一种基于边缘能量图的物体检测方法  

An object detection approach based on edge energy image

作　　者：马勤勇[1] 聂栋栋[2] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004 [2]燕山大学理学院,河北秦皇岛066004

出　　处：《光学技术》2014年第1期40-44,共5页Optical Technique

基　　金：河北省自然科学基金(A2011203053);秦皇岛市科学技术研究与发展计划(2012021A044)

摘　　要：物体的变形、边缘中断等因素通常对物体检测的结果影响很大。为了减小这些因素的影响,提出一种新的基于边缘能量图的物体检测方法。边缘能量图在一定程度上反映了物体的边缘出现的概率,便于计算并具有较好的鲁棒性。计算出图像的边缘图像,在此基础上分别计算出模型和测试图像的边缘能量图。基于多分辨率的滑动窗口,在图像中找出测试图像边缘能量图与模型边缘能量图最相似的区域。基于边缘能量图快速得到目标物体的轮廓。实验结果显示,该方法使用较快的计算速度得到了相对较高的检测率。Object detection is often affected by object deformation and edge discontinuity. A new object detection approach is proposed to reduce the effects based on edge energy image. Edge energy image gives the poSSlDlntY ence of objeet＇s edge to some extent, which is easy to be calculated and has good robustness- Edge energy images of mod- el and test image are generated from edge pixels extracted from images. Based on multiple resolutions sliding window, the region of the testing edge energy image which is most similar to the model edge energy image is chosen as the detect result. OhiecUs contour is located rapidly based on edge energy image. The test results confirm that the proposed algo- rithm achieves fine detection rate with fast speed.

关 键 词：信息光学 物体检测 边缘 能量图 多分辨率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]