用于距离图像2D扫描线的极速边缘检测器  

A Rapid Edge Detector for 2D Scan Line in Range Image

作　　者：王建[1] 姚振强[1] 尹明德[2] 朱瑶洁 李康妹[1] 

机构地区：[1]上海交通大学机械系统与振动国家重点实验室,上海200240 [2]南京航空航天大学机电学院,江苏南京210016 [3]延锋伟世通汽车饰件系统有限公司,上海201206

出　　处：《电子学报》2010年第7期1711-1715,共5页Acta Electronica Sinica

基　　金：国家973重点基础研究发展计划(No.2006B705400);国家科技支撑计划(No.2006BAH02A17);上海市产学研专项(No.06DZ11310)

摘　　要：本文提出一种对距离图像2D扫描线进行粗大噪声去除和边缘识别处理的极速检测器.以扫描线典型轮廓分布规律分析为基础,定义了数个特征描述系数并以此建立判别准则,实现快速去噪和边缘检测.该方法具有内在的平移和旋转不变性,并可通过自适应参数来实现尺度不变性.文中采用基于候选点的参数自学习方法来进行阈值学习.将扫描点参数集计算步骤进行高度集成,可对最大限度消除冗余计算.通过实物扫描实验,对算法的计算复杂度、时间开销、检测准确度和稳定性进行了详细评估.在普通PC平台上,整合模式耗时约0.15ms即可准确检测单条扫描线中的所有边缘点.评估结果表明,本文提出的检测器具有极高的运行效率、检测准确性、鲁棒性和可扩展性,可满足极为苛刻的工业应用要求.In this paper,we propose a novel rapid detector which is used to denoise and recognize edges for 2D scan line in range image.Based on silhouette analysis in a scan line,a few coefficients are defined to characterize large noise and edge points,and several determinative criteria are stipulated for rapid denoising and edge detection.Because all the coefficients used are distances or their ratio,the proposed method is inherently invariant to translation and rotation.The threshold set is trained by selecting some feature point candidates.By merging parameter attributes calculation,redundant computation is decreased to the lowest in an integral mode algorithm.Based on real data experiments,computational complexity,time expenditure,accuracy and stability of the method are evaluated in detail.The proposed detector under integral mode can process a scan line within 0.15 ms on an ordinary PC.The results indicate that the proposed method has excellent performance on efficiency,accuracy,robustness and scalability,and is qualified for real-time processing in stringent industrial occasions.

关 键 词：快速边缘检测 实时数据处理 激光测距雷达 去噪滤波 3D距离图像处理 计算机视觉 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]