光学学报, 2014, 34 (s2): s210004, 网络出版: 2014-11-19
基于阈值分割方法的复杂场景激光图像目标提取算法研究
Research on the Target Extraction Algorithm for Laser Image with Multiple Scene Based on the Threshold Methods for Segmentation
图像处理 激光成像雷达 目标提取 阈值分割 距离像 image processing imaging laser radar target extraction threshold segmentation range image
摘要
激光主动成像雷达与被动红外成像器相比,显著优势是能提供三维几何距离像,可获知视场内多目标的空间位置信息,这极大降低了多建筑物复杂场景目标提取算法的复杂度。利用自研的条纹管激光成像雷达,获取多组复杂场景激光图像。利用三种不同的阈值分割算法对激光距离像进行目标提取,并计算了每种算法的误分率。对于单一目标,误分率可达1.1%;对于整个场景的目标,平均误分率可达7.9%。实验结果表明,阈值分割算法能有效解决复杂场景激光距离像的目标提取问题,其中改进直方图法的误分率最小,可以为激光成像雷达的图像处理方面的工作起到一定的借鉴作用。
Abstract
Compared with passive infrared imager, the laser active imaging radar has an obvious advantage that it can provide a three dimensional range image, and can obtain the spatial position information of multiple targets within the field of vision, which greatly reduces the complexity of extraction algorithm for objects of complex scene with many buildings. The streak tube imaging laser radar made by the research group is used to get many sets of laser images of multiple scene. Using three different threshold algorithms for segmentation to extract the targets in the laser range image, and calculating the error rate on the extraction of each algorithm. For the single target, the error rate on the extraction can reach 1.1%; for the targets of the whole scene, the average error rate on the extraction can reach 7.9%. The experimental results show that, the threshold algorithm for segmentation can effectively solve the problem of target extraction in the laser range image of the multiple scene, and the error rate on the extraction of the improved histogram method is minimum. This study can play a certain reference role for image processing work of imaging laser radar.
周鑫, 孙剑峰, 姜鹏, 熊卫华, 王骐. 基于阈值分割方法的复杂场景激光图像目标提取算法研究[J]. 光学学报, 2014, 34(s2): s210004. Zhou Xin, Sun Jianfeng, Jiang Peng, Xiong Weihua, Wang Qi. Research on the Target Extraction Algorithm for Laser Image with Multiple Scene Based on the Threshold Methods for Segmentation[J]. Acta Optica Sinica, 2014, 34(s2): s210004.