Airport PAPI light anomaly identification and detection method based on aerial video processing and analysis

Abstract

The invention discloses an airport PAPI light anomaly recognition and detection method based on aerial video processing and analysis. The method comprises the steps that a video image of an airport PAPI light area is shot and input into a YOLOV8 target detector for target detection, and a ByteTrack algorithm is used to obtain a target tracking trajectory; calculating the overall mean value A of the PAPI lamps, the distance between the adjacent PAPI lamps and the number mean value of the small lamps in the whole-process monitoring data set, and judging that the brightness of the PAPI lamps is too low, the PAPI lamps are not turned on and the small lamps are not turned on; and constructing a PAPI lamp flicker identification model, performing flicker frequency characteristic identification of each PAPI lamp by using the whole-process monitoring data set, solving a characteristic center of all the PAPI lamps, calculating an Euclidean distance between the flicker frequency characteristic of each PAPI lamp and the characteristic center, and if the Euclidean distance is greater than a flicker characteristic threshold value, judging that flicker is abnormal and giving an alarm prompt. According to the method, the track information of the PAPI lamp in the video is calculated by using the YOLOV8 detector and the ByteTrack algorithm, four abnormal states of the PAPI lamp are detected, and the method has the advantages of high accuracy, high reliability and the like.

Description

Technical Field

[0001] The present invention relates to the field of airport lighting anomaly detection, and in particular to an airport PAPI lighting anomaly recognition and detection method based on aerial video processing and analysis. Background Art

[0002] In the field of modern aviation, the operational safety of airports is of vital importance; the navigation lighting system, especially the precision approach path indicator (PAPI light, usually, each PAPI light contains three to six bulbs side by side) provides visual reference to pilots to help them maintain the correct glide angle during the approach. Abnormal detection of PAPI lights is an important technical means and step to ensure the stable operation of PAPI lights. It is one of the keys to ensure the stable landing of aircraft and plays an important role in the safety of airports. Specifically, abnormal detection of PAPI lights refers to the use of technical means to detect PAPI lights that are in poor working condition, including the following four abnormal conditions: abnormal PAPI light brightness (including too high brightness and too low brightness), PAPI lights are not on, some PAPI bulbs are not on, and PAPI lights have abnormal flashing frequency.

[0003] Traditional PAPI lighting anomaly detection mainly relies on manual regular inspections, which is not only inefficient, but also difficult to detect lighting anomalies in real time. Due to the long intervals between manual inspections, some potential faults may pose a threat to flight safety before they are discovered. Driven by the development of electronic technology, some automatic detection systems based on sensors and monitoring equipment have gradually been introduced. Such technical solutions usually use optical sensors or photodetectors to monitor parameters such as light brightness and color. However, these early automatic detection systems have many limitations. For example, they are more sensitive to interference from environmental factors, such as dust and electromagnetic interference, which may lead to false alarms or missed alarms. In addition, existing detection methods have limitations in system data integration and analysis and scalability. The data collected by some systems lack effective integration and analysis, and it is impossible to extract valuable information from a large amount of data, making it difficult to accurately determine the cause of lighting anomalies. In addition, existing general detection methods are difficult to fully adapt to various actual situations, resulting in a significant reduction in the accuracy and reliability of detection, and the high cost of solution deployment. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problems pointed out by the background technology, and to provide an airport PAPI light anomaly recognition and detection method based on aerial video processing and analysis, to construct a light label data set to train the YOLOV8 target detector for target detection, and the trained YOLOV8 target detector can realize the target detection of PAPI lights and small lights in PAPI lights, and the ByteTrack algorithm associates and attributes the PAPI lights and the small lights in the PAPI lights in a hierarchical manner to form target tracking trajectory data of each PAPI light and the internal small lights in a hierarchical manner; then, the target tracking trajectory data is subjected to target traversal analysis to obtain a full-process monitoring data set, and based on the calculation, analysis and judgment of the full-process monitoring data set, the anomaly detection of PAPI lights and small lights is realized, which provides convenient PAPI light anomaly detection technical support for civil airports and promotes the improvement of the digital level of airport security work.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A method for identifying and detecting abnormal airport PAPI lighting based on aerial video processing and analysis, the method comprising:

[0007] S1. Construct a light label dataset containing video images of the airport PAPI light area and PAPI light label annotations, and input the light label dataset into the YOLOV8 target detector for target detection training. The targets include PAPI lights and small lights in PAPI lights.

[0008] S2, shoot the video image of the airport PAPI light area and input it into the YOLOV8 target detector for target detection and use the ByteTrack algorithm to associate the target data and obtain the target tracking trajectory;

[0009] S3, traversing each frame of target tracking trajectory to form a full process monitoring data set of the target;

[0010] S4. Calculate the overall mean A of the historical brightness average value sequence of the PAPI lamp in the whole process monitoring data set. If the overall mean A is less than the low brightness threshold, it is determined that the brightness of the PAPI lamp is too low and an alarm is issued;

[0011] S5. Calculate the distance between two adjacent PAPI lights in the whole process monitoring data set. If the distance is greater than the light interval threshold, it is determined that there is a completely unlit PAPI light between the two adjacent PAPI lights and an alarm is issued;

[0012] S6. Calculate the number mean of the historical number sequence of small lights in the PAPI lamp in the whole process monitoring data set. If the number mean is less than the small light number threshold, it is determined that there are small lights that are not lit in the PAPI lamp and an alarm is issued;

[0013] S7. Construct a PAPI lamp flicker recognition model and use the whole process monitoring data set to identify the flicker frequency characteristics of each PAPI lamp and find the characteristic center of all PAPI lamps. Calculate the Euclidean distance between the flicker frequency characteristics of each PAPI lamp and the characteristic center. If the Euclidean distance is greater than the flicker characteristic threshold, the corresponding PAPI lamp is judged as flickering abnormality and an alarm is issued.

[0014] In order to better implement the present invention, the image frames of the video image are collected according to the time frame of T1; the PAPI light labeling method is as follows: the area boxes of the belonging levels are marked according to the PAPI lights and the small lights in the PAPI lights, and the PAPI lights with too low brightness and too high brightness are marked respectively.

[0015] Preferably, the YOLOV8 target detector performs frame-by-frame target detection on the video image stream F(t), where t represents the sequential number of the video frames; the ByteTrack algorithm performs data association and target tracking based on the targets of the previous and next frames being the PAPI lights, and the ByteTrack algorithm performs data association and target tracking based on the targets of the previous and next frames being the small lights in the PAPI lights, and the small lights in the PAPI lights are hierarchically associated and belong to the corresponding PAPI lights.

[0016] Preferably, the full-process monitoring data set stores target detection data and trajectory data of n PAPI lights and m small lights attributable to each level; wherein the i-th PAPI light sequence expression of the n PAPI lights is S(i) = {U(i), X(i), Y(i), Z(i)}, U(i) is the historical trajectory coordinate center position sequence of the i-th PAPI light, X(i) is the historical number sequence of small lights contained in the i-th PAPI light, Y(i) is the historical brightness average value sequence of the i-th PAPI light area, and Z(i) is the missed detection video frame position sequence in which no PAPI light label is detected in the trajectory of the i-th PAPI light; the n PAPI lights are respectively organized into a full-process monitoring data set in chronological order.

[0017] Preferably, in step S4, if the overall mean A is greater than the high brightness threshold, the PAPI lamp brightness is determined to be too high and an alarm is issued; the PAPI lamp brightness being too low and the PAPI lamp brightness being too high are collectively referred to as abnormal PAPI lamp brightness.

[0018] Preferably, in step S5, the PAPI light information between two adjacent PAPI lights is determined to be output along with the alarm prompt.

[0019] Preferably, the PAPI light flicker recognition model extracts the difference sequence between the missed video frame sequence and the adjacent non-missed video frame sequence in the full process monitoring data set, calculates its mean and variance as the flicker frequency feature, and the adjacent non-missed video frame sequence is the PAPI light target trajectory data that does not belong to the missed video frame sequence and is located before and after in time sequence.

[0020] Preferably, the characteristic center Ra of the characteristic center is expressed as: Where R(i) represents the flicker frequency characteristic of the i-th PAPI lamp, and n represents the total number of PAPI lamps in the whole process monitoring data set.

[0021] Preferably, the video image of the airport PAPI light area is captured by a drone flying from the far channel end to the near channel end along the arrangement direction of the PAPI light group according to the aerial photography route and the aerial photography altitude set to H1; the video images in the light label data set cover all weather scenes and all light states of the PAPI light group, and all light states of the PAPI light group include that all PAPI light groups are normal, the brightness of the PAPI lights is too low, the brightness of the PAPI lights is too high, the PAPI lights are completely off, some small lights in the PAPI lights are off, and the PAPI lights are flashing abnormally.

[0022] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0023] (1) The present invention constructs a light label data set to train the YOLOV8 target detector for target detection. The trained YOLOV8 target detector can realize the target detection of PAPI lights and small lights in PAPI lights. The ByteTrack algorithm associates and attributes PAPI lights and small lights in PAPI lights in a hierarchical manner to form target tracking trajectory data of each PAPI light and the internal small lights in a hierarchical manner; then, the target traversal analysis is performed on the target tracking trajectory data to obtain a full-process monitoring data set. Based on the calculation, analysis and judgment of the full-process monitoring data set, anomaly detection of PAPI lights and small lights is realized, providing convenient PAPI light anomaly detection technical support for civil airports, and promoting the improvement of the digital level of airport security work.

[0024] (2) The present invention performs PAPI light recognition detection, tracking and analysis based on video image capture of the airport PAPI light area, and achieves the purpose of detecting abnormalities such as abnormal PAPI light brightness, PAPI light not lighting, partial PAPI bulb not lighting, abnormal PAPI light flickering frequency, etc. It does not rely on the light hardware sensor data, and greatly reduces the labor cost and the hardware cost of the sensor deployment cost.

[0025] (3) The present invention uses the YOLOV8 detector and ByteTrack algorithm to calculate the trajectory information of the PAPI light in the video, including the target historical trajectory, average brightness, number of small lights, and flashing frequency. The PAPI light trajectory information analysis based on target detection and data association can simultaneously detect four abnormal states of the PAPI light, with the advantages of high accuracy and high reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0027] Figure 2 Schematic diagram of the YOLOV8 target detector identifying the PAPI light in the embodiment. DETAILED DESCRIPTION

[0028] The present invention is further described in detail below in conjunction with embodiments:

[0029] Example

[0030] like Figure 1 As shown, a method for identifying and detecting anomalies of airport PAPI lights based on aerial video processing and analysis includes:

[0031] S1. Construct a light label dataset containing video images of the airport PAPI light area and PAPI light label annotations, and input the light label dataset into the YOLOV8 target detector for target detection training. The targets include PAPI lights and small lights in PAPI lights. The video images of the airport PAPI light area are taken by setting an aerial route (the route height is 20 meters) with a drone from the far channel end to the near channel end along the arrangement direction of the PAPI light group. Aerial videos under different PAPI light states are collected on this aerial route to ensure that the scenes of the aerial videos cover various scenes (different weather environments or different electromagnetic environments or different working scenes) and different light states of all PAPI light groups (including too dark and too bright states based on brightness grading). Finally, image frames are sampled every 1 second from the aerial video to form a light label dataset D of the PAPI light. The PAPI light labeling method is as follows: according to the PAPI light and the small lights in the PAPI light, the area boxes of the belonging levels are respectively labeled (in order to achieve high-precision PAPI light anomaly detection, each complete PAPI light area in the video image and the light area of ​​each small light in the PAPI light must be labeled), and the PAPI light brightness that is too low and too high are also labeled respectively.

[0032] In some embodiments, the video image of the airport PAPI light area is captured by a drone flying from the far channel end to the near channel end along the arrangement direction of the PAPI light group according to the aerial photography route and the aerial photography height is set to H1 (the height is generally set to 18-26 meters). The video images in the light label data set of the present invention cover all weather scenes and all light states of the PAPI light group. All light states of the PAPI light group include that all PAPI light groups are normal, the brightness of the PAPI lights is too low, the brightness of the PAPI lights is too high, the PAPI lights are completely off, some of the small lights in the PAPI lights are off, and the PAPI lights are flashing abnormally.

[0033] S2. Shoot the video image of the airport PAPI light area and input it into the YOLOV8 target detector for target detection (the image frames of the video image are collected at a time interval of T1) and use the ByteTrack algorithm to associate the target data and obtain the target tracking trajectory. Label the complete PAPI light area and label the PAPI light area with normal brightness as papi_normal (such as Figure 2 As shown in the figure, the schematic diagram of the target recognition of PAPI lights by the YOLOV8 target detector) marks the PAPI light area with too low brightness as papi_low, and the PAPI light area with too high brightness as papi_high. The light area of ​​each small light of each PAPI light is labeled and marked as papi_small. After each video image is labeled according to the above rules, the labeled data L is formed (the labeled data L is part of the light label data set D), and the YOLOV8 target detector uses the labeled data L (including the labeled data L) for training. In the actual scenario involved in this embodiment, the number of small lights contained in each PAPI light is 5.

[0034] S3. Traverse each frame of target tracking trajectory to form a full-process monitoring data set of the target. In this embodiment, the YOLOV8 target detector performs frame-by-frame target detection on the video image stream F(t), and t represents the sequential number of the video frame. The ByteTrack algorithm performs data association and target tracking based on the target of the previous and next frames as the PAPI light. The ByteTrack algorithm performs data association and target tracking based on the target of the previous and next frames as the small lights in the PAPI light. The small lights in the PAPI light are hierarchically associated and attributed to the corresponding PAPI light (the ByteTrack algorithm is based on the small light targets of the previous and next frames for association, and then the small lights are attributed to the PAPI light according to their attribution level, and the PAPI light targets are associated with the previous and next frames at the same time, thereby obtaining the target association of the previous and next frames of the PAPI light, and the small lights inside the PAPI light are also associated with the previous and next frames). The ByteTrack algorithm is based on the small lights of the previous and next frames and the targets of the PAPI light for target association according to the hierarchical attribution, and obtains the trajectory data of the PAPI light as the target association and the trajectory data of the small lights in the PAPI light as the target association.

[0035] In some embodiments, the whole process monitoring data set stores target detection data and trajectory data of n PAPI lights and m small lights belonging to different levels; wherein the ith PAPI light sequence expression of the n PAPI lights is S(i)={U(i), V(i), X(i), Y(i), Z(i)}, U(i) is the historical trajectory coordinate center position sequence of the ith PAPI light, V(i) is the first time that the horizontal and vertical coordinates of the ith PAPI light area are greater than the horizontal and vertical coordinate threshold Thr (the coordinate threshold Thr in the video frame of size 1920*1080 is set to 800). Under normal circumstances, the n PAPI lights will be on, that is, the 1st to the nth PAPI lights will be on. If the i PAPI lights are not on, it will be detected that the i+1th PAPI light is on, and the i+1th PAPI light will be detected. The PAPI light will be recorded as the ith PAPI light. In this case, the horizontal and vertical coordinates of the ith PAPI light will be greater than the horizontal and vertical coordinate threshold Thr. At this time, it is the ith PAPI light counted by the computer. After the PAPI light target association is performed in the ByteTrack algorithm, it will be corrected to the i+1th PAPI light. V(i) will record the video frame position of the ith PAPI light. The horizontal and vertical coordinates are the horizontal and vertical coordinates corresponding to the coordinate system constructed by the video frame image. X(i) is the historical number sequence of small lights contained in the ith PAPI light, Y(i) is the historical brightness average sequence of the ith PAPI light area, and Z(i) is the missed video frame position sequence in which no PAPI light label is detected in the trajectory of the ith PAPI light; the n PAPI lights are respectively sorted into a full-process monitoring data set in time sequence. In this embodiment, the example method is as follows: for the detection and tracking results in the t-th frame image F(t) of the aerial video of the PAPI light group, the target trajectory information T(t, i) of each target as a PAPI light is analyzed, and the temporary full-process monitoring data corresponding to its tracking label trk is associated. For each T(t, i) of each frame image, add the center coordinate position of its current position to the center position sequence of the historical trajectory coordinates of the corresponding temporary full-process monitoring data. For each T(t, i) of each frame image, check the horizontal and vertical coordinates of the center of its current position. If the horizontal and vertical coordinates are greater than the horizontal and vertical coordinate threshold Thr and its tracking label trk is not recorded, then set the V(i) in the corresponding temporary full-process monitoring data to the current frame position t, and record the label trk. For each T(t, i) of each frame image, calculate the IoU (intersection over union) of the current area of ​​the T(t, i) and the current area of ​​all small light trajectories in the current frame, and add the number of small lights with an IoU greater than 0.5 to the small light historical number sequence X(i) in the temporary full-process monitoring data corresponding to T(t, i). For each T(t, i) of each frame image, calculate the regional brightness average of the T(t, i) and add it to the regional historical brightness average sequence Y(i) in the temporary full-process monitoring data corresponding to the trajectory.For each T(t, i), if the difference between the current frame position t and the last detected frame position is greater than 1, and the difference between the current frame position t and the last lost frame position is greater than 20, then the current frame position t is added to the missed frame position sequence Z(i) in the temporary full-process monitoring data corresponding to the trajectory. After all video frames are analyzed, the temporary full-process monitoring data is updated to the full-process monitoring data set S. All temporary full-process monitoring data are added to the full-process monitoring data set S in order of tracking numbers from small to large, forming the full-process monitoring data set S of all PAPI lights.

[0036] S4. Calculate the overall mean A of the historical brightness average sequence of the PAPI lamp in the whole process monitoring data set. If the overall mean A is less than the low brightness threshold thr_low (the low brightness threshold thr_low can be derived from the set value, or from the historical reference threshold obtained by statistical learning of historical data from the light tag data set), then determine that the PAPI lamp brightness is too low and issue an alarm prompt. Preferably, in some embodiments, it is necessary to judge both the PAPI lamp brightness being too low and the PAPI lamp brightness being too high. After the judgment is completed, turn to method S5. The method for judging the PAPI lamp brightness being too high is as follows: if the overall mean A is greater than the high brightness threshold thr_high (the high brightness threshold thr_high can be derived from the set value, or from the historical reference threshold obtained by statistical learning of historical data from the light tag data set), then determine that the PAPI lamp brightness is too high and issue an alarm prompt. In the present invention, the PAPI lamp brightness being too low and the PAPI lamp brightness being too high are collectively referred to as PAPI light brightness abnormality, and the alarm prompt is the PAPI light brightness abnormality, and further specifically outputs an alarm prompt of the PAPI lamp brightness being too low or the PAPI lamp brightness being too high.

[0037] S5. Calculate the distance between two adjacent PAPI lights in the whole process monitoring data set. If the distance is greater than the light interval threshold thr_adj (the light interval threshold thr_adj can be derived from a set value or a historical reference threshold obtained by statistical learning of historical data of a light tag data set), it is determined that there is a completely unlit PAPI light between the two adjacent PAPI lights and an alarm is issued. The present invention outputs the PAPI light information between the two adjacent PAPI lights along with the alarm.

[0038] S6. Calculate the mean value of the historical number sequence of small lights in the PAPI lamp in the whole process monitoring data set. If the mean value is less than the small light number threshold thr_small (the small light number threshold can be derived from the set value or the historical reference threshold obtained by statistical learning of historical data from the light label data set), it is determined that there are unlit small lights in the PAPI lamp and an alarm is issued.

[0039] S7. Construct a PAPI lamp flicker recognition model. Use the whole process monitoring data set to identify the flicker frequency characteristics of each PAPI lamp and find the characteristic center of all PAPI lamps. The characteristic center Ra expression is: Where R(i) represents the flicker frequency characteristic of the i-th PAPI lamp, and n represents the total number of PAPI lamps in the whole process monitoring data set.

[0040] The Euclidean distance between the flicker frequency characteristics and the characteristic center of each PAPI light is calculated. If the Euclidean distance is greater than the flicker characteristic threshold thr_flash (the flicker characteristic threshold can be derived from a set value or obtained by deep learning of a model based on a light label dataset), the corresponding PAPI light is judged to be flickering abnormally and an alarm is issued.

[0041] In some embodiments, the PAPI light flicker recognition model extracts the difference sequence between the missed video frame sequence and the adjacent non-missed video frame sequence in the full process monitoring data set, and calculates its mean and variance as the flicker frequency feature. The adjacent non-missed video frame sequence is the PAPI light target trajectory data that does not belong to the missed video frame sequence and is located before and after in time sequence.

[0042] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying and detecting abnormal airport PAPI lighting based on aerial video processing and analysis, characterized in that: The methods include: S1. Construct a light label dataset containing video images of the airport PAPI light area and PAPI light label annotations, and input the light label dataset into the YOLOV8 target detector for target detection training. The targets include PAPI lights and small lights in PAPI lights. S2, shoot the video image of the airport PAPI light area and input it into the YOLOV8 target detector for target detection and use the ByteTrack algorithm to associate the target data and obtain the target tracking trajectory; S3, traversing each frame of target tracking trajectory to form a full process monitoring data set of the target; S4. Calculate the overall mean A of the historical brightness average value sequence of the PAPI lamp in the whole process monitoring data set. If the overall mean A is less than the low brightness threshold, it is determined that the brightness of the PAPI lamp is too low and an alarm is issued; S5. Calculate the distance between two adjacent PAPI lights in the whole process monitoring data set. If the distance is greater than the light interval threshold, it is determined that there is a completely unlit PAPI light between the two adjacent PAPI lights and an alarm is issued; S6. Calculate the number mean of the historical number sequence of small lights in the PAPI lamp in the whole process monitoring data set. If the number mean is less than the small light number threshold, it is determined that there are small lights that are not lit in the PAPI lamp and an alarm is issued; S7. Construct a PAPI lamp flicker recognition model and use the whole process monitoring data set to identify the flicker frequency characteristics of each PAPI lamp and find the characteristic center of all PAPI lamps. Calculate the Euclidean distance between the flicker frequency characteristics of each PAPI lamp and the characteristic center. If the Euclidean distance is greater than the flicker characteristic threshold, the corresponding PAPI lamp is judged as flickering abnormality and an alarm is issued.

2. The airport PAPI lighting anomaly recognition and detection method based on aerial video processing and analysis according to claim 1 is characterized by: The image frames of the video image are collected according to the time frame T1; the PAPI light labeling method is as follows: the area boxes of the belonging levels are marked according to the PAPI lights and the small lights in the PAPI lights, and the PAPI lights with too low brightness and too high brightness are marked respectively.

3. The airport PAPI lighting anomaly recognition and detection method based on aerial video processing and analysis according to claim 1 is characterized by: The YOLOV8 target detector performs frame-by-frame target detection on the video image stream F(t), where t represents the sequential number of the video frame. The ByteTrack algorithm performs data association and target tracking based on the target of the previous and next frames being the PAPI light. The ByteTrack algorithm performs data association and target tracking based on the target of the previous and next frames being the small lights in the PAPI light. The small lights in the PAPI light are hierarchically associated and belong to the corresponding PAPI lights.

4. The method for identifying and detecting abnormal airport PAPI lighting based on aerial video processing and analysis according to claim 1 is characterized in that: The whole process monitoring data set stores target detection data and trajectory data of n PAPI lights and m small lights attributable to each level; The expression of the ith PAPI light sequence of n PAPI lights is S(i) = {U(i), X(i), Y(i), Z(i)}, where U(i) is the historical trajectory coordinate center position sequence of the ith PAPI light, X(i) is the historical number sequence of small lights contained in the ith PAPI light, Y(i) is the historical brightness average sequence of the ith PAPI light area, and Z(i) is the missed video frame sequence in which no PAPI light label is detected in the trajectory of the ith PAPI light; the n PAPI lights are sorted into a full-process monitoring data set in chronological order.

5. The airport PAPI lighting anomaly recognition and detection method based on aerial video processing and analysis according to claim 1 is characterized by: In step S4, if the overall mean A is greater than the high brightness threshold, the PAPI lamp brightness is determined to be too high and an alarm is issued; the PAPI lamp brightness being too low and the PAPI lamp brightness being too high are collectively referred to as abnormal PAPI lamp brightness.

6. The airport PAPI lighting anomaly recognition and detection method based on aerial video processing and analysis according to claim 1 is characterized by: In step S5, the PAPI light information between two adjacent PAPI lights is determined and outputted along with the alarm prompt.

7. The method for identifying and detecting abnormal airport PAPI lighting based on aerial video processing and analysis according to claim 4 is characterized by: The PAPI light flicker recognition model extracts the difference sequence between the missed video frame sequence and the adjacent non-missed video frame sequence in the whole process monitoring data set, and calculates its mean and variance as the flicker frequency feature. The adjacent non-missed video frame sequence is the PAPI light target trajectory data that does not belong to the missed video frame sequence and is located before and after in time series.

8. The method for identifying and detecting abnormal airport PAPI lighting based on aerial video processing and analysis according to claim 7 is characterized by: The characteristic center Ra expression of the characteristic center is: Where R(i) represents the flicker frequency characteristic of the i-th PAPI lamp, and n represents the total number of PAPI lamps in the whole process monitoring data set.

9. The method for identifying and detecting abnormal airport PAPI lighting based on aerial video processing and analysis according to claim 1 is characterized by: The video images of the airport PAPI light area are captured by a drone flying from the far channel end to the near channel end along the arrangement direction of the PAPI light group according to the aerial photography route and the aerial photography altitude set to H1; the video images in the light label data set cover all weather scenes and all light states of the PAPI light group. All light states of the PAPI light group include that all PAPI light groups are normal, the brightness of the PAPI lights is too low, the brightness of the PAPI lights is too high, the PAPI lights are completely off, some small lights in the PAPI lights are off, and the PAPI lights are flashing abnormally.

Images