Study Of Moving Target Detection And Recognition Based On Laser Radar

Imaging laser radar(LIDAR) is a kind of three-dimensional sensors with high range resolution and angular resolution. The research of three-dimensional moving target detection and recognition technology based on imaging laser radar has gained great attention for both theoretical significance and application potentials, especially in the military. Based on double wedge imaging laser radar, this paper has studied threedimensional target detection and recognition in moving scenarios. The main contents of the research are as follows:(1)The working principle of imaging laser radar technology and double wedge scanning system are introduced, and based on rotating double wedge scanning equation, this paper analyzed scanning paths under different parameters. Ground point cloud distribution in searching scanning and tracking scanning of double wedge laser radar system has been deduced with the actual parameters. In addition, moving scenarios are modeled in a geometrical way, and deduced the scanning point location in the moving scene at every moment. Moreover, the target motion distortion and movement position error of adjacent point on the surface of the target are statistically analyzed. This paper also summarizes the effect of target motion on detection and recognition.(2) A method of real-time moving targets detection has been presented based on the constraints on height difference, natural angle and point cloud location. The algorithms firstly mainly employ these constraints to detect abnormal points, then using a clustering method by setting appropriate threshold of both size and the number of abnormal point to detect the suspected target. Finally, several experiments about target detection under different terrain scene are performed, and derive high detection rates in this paper.(3) Arming at the difficulty in using mature approaches to recognize the target point cloud which has been moving distorted, the idea of real-time correction for moving target points cloud is proposed. The frontal two frames of the data collected by laser radar scanning are utilized to estimate the moving speed and the direction of the target. Then the third frame target data is utilized for correcting and identifying. Moreover, the estimated error of moving speed and direction are analyzed. And several experiments are presented to verify the reliability and effectiveness of the idea about real-time correction.(4) Proposed a method for moving cloud points recognition based on learning. this paper firstly conducts data preprocessing advanced, and then extracts several typical characteristic descriptions of the moving target, including size characteristics, shape characteristics and local spatial distribution characteristics. With the help of SVM classifier, some good effects can be achieved for classifying recognition.