基于三维点云的坦克识别  

Tank Recognition Based on Three-dimensional Point Cloud

作　　者：陈长胜[1] 付文兴[2] 孙永波[3] 马杰[1] 方斌[1] 

机构地区：[1]华中科技大学多谱信息处理技术国家级重点实验室,武汉430074 [2]航天科工集团第三研究院,北京100074 [3]江汉大学物理与信息工程学院信息与控制中心实验室,武汉430074

出　　处：《计算机与数字工程》2016年第12期2302-2305,2385,共5页Computer & Digital Engineering

基　　金：中央高校基本科研业务费项目(编号:2014GF);航天科技创新基金(编号:CASC04);一院高校联合创新基金(编号:CALT201303)资助

摘　　要：当坦克的炮塔发生旋转后,对坦克的识别会产生极大的影响。为了解决这个问题,论文提出了一种快速分割坦克的方法,首先根据霍夫变换获取边界线,然后结合坦克的宽度得到分界面,从而分割坦克,再分别识别炮塔和车身,不仅可以提升坦克的识别率,还可以计算出炮塔和车身之间的相对旋转角度。论文对5种坦克12个炮塔旋转角度下进行实验,实验结果表明该分割算法在速度和正确率上效果较好,坦克的正确识别率很高,炮塔和车身的相对旋转角度也基本准确。The rotating of turret can make an enormous impact on the tank recognition, A fast tank segmentation meth-od is proposed in this paper in order to eliminate the influence. Firstly , Hough transformation theory is used to calculate the boundary line , which is used to get the boundary plane together with the information of tankas width. After segmentation , the recognition of turret and body is needed, which not only can promote the recognition rate of tank,but also can calculate the relative rotation angle of the turret and body. The experiments are performed by using the simulated cloud data of 5 tanks under 12 different rotation angles of the tank turret. The resutt shows that this method has a good performance of the speed and the accuracy. The recognition rate is high , the relative rotation angle of the turret and bady is accurate.

关 键 词：坦克 炮塔旋转 霍夫变换 姿态估计 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]