Research On Target Dots-cohesion Processing Algorithm

With the rapid development of artificial intelligence and machine learning,radar tends to be sophisticated,and the dots data with target information got by the receiver increase greatly,which brings great inconvenience to the following target tracking and track processing.In order to detect the target accurately and ensure the accuracy of the track information while tracking,a good dots-cohesion algorithm becomes particularly important.In this paper,two kinds of dots-cohesion algorithms are studied,one is the DBSCAN dots-cohesion algorithm based on KD-tree space search,and the other is the dots-cohesion algorithm based on the combination of contour tracking and region growing.The two algorithms use DBSCAN to cluster target pixels and extract target contours to achieve target detection seperately.Target detection is the central part of the dots-cohesion and the key problems are studied as follows:1.Aiming at the real-time problem of traditional DBSCAN dots-cohesion algorithm,this paper uses KD-tree in DBSCAN,instead of its Eps-neighborhood search path,and studies a DBSCAN dots-cohesion algorithm based on KD-tree with higher computing efficiency.By using the spatial search algorithm,only a limited KD-tree paths are traversed when searching the Eps-neighborhood,which reduces the target clustering time,improves the dots-cohesion efficiency,and the time complexity of the algorithm is reduced from O(n~2)to O(nlogn).Through a large number of comparative experiments on the measured radar data by MATLAB,the algorithm is proved to be feasible.At the same time,the experimental results show that the algorithm is better for large data.2.In order to solve the problem that the traditional contour tracking algorithm can not extract the contour of the target of non-connected pixels,based on 8-chain code,this paper proposes a contour tracking algorithm based on(m,n)-mask.While tracking,the algorithm uses a variable size mask to extract the target contour flexibly and realize the target detection.Through the experimental verification,for the target of non-connected pixels,(m,n)-mask works fine on contour tracking;for the target of connected pixels,(m,n)-mask realizes the data compression of contour points.The experimental results show that for the same radar image,after contour tracking,the number of contour points of(3,3)-mask is 58.6%less than that of 8-chain code.3.In order to ensure the estimation accuracy of the target dots after cohering,this paper studies the region growing algorithm.On the basis of the contour tracking algorithm,the region growing algorithm is combined with it,and a dots-cohesion algorithm based on the contour tracking region growing is studied.The target contour is regarded as the seed points,and the 8-neighborhood growth criterion is used to grow the target.After obtained the target all pixels,the algorithm coheres the target dots.The feasibility of the algorithm is verified by experiments on the measured and simulated data.