一种降低激光雷达行人识别错误率的多算法组合的研究

针对2维激光雷达获取的点云信息，在类圆弧人腿形状识别算法的基础上，提出了一种可以降低行人识别错误率的多算法组合。该方法先采用高斯滤波算法降低噪声的影响，然后利用近邻聚类算法对数据进行聚类处理，再利用组合的聚类中心角算法和最小二乘圆拟合算法对聚类后的数据进行行人腿部检测，完成行人的识别。该算法混合利用LabVIEW和matlab 软件平台，并使用激光雷达对现场的行人进行了识别验证，该多算法组合与单一使用一种识别算法相比，行人识别的错误率由40%降低到了10%以下，充分说明该算法组合具有良好的性能。

A STUDY ON MULTI ALGORITHM COMBINATION FOR REDUCING PEDESTRIAN RECOGNITION ERROR RATE OF LASER RADAR

According to the point cloud information obtained from the 2 dimension laser radar, a new algorithm is proposed, which can reduce the recognition error rate. This method first uses Gauss filter algorithm to reduce noise, and then use the nearest neighbor clustering algorithm for data clustering, and then use the clustering center combination angle algorithm and least square circle fitting method for Pedestrian Leg detection of clustering data after completion of pedestrian recognition. The hybrid algorithm using LabVIEW and MATLAB software platform, and the scene of the pedestrian recognition verification using laser radar, compared to a recognition algorithm of the algorithm in combination with a single use, pedestrian recognition error rate reduced from 40% to 10%, and shows that the combined algorithm has good performance.