Clustering With Multidimensional Features And Target Feature Extraction In Coherent Radar Range-doppler Domain

Radar target recognition has decades of research history, the direction is not only has important theoretical significance, also has a very broad application prospects, especially of great military value. At present, the technology of the radar target recognition has already accumulated rich theoretical knowledge and appeared a series of fruitful research achievements. However, With the Electromagnetic environment becomes complex and diversified and interference technology grows quickly, the challenge for the radar target recognition technology has become increasingly serious. This article focuses on target detection problem in the coherent radar’s range-doppler domain. By means of analyzing measured range-doppler data, this paper deeply study the clustering algorithm which is suitable for the range-doppler domain and acquire the multidimensional features which can reflect essential information of different targets, finally, use the features to recognize different targets. The specific work arrangement is summarized as follows:Chapter 1 introduces the research status and development trend of the technology of target recognition as well as the background and significance of the subject and pointes out the problems which are to be solved in coherent radar feature extraction, and summarizes the main work done in this paper.Chapter 2 introduces the basic concept of clustering analysis and its application in the range-doppler domain. Through the different division of clustering algorithm, this paper introduces the various clustering analysis algorithm which has been widely applied at present. In view of the different types of data objects, clustering algorithm using the objective function and distance function is very different. How to correctly select the objective function is the key to effective clustering algorithm. In this paper, through the analysis of clustering algorithm in different areas to learn useful information and analyses problems which clustering algorithm faced for in range-doppler domain, a kind of clustering algorithm which suitable for range-doppler domain is been put forward.Chapter 3 puts forward the amplitude aiding dual-threshold clustering algorithm. It mainly introduced the basic principle of the amplitude aiding dual-threshold clustering algorithm. On each link of the algorithm, in turn, gives the principle of edge detection technology, the way of data standardization processing, coarse clustering process and the secondary process. Algorithm using the range features, doppler feature and amplitude feature complete data clustering, The application of edge detection algorithm for clustering provides constraints, and reduce the computation complexity of the algorithm. The algorithm has strong adaptability, able to handle range-doppler domain data, get a satisfactory clustering result.Chapter 4 studies the multidimensional feature extraction technology and its separability analysis among different target. It mainly introduces the significance and methods of the feature extraction and the basic theory of SVM. This paper put forward multidimensional characteristics for target classification and recognition, such as the amplitude entropy, disperse coefficient, etc. Finally, use the data which measured in coherent radar range-doppler to validate the effectiveness of the features through the SVM algorithm.Systematic conclusions are finally drawn in Chapter 5, as well as the defects of this paper and the prospects of further research.