Image data processing system for fault diagnosis of landing lamp of unmanned aerial vehicle

Abstract

The invention relates to the technical field of image processing, in particular to an image data processing system for unmanned aerial vehicle landing lamp fault diagnosis. According to the system, a first suspected fault area in a light emitting element area is counted by using a brightness value under a fixed gear, brightness changes generated by changes of different gears are analyzed at different heights, and a second suspected fault area is further screened out through distribution of brightness change amplitudes. And analyzing the relevance of the brightness variation amplitudes of different position areas in the second suspected fault area, and further correcting the fault credibility in combination with the distance between the unmanned aerial vehicle and the landing lamp image acquisition equipment at each height. And determining a final fault probability and carrying out accurate fault positioning. According to the invention, each light-emitting element area is effectively analyzed based on factors such as brightness change, adjacent array brightness influence, distance between the image acquisition equipment and the landing lamp and the like, so that an accurate final fault probability is obtained and effective fault positioning is carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to an image data processing system for fault diagnosis of an unmanned aerial vehicle (UAV) landing light. Background Art

[0002] A UAV landing light is a lighting device installed on a UAV, mainly used to provide auxiliary lighting when the UAV lands. It can provide sufficient brightness under various ambient light conditions, illuminate the landing area below the UAV in a darker environment or low visibility conditions, help the operator see the UAV more clearly, and judge the height, speed, etc. of the UAV to ensure the safe and accurate landing of the UAV.

[0003] A UAV landing light generally consists of multiple light-emitting elements. If a certain light-emitting element fails, a brightness significantly different from that of other light-emitting elements will be generated. Therefore, in the prior art, through computer vision technology, by collecting the landing light image and analyzing the brightness difference between the light-emitting element areas, the fault area can be automatically and intelligently located. However, in actual situations, the collected images are affected by factors such as ambient light, light between arrays, and brightness attenuation caused by the distance between the image acquisition device and the landing light. These factors will affect the brightness performance in the image, and further make the fault area located directly based on the brightness difference between the arrays inaccurate. Summary of the Invention

[0004] In order to solve the technical problem that the prior art does not consider factors such as the influence of ambient light on the landing light image, the influence of light between arrays, and the influence of brightness attenuation caused by the distance between the image acquisition device and the landing light, resulting in inaccurate fault areas located directly based on the brightness difference between the arrays, the purpose of the present invention is to provide an image data processing system for fault diagnosis of a UAV landing light, and the specific technical solution adopted is as follows: The present invention proposes an image data processing system for fault diagnosis of a UAV landing light, and the system includes: A first fault analysis module, configured to obtain a landing light image of the UAV landing light; the landing light image includes multiple light-emitting element areas; at all heights, according to the brightness values of the light-emitting element areas under a preset brightness level, a first suspected fault area is statistically obtained; A second fault analysis module, configured to, at each height, change the brightness level, according to the brightness change amplitude of the first suspected fault area before and after the change of the level, and according to the distribution of the brightness change amplitude, screen out a second suspected fault area; divide the second suspected fault area into a central area and a boundary area, and the boundary area is adjacent to other light-emitting element areas; according to the correlation of the brightness change amplitude between the central area and the boundary area when the brightness level changes, obtain the fault credibility of each second suspected fault area at the current height; A fault probability determination module, configured to obtain a distance weight according to the distance between the drone and the landing light image acquisition device at each altitude, and use the distance weight to perform weighted integration on each second suspected fault area at all altitudes to obtain an initial fault probability for each light-emitting element area; according to the number of times the light-emitting element area is determined as a second suspected fault area at all altitudes and the initial fault probability, obtain the final fault probability for each light-emitting element area; A fault location module, configured to screen out the faulty light-emitting element areas according to the final fault probability.

[0005] Further, the screening method for the first suspected fault area includes: At one altitude, obtain the average brightness of each light-emitting element area at a preset brightness level, cluster all the average brightnesses to obtain a plurality of first clustering clusters; for each first clustering cluster, obtain a first normal probability according to the number of light-emitting element areas in the first clustering cluster and the average brightness; select the light-emitting element areas in the first clustering cluster with the largest first normal probability as the first normal light-emitting element areas, and the other light-emitting element areas except the first normal light-emitting element areas as the third suspected fault areas; Statistically analyze all altitudes. If a light-emitting element area is identified as a third suspected fault area at any one altitude, then regard this light-emitting element area as the first suspected fault area.

[0006] Further, the screening method for the second suspected fault area includes: For each brightness level change process, cluster the first suspected fault areas according to the brightness change amplitude to obtain a second clustering cluster; according to the number of first suspected fault areas in the second clustering cluster and the variance of the brightness change amplitude, obtain a second normal probability; select the first suspected fault areas in the second clustering cluster with the largest second normal probability as the second normal light-emitting element areas, and the other first suspected fault areas except the second normal light-emitting element areas as the fourth suspected fault areas; Statistically analyze all brightness level change processes. If a first suspected fault area is judged as a fourth suspected fault area in all brightness level change processes, then regard this first suspected fault area as the second suspected fault area.

[0007] Further, the method for obtaining the fault credibility includes: For each brightness level change process, for any second suspected fault area, if the other light-emitting element areas adjacent to the boundary area are not the second suspected fault area, then the boundary area is used as the boundary area to be analyzed; the difference in the brightness change amplitude between each boundary area to be analyzed and the central area is negatively correlated and normalized to obtain the brightness change amplitude correlation, and the average brightness change amplitude correlation corresponding to all boundary areas to be analyzed is used as the initial fault credibility of the second suspected fault area; if all other light-emitting element areas adjacent to the boundary areas are the second suspected fault areas, then the brightness change amplitude of the central area is negatively correlated and normalized to obtain the initial fault credibility of the second suspected fault area; Count all brightness level change processes, and use the average initial fault credibility of the second suspected fault area in all brightness level change processes as the fault credibility.

[0008] Furthermore, the distance weight is the reciprocal of the distance between the drone and the landing light image acquisition device at each altitude.

[0009] Furthermore, the method for obtaining the initial fault probability includes: For any light-emitting element area, if the light-emitting element area is not judged as the second suspected fault area at each altitude, then set the initial fault probability to 0; otherwise, count all altitudes, and use the distance weight to weighted average the fault credibility corresponding to the light-emitting element area to obtain the initial fault probability.

[0010] Furthermore, the method for obtaining the final fault probability includes: For any light-emitting element area, use the ratio of the number of times the light-emitting element area is judged as the second suspected fault area to the number of altitudes as the adjustment weight; normalize the product of the adjustment weight and the initial fault probability to obtain the final fault probability.

[0011] Furthermore, use the light-emitting element areas with the final fault probability greater than the preset probability threshold as the fault light-emitting element areas.

[0012] Furthermore, the method for dividing the central area and the boundary area includes: The light-emitting element area is a rectangular area; For each second suspected fault area, starting from the boundary pixel points of the second suspected fault area, traverse a preset number of pixel points in the direction of the area center point to obtain the area boundary pixel points and the central area boundary of the central area; each central area boundary and the nearest and parallel second suspected fault area boundary form a boundary area.

[0013] Further, the system further includes a fault degree evaluation module for evaluating the fault degree according to the number of the faulty light-emitting element areas. If the fault degree is greater than a preset fault degree threshold, a maintenance and replacement command is fed back.

[0014] The present invention has the following beneficial effects: The present invention first uses the brightness values at fixed gears to count the first suspected fault areas in the light-emitting element areas. The first suspected fault areas are inaccurate fault screening results affected by multiple factors. Therefore, the present invention further analyzes the brightness changes generated by different gear changes at different heights, and further screens out the second suspected fault areas through the distribution of the brightness change amplitudes. And considering the brightness influence between arrays, the correlation of the brightness change amplitudes of different position areas in the second suspected fault areas is analyzed, and then the fault credibility of the second suspected fault areas at each height is obtained. By combining the distance between the drone and the landing light image acquisition device at each height, the fault credibility is further corrected to obtain the initial fault probability of each light-emitting element area. The final fault probability can be determined by counting the number of times each light-emitting element area is identified as the second suspected fault area in the whole process, and accurate fault location can be performed. The present invention effectively analyzes each light-emitting element area based on factors such as brightness change, adjacent array brightness influence, and the distance between the image acquisition device and the landing light, obtains the accurate final fault probability, and performs effective fault location. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a block diagram of an image data processing system for drone landing light fault diagnosis provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the segmentation of the light-emitting element areas provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the division of a second suspected fault area provided by an embodiment of the present invention. Detailed Embodiments

[0017] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of an image data processing system for UAV landing light fault diagnosis proposed in accordance with the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following describes in detail a specific solution of an image data processing system for UAV landing light fault diagnosis provided by the present invention in conjunction with the accompanying drawings.

[0020] See also Figure 1 , which shows a block diagram of an image data processing system for UAV landing light fault diagnosis provided by an embodiment of the present invention. The system includes: a first fault analysis module 101, a second fault analysis module 102, a fault probability determination module 103 and a fault location module 104.

[0021] Because the embodiment of the present invention aims to analyze the brightness changes of each light-emitting element area during the change of the brightness level of the landing light at different heights by image processing means, and then determine the final failure probability. Therefore, the scenario for the implementation of the present invention can be the working stage of the drone, and the image of the drone in the air is captured by the shooting equipment, and then the landing light image is segmented for processing by the system of the embodiment of the present invention. During the shooting process, it is necessary to ensure that the camera field of view is wide, and to avoid interference from other background factors as much as possible, and the camera parameters of the image acquisition equipment need to ensure that clear and complete details of the landing light can be captured at a certain distance. The landing light in the embodiment of the present invention includes three gears: low, medium, and high. And four heights are set for image acquisition, namely: 3 meters, 7 meters, 9 meters, and 11 meters.

[0022] It should be noted that in the embodiments of the present invention, a pre-trained neural network can be used to segment the landing light image from the image. Specifically, a Mask R-CNN neural network (Mask Region-based Convolutional Neural Network) can be used. The specific network structure and training methods are well known to those skilled in the art and will not be detailed here.

[0023] The first fault analysis module 101 is configured to obtain an image of the landing light of the drone. Usually, the landing light is composed of multiple circular light-emitting elements, so the landing light image includes multiple light-emitting element areas.

[0024] In the embodiment of the present invention, since the light-emitting element is the light source of the landing light, the corresponding brightness value has significant characteristics. The Otsu threshold segmentation algorithm can be used to screen out the highlighted connected areas with larger brightness. Each highlighted connected area represents a light-emitting element. For the convenience of subsequent area analysis, the minimum bounding rectangle of the obtained highlighted connected area is used as the light-emitting element area.

[0025] In another embodiment of the present invention, considering that the landing light is usually a device with regular shape and the light-emitting elements are neatly arranged in the landing light, after obtaining the landing light image, the landing light image can be directly and evenly segmented according to the specifications of the light-emitting elements in the landing light to directly obtain the light-emitting element areas. Please refer to Figure 2 , which shows a schematic diagram of the segmentation of the light-emitting element area provided by an embodiment of the present invention. Since the light-emitting elements are evenly arranged, each light-emitting element area can be directly determined by uniform area division.

[0026] It should be noted that in the embodiment of the present invention, when performing brightness analysis, the gray value in the landing light image is used as the brightness value.

[0027] The embodiment of the present invention aims at the technical problem that the prior art does not consider factors such as the influence of environmental light on the landing light image, the influence of light between arrays, and the influence of brightness attenuation caused by the distance between the image acquisition device and the landing light, resulting in inaccurate positioning of the fault area directly based on the brightness difference between arrays. The embodiment of the present invention gradually improves the screening accuracy through three screenings, and then obtains an accurate fault area. Therefore, in order to avoid missed detection, the first fault analysis module 101 statistically obtains the first suspected fault area according to the brightness values of the light-emitting element areas at a preset brightness level at all heights. That is, compared with the prior art, the first fault analysis module 101 analyzes the brightness of the light-emitting elements at a fixed level at multiple heights, avoiding the inaccurate influence caused by certain factors at a certain height. By statistically analyzing the brightness values of the light-emitting element areas at all heights, the result of the first screening can be obtained. It should be noted that since normal light-emitting elements have fixed and uniform light-emitting characteristics, the generated brightness and color are relatively stable, and the brightness information on the image is relatively consistent, while faulty light-emitting elements will generate significantly abnormal brightness values. Therefore, the first suspected fault areas are all areas with abnormal brightness at a certain height or multiple heights. However, it is uncertain whether these abnormal brightnesses are caused by real component failures and need to be further analyzed in subsequent modules.

[0028] Preferably, in the embodiments of the present invention, the method for screening the first suspected fault area includes: At a certain height, obtain the average brightness of each light-emitting element area at a preset brightness level. That is, the average brightness is the average gray value of the current light-emitting element area. In the embodiments of the present invention, the preset brightness level is set to the medium level.

[0029] At a certain height, an image of the landing light at the preset brightness level will be obtained. The image contains multiple light-emitting element areas, that is, there are multiple average brightness values. If all the light-emitting elements are normal elements, the average brightness will be relatively uniform. Therefore, the embodiments of the present invention adopt a classification strategy for the first abnormal screening.

[0030] Cluster all the average brightness values to obtain multiple first clustering clusters. Since in the array composed of light-emitting elements, the faulty elements are a small probability event, in the first clustering cluster, the more corresponding light-emitting element areas and the greater the average brightness, it indicates that the first clustering cluster is more likely to be a clustering cluster formed by normal light-emitting element areas. Therefore, for each first clustering cluster, according to the number of light-emitting element areas and the average brightness in the first clustering cluster, obtain the first normal probability.

[0031] In the embodiments of the present invention, the average brightness is normalized by dividing it by 255, and the first normal probability can be obtained by multiplying the normalized result by the number of light-emitting element areas.

[0032] In the embodiments of the present invention, the clustering method can be selected as density clustering, which is a well-known technical means for those skilled in the art and will not be elaborated here.

[0033] Select the light-emitting element areas in the first clustering cluster with the largest first normal probability as the first normal light-emitting element areas, and the other light-emitting element areas except the first normal light-emitting element areas are the third suspected fault areas.

[0034] So far, there is a set of third suspected fault areas at each height. The UAV landing light is generally located on the front side of the nose. Therefore, during the shooting process, the camera and the landing light cannot be completely parallel, and there will be a certain angular difference. So the height will affect the acquisition angle of the image acquisition device, and then form a certain light reflection deviation. It is possible that the fault area will be missed at a certain height. Therefore, the embodiments of the present invention count all heights. If a light-emitting element area is identified as a third suspected fault area at any height, then this light-emitting element area is used as the first suspected fault area. That is, select the union of the sets of third suspected fault areas at all heights as the set of first suspected fault areas.

[0035] The second fault analysis module 102 is used to further analyze the first suspected fault area, perform a second screening process and analyze for a third screening. Considering the influence of light reflection, accurate fault areas cannot be obtained only through the information of brightness itself. Therefore, the second fault analysis module 102 changes the brightness level at each height, and further analyzes the brightness change amplitude of the first suspected fault area during the process of changing the brightness level, and then performs the second screening process. Since the normal light-emitting element area operates stably, the brightness change amplitude generated during the process of changing the level should be uniform. Therefore, for all the first suspected fault areas, the more the distribution of the brightness change amplitude before and after the level change deviates from the whole, the more likely it is a fault area. Thus, the second suspected fault areas can be screened out for the second screening process.

[0036] According to the structural characteristics of the landing light, the light-emitting elements can be arranged in an orderly and compact manner in the landing light. Therefore, the brightness of each light-emitting element area will be affected by the brightness of other areas. For example, when a certain element fails, if it exists alone, the brightness of the corresponding area will not change or change little with the adjustment of the brightness level; however, if it is adjacent to a normal element, the adjacent normal element will "supplement light" to it, that is, the brightness of the faulty element will be affected by the adjacent normal element and generate a synchronous brightness change. These brightness changes will affect the judgment of the fault area. Therefore, although the brightness change of the second suspected fault area is analyzed, there is still a risk of inaccuracy. Therefore, further analysis of the second suspected fault area is required.

[0037] The second fault analysis module 102 further divides the second suspected fault area into a central area and a boundary area, and the boundary area is adjacent to other light-emitting element areas. That is, during the process of changing the brightness level, the brightness change of the central area represents the brightness change formed by the elements in the second suspected fault area itself, and the boundary area is the brightness change that may be affected by the adjacent light-emitting element areas. Therefore, by analyzing the correlation of the brightness change amplitudes of the central area and the boundary area, the fault credibility of each second suspected fault area at the current height can be obtained. That is, the greater the correlation of the change amplitudes, the less the current second suspected fault area is affected by the supplementary light of adjacent elements, and the brightness changes in the area are all generated by its own normal level changes, so the fault credibility is smaller; the smaller the correlation of the change amplitudes, the greater the influence of the boundary area by the supplementary light of adjacent elements, and the central element area has a small change or no change due to the fault, so a large change amplitude difference is generated, and the fault credibility of the second suspected fault area is greater.

[0038] It should be noted that in the embodiment of the present invention, since there are three brightness levels, two brightness level change processes are set, that is, from the low level to the middle level, and from the middle level to the high level.

[0039] In an embodiment of the present invention, the amplitude of the brightness change of a region is the absolute value of the difference between the average gray values of the region before and after the change of the brightness level.

[0040] In another embodiment of the present invention, the amplitude of the brightness change is the absolute value of the difference between the lowest brightness value of the low-level brightness level and the highest brightness value of the high-level brightness level.

[0041] Preferably, in an embodiment of the present invention, the method for screening the second suspected fault region includes: Similar to the method for screening the first suspected fault region in an embodiment of the present invention above, since normal components have a unified amplitude of brightness change during the change of the brightness level, the method of clustering classification can also be used for the second screening.

[0042] For each process of changing the brightness level, cluster the first suspected fault regions according to the amplitude of the brightness change to obtain the second cluster. For the second cluster, the more the number of the first suspected fault regions it contains and the more unified the distribution of the amplitude of the brightness change, the more likely it is that the cluster is formed by the normal component regions. Therefore, according to the number of the first suspected fault regions in the second cluster and the variance of the amplitude of the brightness change, obtain the second normal probability.

[0043] In an embodiment of the present invention, the ratio of the number of the first suspected fault regions to the variance of the amplitude of the brightness change is used as the second normal probability of the second cluster. The smaller the variance, the more unified the amplitude of the brightness change, and the greater the second normal probability.

[0044] Select the first suspected fault regions in the second cluster with the largest second normal probability as the second normal light-emitting component regions, and the other first suspected fault regions except the second normal light-emitting component regions as the fourth suspected fault regions; Since the landing light includes multiple levels, there are multiple processes of changing the brightness level. Count all the processes of changing the brightness level. If a first suspected fault region is judged as a fourth suspected fault region in all the processes of changing the brightness level, then regard this first suspected fault region as the second suspected fault region. That is, each process of changing the brightness level corresponds to a set of fourth suspected fault regions, and the intersection of all the sets of fourth suspected fault regions is used as the set of second suspected fault regions.

[0045] Preferably, in an embodiment of the present invention, since the light-emitting element region is rectangular, for each second suspected fault region, starting from the boundary pixel points of the second suspected fault region, a preset number of pixel points are traversed in the direction of the region center point to obtain the region boundary pixel points and the center region boundary of the central region; each center region boundary and the nearest and parallel second suspected fault region boundary form one of the boundary regions. Please refer to Figure 3 , which shows a schematic diagram of the division of a second suspected fault region provided by an embodiment of the present invention. The shaded region in the figure is the central region. The extension lines of the four boundaries of the central region and the nearest and parallel second suspected fault region boundary form a boundary region, that is, the four boundary regions are of the same size, and there are parts of the same size overlapping at the four corners.

[0046] Preferably, in an embodiment of the present invention, the method for obtaining the fault credibility includes: For each brightness level change process, for any second suspected fault region, if the other light-emitting element regions adjacent to the boundary region are not the second suspected fault region, that is, the other adjacent light-emitting element regions are normal element regions determined after two screenings, then the boundary region is used as the boundary region to be analyzed. The difference in the brightness change amplitude between each boundary region to be analyzed and the central region is negatively correlated and normalized to obtain the brightness change amplitude correlation. That is, the smaller the difference, the closer the brightness change amplitudes between the central region and the boundary region to be analyzed, and the greater the brightness change amplitude correlation. Since there may be multiple boundary regions to be analyzed, the average brightness change amplitude correlation corresponding to all boundary regions to be analyzed is used as the initial fault credibility of the second suspected fault region; If all the other light-emitting element regions adjacent to the boundary regions are the second suspected fault regions, it can be considered that the boundary regions may not be affected. At this time, the brightness change amplitude of the central region can be directly analyzed. The smaller the brightness change amplitude, the more likely it is a faulty element, and the greater the initial fault credibility. Therefore, the brightness change amplitude of the central region is negatively correlated and normalized to obtain the initial fault credibility of the second suspected fault region.

[0047] It should be noted that the negative correlation mapping method in the embodiment of the present invention can be implemented by first normalizing and then subtracting the normalized result from the positive integer 1. The normalization can be implemented by multiple existing technologies such as linear normalization and function mapping method.

[0048] Since there are multiple brightness level change processes at one height, all brightness level change processes are statistically analyzed, and the average initial fault credibility of the second suspected fault region in all brightness level change processes is used as the fault credibility.

[0049] After being screened and analyzed by the second fault analysis module 102, each second suspected fault area corresponds to a fault credibility at each altitude. However, since the shooting distance affects the image quality, the lower the altitude of the UAV landing light, the higher the confidence of the extracted image features should be. Therefore, the fault probability determination module 103 further obtains the distance weight according to the distance between the UAV and the landing light image acquisition device at each altitude. That is, the confidence of the extracted features is characterized by the distance weight, and the distance weight can be used to weightedly integrate each second suspected fault area at all altitudes, and then the initial fault probability of each light-emitting element area can be obtained. Combining the number of times each light-emitting element area is judged as a second suspected fault area under the entire process at all altitudes, the final fault probability of each light-emitting element area can be obtained. That is, the more times it is judged as a second suspected fault area, the more likely it is to be a real fault, and the greater the final fault probability.

[0050] Preferably, in an embodiment of the present invention, the distance weight is the reciprocal of the distance between the UAV and the landing light image acquisition device at each altitude. Then the further method for obtaining the initial fault probability is as follows: For any light-emitting element area, if the light-emitting element area is not judged as a second suspected fault area at each altitude, the initial fault probability is set to 0; otherwise, count all altitudes, and weightedly average the fault credibility corresponding to the light-emitting element area using the distance weight to obtain the initial fault probability. For example, in the embodiment of the present invention, there are a total of four altitude levels: 3 meters, 7 meters, 9 meters, and 11 meters. After screening and judgment, a certain light-emitting element area is identified as a second suspected fault area at the altitudes of 3 meters and 7 meters. Then, one-third and one-seventh are used as the distance weights respectively, multiplied by the fault credibility at the corresponding altitude to achieve weighting; further, add the weighted results and divide by 2 to achieve averaging, and obtain the initial fault probability.

[0051] Preferably, in the embodiment of the present invention, the method for obtaining the final fault probability includes: For any light-emitting element area, the ratio of the number of times the light-emitting element area is judged as a second suspected fault area to the number of altitudes is used as the adjustment weight. That is, the larger the adjustment weight, the more times the light-emitting element area is identified as a second suspected fault area, indicating that the corresponding final fault probability should be greater.

[0052] Normalize the product of the adjustment weight and the initial fault probability to obtain the final fault probability.

[0053] The normalization processing in the embodiment of the present invention can be implemented by using existing technologies such as linear normalization and function mapping method, which will not be elaborated and limited here.

[0054] In the embodiments of the present invention, the light-emitting element area with a final failure probability greater than the preset probability threshold is used as the failed light-emitting element area. Since the final failure probability is normalized in an embodiment of the present invention, the probability threshold is set to 0.6.

[0055] Preferably, in an embodiment of the present invention, the image data processing system further includes a failure degree evaluation module for evaluating the failure degree according to the number of the failed light-emitting element areas. If the failure degree is greater than the preset failure degree threshold, a maintenance and replacement command is fed back. In an embodiment of the present invention, the failure degree may be the proportion of the number of the failed light-emitting element areas in all the light-emitting element areas, and the failure degree threshold is set to 0.5. If the failure degree is not greater than the preset failure degree threshold, the failure location result is fed back to the terminal.

[0056] In summary, in the embodiments of the present invention, the first suspected failure area in the light-emitting element area is counted by using the brightness value in the fixed gear. Further, the brightness change generated by the change of different gears at different heights is analyzed, and the second suspected failure area is further screened out through the distribution of the brightness change amplitude. The correlation of the brightness change amplitude of different position areas in the second suspected failure area is analyzed, and then the failure credibility of the second suspected failure area at each height is obtained. Combining the distance between the drone and the landing light image acquisition device at each height further corrects the failure credibility, obtains the initial failure probability of each light-emitting element area, and determines the final failure probability by counting the number of times each light-emitting element area is identified as the second suspected failure area in the whole process, so as to perform accurate failure location. The present invention effectively analyzes each light-emitting element area based on factors such as brightness change, adjacent array brightness influence, and the distance between the image acquisition device and the landing light, obtains the accurate final failure probability, and performs effective failure location.

[0057] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the advantages or disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0058] Each embodiment in this specification is described in a progressive manner, and the same or similar parts among the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light, characterized in that The system includes: A first fault analysis module, configured to obtain a landing light image of a drone landing light; the landing light image includes a plurality of light-emitting element areas; at all altitudes, according to the brightness values of the light-emitting element areas in a preset brightness level, a first suspected fault area is statistically obtained; A second fault analysis module, configured to, at each altitude, change the brightness level, according to the brightness change amplitude of the first suspected fault area before and after the change of the level, and according to the distribution of the brightness change amplitude, screen out a second suspected fault area; divide the second suspected fault area into a central area and a boundary area, and the boundary area is adjacent to other light-emitting element areas; according to the correlation of the brightness change amplitudes of the central area and the boundary area when the brightness level changes, obtain the fault credibility of each second suspected fault area at the current altitude; A fault probability determination module, configured to obtain a distance weight according to the distance between the drone and the landing light image acquisition device at each altitude, and use the distance weight to perform weighted integration on each second suspected fault area at all altitudes to obtain an initial fault probability of each light-emitting element area; according to the number of times that the light-emitting element area is determined as the second suspected fault area at all altitudes and the initial fault probability, obtain the final fault probability of each light-emitting element area; A fault location module, configured to screen out the faulty light-emitting element area according to the final fault probability.

2. The image data processing system for drone landing light fault diagnosis according to claim 1, characterized in that, The screening method of the first suspected fault area includes: At one altitude, obtain the average brightness of each light-emitting element area in a preset brightness level, cluster all the average brightnesses to obtain a plurality of first clustering clusters; for each first clustering cluster, according to the number of light-emitting element areas in the first clustering cluster and the average brightness, obtain a first normal probability; select the light-emitting element areas in the first clustering cluster with the largest first normal probability as the first normal light-emitting element areas, and the other light-emitting element areas except the first normal light-emitting element areas as the third suspected fault areas; Statistically count all altitudes, and if a light-emitting element area is determined as the third suspected fault area at any altitude, then use this light-emitting element area as the first suspected fault area.

3. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light according to claim 1, wherein, The screening method of the second suspected fault area includes: For each brightness level change process, cluster the first suspected fault areas according to the brightness change amplitude to obtain a second clustering cluster; according to the number of first suspected fault areas in the second clustering cluster and the variance of the brightness change amplitude, obtain a second normal probability; select the first suspected fault areas in the second clustering cluster with the largest second normal probability as the second normal light-emitting element areas, and the other first suspected fault areas except the second normal light-emitting element areas as the fourth suspected fault areas; Statistically count all brightness level change processes, and if a first suspected fault area is determined as the fourth suspected fault area in all brightness level change processes, then use this first suspected fault area as the second suspected fault area.

4. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light according to claim 1, characterized in that, The method for obtaining the fault credibility includes: For each brightness level change process, for any second suspected fault area, if the other luminous element areas adjacent to the boundary area are not the second suspected fault area, then the boundary area is used as the boundary area to be analyzed; the difference in the brightness change amplitude between each boundary area to be analyzed and the central area is negatively correlated and normalized to obtain the brightness change amplitude correlation, and the average brightness change amplitude correlation corresponding to all boundary areas to be analyzed is used as the initial fault credibility of the second suspected fault area; if the other luminous element areas adjacent to all boundary areas are the second suspected fault area, then the brightness change amplitude of the central area is negatively correlated and normalized to obtain the initial fault credibility of the second suspected fault area. Count all brightness level change processes, and use the average initial fault credibility of the second suspected fault area in all brightness level change processes as the fault credibility.

5. An image data processing system for drone landing light fault diagnosis according to claim 1, characterized in that, The distance weight is the reciprocal of the distance between the drone and the landing light image acquisition device at each height.

6. The image data processing system for UAV landing light fault diagnosis according to claim 5, wherein The method for obtaining the initial fault probability includes: For any luminous element area, if the luminous element area is not judged as the second suspected fault area at each height, then set the initial fault probability to 0; otherwise, count all heights, and use the distance weight to weighted average the fault credibility corresponding to the luminous element area to obtain the initial fault probability.

7. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light according to claim 1, characterized in that, The method for obtaining the final fault probability includes: For any luminous element area, use the ratio of the number of times the luminous element area is judged as the second suspected fault area to the number of heights as the adjustment weight; normalize the product of the adjustment weight and the initial fault probability to obtain the final fault probability.

8. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light according to claim 1, characterized in that, Use the luminous element area with the final fault probability greater than the preset probability threshold as the fault luminous element area.

9. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light according to claim 1, characterized in that, The division method of the central area and the boundary area includes: The luminous element area is a rectangular area. For each second suspected fault area, starting from the boundary pixel points of the second suspected fault area, traverse a preset number of pixel points in the direction of the area center point to obtain the area boundary pixel points and the central area boundary of the central area; each central area boundary and the nearest and parallel second suspected fault area boundary form a boundary area.

10. An image data processing system for fault diagnosis of an unmanned aerial vehicle landing light according to claim 1, characterized in that, The system further includes a fault degree evaluation module, which is used to evaluate the fault degree according to the number of the fault luminous element areas. If the fault degree is greater than the preset fault degree threshold, then feedback a maintenance and replacement command.

Images