基于子块运动补偿的运动目标检测

发布时间：2018-03-13 05:21

本文选题：运动目标检测　切入点：子块运动补偿　出处：《电子学报》2017年01期 　论文类型：期刊论文

【摘要】：鱼眼相机成像视角大,获得信息丰富,在车载应用中具有广阔应用前景.本文提出了一种适用于移动单目鱼眼相机的运动目标检测方法.首先,提出一种子块运动补偿模型补偿图像背景运动,解决了现有运动补偿模型对强视差背景补偿效果不好的问题.其次,在子块运动补偿模型参数求解时,通过引入自车运动参数简化模型参数个数,并结合直接方法求解,避免了传统基于特征点匹配方法求解参数时易受误匹配特征点影响的问题.然后,针对鱼眼相机的成像形变问题,本文提出了一种三平面校正方法获取鱼眼图像的子块运动补偿图像.最后,利用鱼眼图像的子块运动补偿图像和真实拍摄图像的差异信息实现运动目标检测.多种测试场景下的实验结果表明了本文方法的有效性.
[Abstract]:The fish-eye camera has a wide application prospect because of its large angle of view and abundant information. In this paper, a moving target detection method for moving single-eye fish-eye camera is proposed. A subblock motion compensation model is proposed to compensate the background motion of the image, which solves the problem that the existing motion compensation model is not effective for the strong parallax background compensation. Secondly, when the parameters of the subblock motion compensation model are solved, By introducing the number of parameters of the model to simplify the motion parameters of the vehicle, and combining with the direct method, the problem that the traditional matching method based on the feature points is easy to be affected by the mismatched feature points is avoided. In order to solve the problem of imaging deformation of fish-eye camera, a three-plane correction method is proposed to obtain subblock motion compensation image of fish-eye image. The difference information between the subblock motion compensation image and the real shot image of the fish-eye image is used to detect the moving target. The experimental results in various test scenarios show the effectiveness of the proposed method.
【作者单位】：东北大学研究院;东软集团股份有限公司;
【基金】：国家自然科学基金(No.61273239)
【分类号】：TP391.41

【相似文献】