Algorithm Research And System Implementation Of Landing Gear Monitoring Based On Machine Vision

With the popularity of aircraft in daily travel,the flight safety of aircraft has become the focus of people’s attention.The landing gear failure,as the largest proportion of flight failures in the past,has also become the key point of the current aircraft safety guarantee.At present,the safety and health of the landing gear is tested,or relying on expert knowledge and previous fault experience for probabilistic fault analysis,or using existing hardware facilities to physically operate the landing gear structure to observe the health status of the landing gear structure.While these traditional analytical methods can perform multifaceted health detection,they do not utilize rich visual information to analyze structural anomalies.The core work of thesis is to analyze the abnormal state of the landing gear structure from the perspective of machine vision,study the algorithm for identifying the abnormal information of the landing gear structure,and design the landing gear monitoring system combined with the algorithm research.The system collects data from the aircraft landing gear through the camera,analyzes it,and identifies possible structural defects and foreign object information.The study can be seen as a complement to,rather than a substitute,for existing studies.The main research content of thesis is as follows:（1）A judges jitter algorithm and a debounce method that combine distortion transformation with adversarial network are proposed for high-frequency jitter of video.The judges jitter algorithm mainly determines the jitter state of the video frame by calculating the probability distribution of the global optical flow vector offset angle.The debouncing algorithm proposes a deep change matrix through a deep neural network,and uses the distortion frame to map the stable frame,which improves the video stability maintenance effect.（2）Aiming at the complex and rapidly changing background of the landing gear,a defect detection algorithm with low delay and high accuracy and a modeling algorithm for real-time background update are proposed.In order to eliminate sensitive edge noise,adaptive threshold setting is used,and the time-consuming time of defect detection algorithm is improved to improve the calculation efficiency.Based on the rapidly changing scene information,thesis optimizes the Vi Be+ background modeling algorithm,eliminates the Ghost area,performs real-time background update,and realizes the detection of foreign body information on the landing gear through the comparative analysis of the real-time frame and the template frame.（3）Realized the prototyping of the real-time monitoring system.Through the two aspects of video image stabilization and structural anomaly detection,the algorithm of the core system is systematically evaluated and experimented,which effectively proves the detection ability of the proposed algorithm and builds a landing gear monitoring system based on machine vision.