment. Otsu's method is an automatic global threshold selection algorithm widely used in image processing. Its core idea is to find an optimal threshold for a grayscale image, dividing pixels into "foreground" and "background" categories, maximizing the difference between these two categories. In this embodiment, when using Otsu's method for clustering, the morphological differences between the suspected reference and control edges are replaced with grayscale values ​​from the Otsu method for clustering.

[0104] It should be further explained that, since it is necessary to divide the suspected reference edges into those that can be further processed and those that do not, and since the Otsu method clusters the data into two categories, the final reference edges are obtained by using the Otsu method.

[0105] Furthermore, if the first The first in the final region The first line of contrasting edges If the number of suspected reference edges in the combined pairs is 2, then the th pair will be... The first in the final region The first line of contrasting edges The suspected reference edges in each of the combined pairs are denoted as the i-th. The first in the final region The first line of contrasting edges The final reference edge in each combination pair. The threshold of 2 is used in this section because a minimum of 3 data points are required when using Otsu's method to cluster the data.

[0106] Thus, the first... The first in the final region The first line of contrasting edges The final reference edge and contrast edge in each combination pair.

[0107] It should be noted that this step does not include the first... The first in the final region The first line of contrasting edges Among all suspected reference edges in a pair of combinations, the smallest morphological difference between the reference edge and the control edge is used as the similarity between the two edges in that pair. This is because the shape of an object may undergo slight changes during movement, and morphological difference is obtained in a relatively coarse way to measure the difference between the two edges. Therefore, several edges with small differences are selected as the final reference edges. Then, the final reference edges in each pair are compared with the control edges to obtain the similarity between the two edges in each pair.

[0108] The operation as follows:

[0109] Furthermore, according to the first The first in the final region The first line of contrasting edges The final reference edge and contrast edge in the nth combination pair are used to obtain the nth final region. The contrast edge and the first The specific steps for edge similarity comparison are as follows:

[0110] Furthermore, if the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The number and the first The first in the final region The first line of contrasting edges The first chain code of the contrast edge in the first combination pair If the values ​​of the nth numbers are the same, then the nth number... The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The difference is 0.

[0111] Furthermore, if the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The number and the first The first in the final region The first line of contrasting edges The first chain code of the contrast edge in the first combination pair If the values ​​of the nth and nth numbers are different, then the nth number... The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The difference is 1.

[0112] Furthermore, with the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge Centered on the first number, generate a window containing five numbers, denoted as the first. The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge A comparison window for the number. Among them, if the number... The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge If the number of digits on one or both sides of a number is less than two, the missing digits are not added. The reason for selecting two digits on each side of a number to form a window is that when the position of a pixel on the edge changes, it affects two chain code values. To ensure that the window contains chain code values ​​unaffected by changes in the position of the same pixel, at least two chain code values ​​need to be selected on one side. Since a chain code value may contain chain code values ​​on both sides, two digit values ​​need to be selected on each side when constructing the window for each number. Therefore, the length of the comparison window in this section is 5.

[0113] Furthermore, with the first The first in the final region The first line of contrasting edges The first chain code of the corresponding edge in the first combination pair Centered on the first number, generate a window containing five numbers, denoted as the first. The first in the final region The first line of contrasting edges The first of the combinations The first in the chain code of the first contrast edge A comparison window for each number. Among them, the first... The first in the final region The first line of contrasting edges The first of the combinations The first in the chain code of the first contrast edge The comparison window for the first number and the second number The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The length of the comparison window for each number is the same.

[0114] Furthermore, obtain the first The first in the final region The first line of contrasting edges The first of the combinations The specific formula for calculating the edge similarity between the final reference edge and the control edge is as follows:

[0115]

[0116]

[0117] In the formula, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations Edge similarity between the final reference edge and the control edge. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The final difference, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge One difference, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The number of digits contained in the chain code of the final reference edge, i.e. The length of the chain code. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The number of digits contained in the comparison window. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The degree of difference; if the first... The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The number, with the first The first in the final region The first line of contrasting edges The first chain code of the corresponding edge in the first combination pair The first number in the comparison window If the values ​​of the numbers are the same, then If the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The number, with the first The first in the final region The first line of contrasting edges The first chain code of the corresponding edge in the first combination pair The first number in the comparison window If the values ​​of the numbers are different, then .

[0118] It should be noted that, The value is obtained through the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The numerical values ​​surrounding the number and the first The first in the final region The first line of contrasting edges The first chain code of the corresponding edge in the first combination pair To determine if a number has the same value as the numbers surrounding it. This reduces the impact of localized deformation during an object's movement, which can cause differences between two edges that should be identical. The impact of local deformation is because when local deformation occurs, other pixels a certain distance away from the deformed pixel may not be affected by the deformation. This is mitigated by using other chain codes surrounding each chain code to reduce the impact of local deformation. The impact.

[0119] Furthermore, the first The first in the final region The first line of contrasting edges The maximum value among the edge similarities of all final reference edges and control edges in the combined pairs is denoted as the i-th... The first in the final region The contrast edge and the first Edge similarity of the contrasting edges.

[0120] At this point, the steps are complete. Finish.

[0121] Furthermore, using Otsu's method on the first The first in the final region Clustering the edge similarity of the first contrasting edge with all other contrasting edges yields two clusters. The edge similarities within the cluster with the highest mean edge similarity are denoted as the i-th cluster. The first in the final region The final edge similarity of the compared edges is calculated. The maximum edge similarity among the clusters with the smallest mean edge similarity is denoted as the i-th cluster. The first final region The difference is similar.

[0122] It should be further explained that Otsu's method is used to cluster the data because, under normal circumstances, each edge in each region will be more similar to some edges in that region and less similar to some edges in that region. That is, it is necessary to divide the edge similarity into two categories, and Otsu's method can divide the data into two categories, so Otsu's method is used.

[0123] Furthermore, if the first The first in the final region The similarity of each final edge of the contrasting edges is greater than that of the first edge. For each difference similarity in the final region, the first... The first in the final region The final edge similarity of the contrasting edges is denoted as the i-th edge. The first in the final region The final edge similarity representation of the contrasting edges, and the first edge The first in the final region The contrasting edges are denoted as blurred representation edges.

[0124] It should be noted that since the shape of one edge in a region is unlikely to be strongly similar to all edges in that region, Otsu's method is used to cluster the difference similarity to obtain edges that are relatively similar to each other. Because the similarity between two dissimilar edges is weaker than that between two similar edges, and because Otsu's method inevitably divides the data into two classes when classifying the data, each difference similarity of each final region is used as a threshold to remove edges that are not actually very similar.

[0125] Furthermore, the first The first in the final region The sum of the final edge similarity representations of all contrasting edges, denoted as the i-th. The first in the final region The degree of fuzziness in contrasting edges.

[0126] Furthermore, the first The mean of the fuzziness of all contrast edges within the final region is denoted as the i-th. The final region is a blurry representation.

[0127] At this point, the blurred representation of each final region is obtained, along with the blurred representation edges within each final region.

[0128] Step S003: Based on the gradient magnitude and distribution of pixels on each contrast edge in each final region, and the blurring performance of each final region, obtain the final blur level of each final region; based on the final blur level of each final region, obtain the length and width of the blur kernel of each final region; based on the length and width of the blur kernel of each final region, obtain the blur kernel of each final region; based on the blur kernel of each final region in the image, complete the image deblurring process.

[0129] It should be noted that when obtaining the deblurring intensity of each region based on the similarity of different edges within a region, some objects inherently possess regularly repeating texture structures, such as the black and white stripes on a zebra. Even when these objects are stationary or in weak motion, their image areas still contain a large number of similar edges. This means that when calculating the deblurring intensity of each region based on edge similarity, regions less affected by motion (i.e., those that were originally relatively clear) may be assigned a high level of blurring, leading to overprocessing. Therefore, a second adjustment is performed on the blurring representation of each region to obtain the final blur level for each region.

[0130] It should be further explained that when performing a second adjustment to the blur representation of each region to obtain the final blur level for each region, motion blur causes sharp edges of objects to degenerate into gentle gradient transitions. Edges within blurred regions become weak, and the gradient magnitude of pixels at those edges is relatively small. Conversely, the gradient magnitude of pixels at the edges within sharp regions is typically larger. Therefore, the final blur level for each region is calculated based on the gradient magnitude of pixels at the edges within that region.

[0131] It should be further explained that, because the grayscale values ​​of pixels in regions with motion artifacts are obtained by superimposing the grayscale values ​​of pixels at multiple different spatial locations during exposure, the gradient amplitudes of pixels at the edges of these regions originate from diverse sources, resulting in a high degree of dispersion in the statistical distribution of gradient amplitudes. For sharp regions unaffected by motion, the grayscale changes of their edge pixels mainly originate from differences in the object's material or texture, making the gradient amplitudes of pixels at the edges of these regions more consistent in origin, thus resulting in a more concentrated distribution of gradient amplitudes at these edges. Therefore, the final blur level of each region is obtained based on the gradient amplitudes of pixels at the edges and the distribution of pixel gradient amplitudes within each region.

[0132] Specifically, to obtain the first The first in the final region The specific formula for calculating the motion performance of contrasting edges is as follows:

[0133]

[0134] In the formula, Indicates the first The first in the final region The motion performance of the contrasting edges Indicates the first The first in the final region The variance of the gradient magnitude of all pixels on the contrast edge. Indicates the first The first in the final region The average gradient magnitude of all pixels on the contrast edge.

[0135] It should be noted that, The larger the value, the more significant the [value]. The first in the final region The more discrete the gradient magnitudes of pixels on a contrasting edge, the more consistent it is with the characteristic that when an edge is a motion-induced afterimage, the gradient magnitudes of pixels on that edge are more discrete; through Divide by This is used to prevent situations where the gradient magnitudes of pixels on an edge are all small. Even if the gradient magnitudes of pixels are relatively discrete, their variance is still relatively small because the overall range of gradient magnitude variation is relatively small. In other words, when the gradient magnitude of a pixel is small, its range of variation is also relatively small. Divide by This will resolve the issues mentioned in this paragraph.

[0136] Furthermore, the first The first in the final region The average gradient magnitude of all pixels on the edge of the comparison is denoted as the i-th. The first in the final region The gradient mean of the contrasting edges.

[0137] Furthermore, obtain the first The specific formula for calculating the final ambiguity of each final region is as follows:

[0138]

[0139] In the formula, Indicates the first The final ambiguity of the final region, Indicates the first The mean of the motion performance of all contrasting edges within the final region. Indicates the first The average of the gradients of all contrasting edges within the final region. Indicates the first The blurry representation of the final region This represents the sigmoid function. In this example, it is used for normalization.

[0140] It should be noted that, The smaller the value, the better it conforms to the characteristic that when an edge is greatly affected by motion, the gradient magnitude of the pixels on that edge should be smaller. The function is used to prevent large gradient magnitudes at the edges of pixels in regions less affected by motion, so that if directly passed through... right When making adjustments, it will be to The problem is to make a large adjustment to the value, and thus obtain the final blur of each final region.

[0141] Furthermore, let the length of the minimum fuzzy kernel be 3, and obtain the first... The specific formula for calculating the length of the fuzzy kernel of each final region is as follows:

[0142]

[0143] In the formula, Indicates the first The length of the fuzzy kernel of the final region, This represents the minimum value of the final ambiguity among all final regions. This represents the floor function. The length of the blur kernel for each final region is less than or equal to 31. 31 is used as the maximum length of the blur kernel because, generally, the maximum length of the blur kernel is 31 when performing deblurring operations.

[0144] Furthermore, a blur kernel for each final region is generated based on the length and width of the blur kernel for each final region. The method for generating the blur kernel based on its length is a well-known existing technique and will not be described in detail in this embodiment. Furthermore, the length and width of each blur kernel are the same. And the first... The number of pixels within the blur kernel of each final region should be: take indivual.

[0145] Furthermore, the image acquired by the UAV is deblurred based on the blur kernel of each final region, thus completing the image deblurring process. The method of deblurring each region based on the blur kernel used in the deblurring process is a well-known existing technique and will not be described in detail in this embodiment.

[0146] This concludes the embodiment.

[0147] Another embodiment of the present invention provides a dynamic blurry target recognition system for unmanned aerial vehicles (UAVs). The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the above-described method steps S001 to S003.

Claims

1. A method for identifying dynamically blurred targets from unmanned aerial vehicles (UAVs), characterized in that, The method includes the following steps: Use an image acquisition device mounted on the drone to acquire images captured by the drone; Edge detection algorithms are used to detect edges in images captured by drones. Based on these edges, final regions and the edges within those regions are identified. The intersection points of the edges within each final region are used as segmentation points to segment the edges, resulting in contrasting edges within each final region. An 8-connected chain code is used to obtain the chain code for each contrasting edge within each final region. The chain codes of different contrasting edges within each final region are compared to determine the morphological differences between each contrasting edge and other contrasting edges within each final region. Based on these morphological differences, the blur representation of each final region is determined. Based on the gradient magnitude and distribution of pixels on each contrast edge in each final region, and the blurring behavior of each final region, the final blur level of each final region is obtained; based on the final blur level of each final region, the length and width of the blur kernel of each final region are obtained; based on the length and width of the blur kernel of each final region, the blur kernel of each final region is obtained; based on the blur kernel of each final region in the image, the image deblurring process is completed.

2. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 1, characterized in that, The specific calculation steps for obtaining the final region based on the edges in the image, and the edges located within the final region, are as follows: The closed region enclosed by the edge in the image taken by the drone is recorded as the initial region, and all the initial regions in the image taken by the drone are obtained. If the image captured by the drone contains the first If an initial region is not part of any other initial region in the image, it is recorded as the final region. Will be located in the The edges of the final region and the surrounding area The edge of each final region is denoted as the first... The edge of the final region.

3. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 1, characterized in that, The specific calculation steps for comparing the chain codes of different contrast edges in each final region to obtain the morphological differences between each contrast edge and other contrast edges in each final region are as follows: The first The first in the final region The contrast edge and the first By combining the contrasting edges, we obtain the first... The first in the final region The first line of contrasting edges A number of combinations; If the first The first in the final region The first line of contrasting edges If the chain code lengths of the two edges in the combined pair are the same, then the first pair will be... The first in the final region The contrast edge and the first The contrast edges are respectively denoted as the first. The first in the final region The first line of contrasting edges The control edge and the final reference edge in each combination pair; If the first The first in the final region The first line of contrasting edges If the chain code lengths of the two contrasting edges in the first pair are different, then the first pair will be... The first in the final region The contrast edge and the first The longest and shortest chain code lengths of the edges in the comparison edges are denoted as the first and second edges, respectively. The first in the final region The first line of contrasting edges The long edge and the control edge in each combination pair; For the first The first in the final region The first line of contrasting edges By comparing and analyzing the chain codes of the two contrasting edges in the first combination pair, we obtain the first... The first in the final region The first line of contrasting edges The final reference edge in each combination pair; The first The first in the final region The first line of contrasting edges The chain codes of the corresponding edges in each combination pair are compared one by one with the final reference edges to obtain the first... The first in the final region The contrast edge and the first Edge similarity of contrasting edges; According to the The edge similarity of different contrasting edges within each final region is used to obtain the blurred representation of each final region.

4. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 3, characterized in that, The first The first in the final region The first line of contrasting edges By comparing and analyzing the chain codes of the two contrasting edges in the first combination pair, we obtain the first... The first in the final region The first line of contrasting edges The specific steps for calculating the final reference edge in each combination pair are as follows: With the first The first in the final region The first line of contrasting edges The chain code length of the control edge in the combined pair is the window scale, with a step size of 1, starting from the . The first in the final region The first line of contrasting edges Slide the chain code on the long edge of each combination pair from beginning to end, and denote the resulting equal-length chain code segments as the th... The first in the final region The first line of contrasting edges Suspected reference edges in a combination pair; Get the The first in the final region The first line of contrasting edges The first of the combinations The specific formula for calculating the morphological difference between a suspected reference edge and a control edge is as follows: In the formula, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The morphological differences between a suspected reference edge and a control edge. Indicates the first The first in the final region The first line of contrasting edges The edge morphology of the contrast edge in each combination pair Indicates the first The first in the final region The first line of contrasting edges The first of the combinations An edge shape that is suspected to be a reference edge. Indicates the ordinal value. Indicates the first The first in the final region The first line of contrasting edges The number of digits contained in the chain code of the contrast edge in each combination pair. Indicates the first The first in the final region The first line of contrasting edges The numbers on the chain code at the contrast edge of each combination pair Quantity, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations Numbers on a chaincode that appears to be at a reference edge Quantity, Represents the absolute value function; where the first... The first in the final region The first line of contrasting edges The first of the combinations The morphological differences between the suspected reference edge and the control edge are as follows: The first in the final region The first line of contrasting edges The first of the combinations A suspected reference edge, with the first The first in the final region The first line of contrasting edges Morphological differences in the control edges of each combination pair; According to the The first in the final region The first line of contrasting edges The morphological differences between each suspected reference edge and control edge in each combination pair were analyzed using the Otsu method. The first in the final region The first line of contrasting edges Clustering the suspected reference edges in each combination pair yields two clusters; Connect all suspected reference edges in the two clusters with the first... The first in the final region The first line of contrasting edges The cluster containing the lowest mean morphological difference in the control edges among the 10 pairs of combinations is denoted as the 1st cluster. The first in the final region The first line of contrasting edges The final reference edge in each combination pair; If the first The first in the final region The first line of contrasting edges If the number of suspected reference edges in the combined pairs is 2, then the th pair will be... The first in the final region The first line of contrasting edges The suspected reference edges in each of the combined pairs are denoted as the i-th. The first in the final region The first line of contrasting edges The final reference edge in each combination pair.

5. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 3, characterized in that, The first The first in the final region The first line of contrasting edges The chain codes of the corresponding edges in each combination pair are compared one by one with the final reference edges to obtain the first... The first in the final region The contrast edge and the first The specific steps for calculating the edge similarity of the contrasting edges are as follows: If the first The first in the final region The first line of contrasting edges The contrast edge in the first combination pair and the first The first in the chaincode of the final reference edge If the values ​​of the nth numbers are the same, then the nth number... The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The difference is 0 if the values ​​are different; if the values ​​are different, the difference is 1. By using the first chain code in each edge The method of generating a window containing five numbers centered on the nth number yields the nth... The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The comparison window for the number, and the number The first in the final region The first line of contrasting edges The first of the combinations The first in the chain code of the first contrast edge A comparison window for each number; Get the The first in the final region The first line of contrasting edges The first of the combinations The specific formula for calculating the edge similarity between the final reference edge and the control edge is as follows: In the formula, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations Edge similarity between the final reference edge and the control edge. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge The final difference, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The final reference edge and the first contrast edge One difference, Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The number of digits contained in the chaincode of the final reference edge. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The number of digits contained in the comparison window. Indicates the first The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The degree of difference; if the first... The first in the final region The first line of contrasting edges The first of the combinations The first in the chaincode of the final reference edge The first number in the comparison window The number, with the first The first in the final region The first line of contrasting edges The first chain code of the corresponding edge in the first combination pair The first number in the comparison window If the values ​​of the numbers are the same, then ,otherwise ; The first The first in the final region The first line of contrasting edges The maximum value among the edge similarities of all final reference edges and control edges in the combined pairs is denoted as the i-th... The first in the final region The contrast edge and the first Edge similarity of the contrasting edges.

6. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 3, characterized in that, According to the first The specific calculation steps for obtaining the blurred representation of each final region by calculating the edge similarity of different contrasting edges within each final region are as follows: Using Otsu's method to... The first in the final region Clustering is performed on the edge similarity of the single contrasting edge and all contrasting edges to obtain two clusters; Let the edge similarities within the cluster with the largest mean edge similarity between two clusters be denoted as the i-th. The first in the final region The final edge similarity of the contrasting edges; The maximum value of the edge similarity among the clusters with the smallest mean edge similarity is denoted as the i-th cluster. The first final region One difference similarity; If the first The first in the final region The similarity of each final edge of the contrasting edges is greater than that of the first edge. For each difference similarity in the final region, the first... The first in the final region The final edge similarity of the contrasting edges is denoted as the i-th edge. The first in the final region The final edge similarity representation of the contrasting edges, and the first edge The first in the final region A single contrasting edge is denoted as a blurred representation edge; The first The first in the final region The sum of the final edge similarity representations of all contrasting edges, denoted as the i-th. The first in the final region The degree of fuzziness in contrasting edges; The first The mean of the fuzziness of all contrast edges within the final region is denoted as the i-th. The final region is a blurry representation.

7. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 1, characterized in that, The specific steps for calculating the final blur level of each final region based on the gradient magnitude and distribution of pixels on each contrast edge within each final region, and the blur performance of each final region, are as follows: Get the The first in the final region The specific formula for calculating the motion performance of contrasting edges is as follows: In the formula, Indicates the first The first in the final region The motion performance of the contrasting edges Indicates the first The first in the final region The variance of the gradient magnitude of all pixels on the contrast edge. Indicates the first The first in the final region The average gradient magnitude of all pixels on the contrast edge; The first The first in the final region The average gradient magnitude of all pixels on the edge of the comparison is denoted as the i-th. The first in the final region The gradient mean of the contrasting edges; Get the The specific formula for calculating the final ambiguity of each final region is as follows: In the formula, Indicates the first The final ambiguity of the final region, Indicates the first The mean of the motion performance of all contrasting edges within the final region. Let represent the average of the gradients of all contrast edges within the nth final region. Indicates the first The blurry representation of the final region This represents the sigmoid function.

8. The method for identifying dynamic blurred targets from unmanned aerial vehicles according to claim 1, characterized in that, The specific calculation steps for obtaining the length and width of the blur kernel of each final region based on the final blur degree of each final region are as follows: Get the The specific formula for calculating the length of the fuzzy kernel of each final region is as follows: In the formula, Indicates the first The length of the fuzzy kernel of the final region, This represents the minimum value of the final ambiguity among all final regions. This represents the floor function; and the length of the blur kernel for each final region is less than or equal to 31; where, the first... The length and width of the fuzzy kernel of each final region are the same.

9. A dynamic fuzzy target recognition system for unmanned aerial vehicles (UAVs), characterized in that, The system includes the following modules: Image acquisition module: Uses an image acquisition device mounted on the drone to acquire images captured by the drone; Initial blurriness judgment module: Uses an edge detection algorithm to perform edge detection on the image captured by the drone to obtain the edges in the image; Based on the edges in the image, it obtains the final region and the edges located within the final region; The intersection points of the edges in each final region are used as segmentation points to segment the edges and obtain the contrast edges in each final region. The chain code of each contrast edge in each final region is obtained by using 8-connected chain code. The chain codes of different contrast edges in each final region are compared to obtain the morphological differences between each contrast edge in each final region and other contrast edges. Based on the morphological differences between each contrast edge in each final region and other contrast edges, the blur representation of each final region is obtained. Deblurring module: Based on the gradient magnitude and distribution of pixels on each contrast edge in each final region, and the blurring performance of each final region, the final blur level of each final region is obtained; based on the final blur level of each final region, the length and width of the blur kernel of each final region are obtained. The fuzzy kernel for each final region is obtained based on the length and width of the fuzzy kernel for each final region. The image deblurring process is completed based on the blur kernel of each final region within the image.

Images