红外与激光工程
紅外與激光工程
홍외여격광공정
INFRARED AND LASER ENGINEERING
2015年
4期
1115-1120
,共6页
吕丹%孙剑峰%李琦%王骐
呂丹%孫劍峰%李琦%王騏
려단%손검봉%리기%왕기
3D姿态角估计%激光雷达%模型坐标系%距离像
3D姿態角估計%激光雷達%模型坐標繫%距離像
3D자태각고계%격광뢰체%모형좌표계%거리상
3D pose estimation%ladar%model coordinate system%range image
在激光雷达目标识别中,目标姿态的精确估计可以有效地简化识别过程。现有的PDVA算法主要是针对地面结构化目标而提出的一种3D目标姿态估计方法。该方法利用模型坐标系(MCS)各个坐标轴的正方向向量来确定目标的三维姿态角,其有效性通过实验得到了验证。但该方法在确定MCS各坐标轴的正方向向量时,所消耗的时间比较多,影响了算法的执行效率。文中提出了一种改进的PDVA算法,利用聚类中心邻域判别CCND法来加速MCS各坐标轴的正方向向量的确定过程。采用四种地面军用车模型目标进行了仿真实验,实验结果显示,改进的PDVA算法的平均运行时间约占PDVA算法的66%,极大地提高了目标3D姿态估计的执行效率。
在激光雷達目標識彆中,目標姿態的精確估計可以有效地簡化識彆過程。現有的PDVA算法主要是針對地麵結構化目標而提齣的一種3D目標姿態估計方法。該方法利用模型坐標繫(MCS)各箇坐標軸的正方嚮嚮量來確定目標的三維姿態角,其有效性通過實驗得到瞭驗證。但該方法在確定MCS各坐標軸的正方嚮嚮量時,所消耗的時間比較多,影響瞭算法的執行效率。文中提齣瞭一種改進的PDVA算法,利用聚類中心鄰域判彆CCND法來加速MCS各坐標軸的正方嚮嚮量的確定過程。採用四種地麵軍用車模型目標進行瞭倣真實驗,實驗結果顯示,改進的PDVA算法的平均運行時間約佔PDVA算法的66%,極大地提高瞭目標3D姿態估計的執行效率。
재격광뢰체목표식별중,목표자태적정학고계가이유효지간화식별과정。현유적PDVA산법주요시침대지면결구화목표이제출적일충3D목표자태고계방법。해방법이용모형좌표계(MCS)각개좌표축적정방향향량래학정목표적삼유자태각,기유효성통과실험득도료험증。단해방법재학정MCS각좌표축적정방향향량시,소소모적시간비교다,영향료산법적집행효솔。문중제출료일충개진적PDVA산법,이용취류중심린역판별CCND법래가속MCS각좌표축적정방향향량적학정과정。채용사충지면군용차모형목표진행료방진실험,실험결과현시,개진적PDVA산법적평균운행시간약점PDVA산법적66%,겁대지제고료목표3D자태고계적집행효솔。
In the target recognition of ladar, the accurate estimation of target pose can effectively simplify the recognition process. The existing PDVA algorithm as a method of target 3D pose estimation is mainly for ground structured targets. This method uses the planar normals of rigid targets as the vectors in the positive direction of the axes in model coordinate system(MCS) to estimate the 3D pose angles of targets, and its effectiveness has been verified by experiments. However, it is time consuming when determining the positive direction vectors of the axes in MCS and affecting the efficiency of the algorithm. In this paper, an improved PDVA algorithm was proposed and a method of clustering center neighborhood discriminant (CCND) was used for accelerating the determination process of positive direction vectors of the axes in MCS. The simulation experiments were performed with four military vehicle models. The results show that the average running time of the improved PDVA algorithm only accounts for about 66%of the PDVA algorithm, and it greatly improves the efficiency of target 3D pose estimation.