一种立体视觉的障碍物检测的细化方法

A Refined Method for Stereo Vision Obstacle Detection

在立体视觉的障碍物检测过程中,由于室内地面的纹理特征较弱,使得检测的障碍物区域包含了较大的地面干扰,获得障碍物目标位置并不理想,该文使用了一种新的细化方法,改善了目标检测精度。该算法基于双目视觉获得深度信息,采用深度图像分割法提取出目标的粗略区域;然后结合其Canny边缘坐标信息对障碍物区域进行细化,进而得到准确的障碍物位置信息。实验结果表明了该方法具有较强的自适应性。

In the obstacle detection process of stereo vision,due to the low texture of the indoor ground,the detected obstacle area contains a large ground interference,and obtaining the target position of the obstacle is not ideal.This paper uses a new refinement method,improved target detection accuracy.The algorithm obtains depth information based on binocular vision,and uses the depth image segmentation method to extract the rough area of the target;then combines the Canny edge coordinate information to refine the obstacle area to obtain accurate obstacle position information.Experimental results show that this method has strong adaptability.