基于激光雷达与红外图像融合的车辆目标识别算法

为了提高车辆目标在不同测试条件下的识别效率,降低系统的漏检率和误检率,提出了一种基于激光雷达与红外图像融合的车辆目标识别算法。该算法利用目标原点矩参量表征目标的红外特征,用匹配相似度表征目标的点云特征,再经过轴系对齐和尺度变换实现图像融合。实验采用激光雷达与红外同轴光路获取的两类数据进行图像融合,再利用目标匹配阈值进行迭代筛选,最终识别车辆目标。对比了1帧、20帧和40帧图像中具有不同属性的车辆目标识别效果,结果显示,本算法输出的目标识别区域正确适当。在1000帧图像的多种测试条件的实验中,本算法的漏检率均小于100,误检率均小于50,明显优于传统的距离向数据分类法和光谱分类法,验证了其具有较好的鲁棒性。

Vehicle target recognition algorithm based on fusion of lidar and infrared image

In order to improve the recognition efficiency of vehicle targets under different test conditions and reduce the missed detection rate and false detection rate of the system,a vehicle target recognition algorithm based on the fusion of lidar and infrared images is proposed.In this algorithm,the target moment parameter is used to characterize the infrared feature of the target,and the matching similarity is used to characterize the point cloud feature of the target.The system uses axis alignment and scale transformation to achieve image fusion.Lidar and infrared coaxial optical paths are used in the experiment.After the two types of data are fused with images,iterative filtering is completed through the target matching threshold,and finally the vehicle target is identified.Comparing the effects of vehicle target recognition with different attributes in 1 frame,20 frame and 40 frame images,the results show that the target recognition area output by this algorithm is correct and appropriate.In experiments with multiple test conditions of 1000 frames of images,the missed detection rate of this algorithm is less than 10.0%,and the false detection rate is less than 5.0%.It is obviously superior to traditional distance data classification and spectral classification,and it has better robustness.