单目图像内静态物体深度的自动测量方法  

An Automatic Method of the Depth Measurement of Static Objects in a Monocular Image

作　　者：何立新[1,2,3] 孔斌[2] 杨静[2] 

机构地区：[1]中国科学技术大学自动化系,安徽合肥230027 [2]中国科学院合肥智能机械研究所,安徽合肥230031 [3]合肥学院网络与智能信息处理重点实验室,安徽合肥230601

出　　处：《武汉大学学报（信息科学版）》2016年第5期635-641,共7页Geomatics and Information Science of Wuhan University

基　　金：国家自然科学基金(91120307;61005010);安徽省教育厅自然科学基金(KJ2013B230;KJ2013A226;KJ2015A162);安徽省高等学校质量工程项目(2015ckjh047;2015ckjh048;2015ckjh058;2015ckjh061;2015zy054);合肥学院重点建设学科基金(2014xk08);合肥学院学科带头人培养对象基金(2014dtr08)~~

摘　　要：提出了一种从普通单目相机拍摄的两幅图像中自动获取其中物体深度的方法。该方法首先保持相机的参数不变,相机沿光轴方向移动,在物距为u和u+d处各拍摄一张图像,然后采用局部二值拟合能量(local binary fitting energy,LBF)模型的方法分割出图像中的物体,再将物体像的熵的相对变化率和加权Hu氏不变矩结合起来实现图像内物体的自动匹配,最后运用本文推导的公式计算出各个物体深度。实验结果表明了该方法的有效性。Measurement of object depth information in an image is an important task in the stereo vision domain. An automatic method of the depth measurement from two images captured by an ordinary monocular camera is proposed. While keeping the parameters of camera constant, two images are captured at the distance between object and the lens definted as u and u＋d, respectively. Objects in the images are acquired by the image segment method using the LBF model. The objects are matched automatically using the combined the relative difference ratios of entropy from the object image and the weighted HU invariant moment. The object depth can be calculated by a formula presented in this paper.. Experimental results show that the method is effective.

关 键 词：单目图像 深度测量 图像分割 物体匹配 

分 类 号：P237.4[天文地球—摄影测量与遥感] TP391.4[天文地球—测绘科学与技术]