摘要
针对移动机器人和自动驾驶等研究领域中视觉系统的需要,提出了一种在图像分割基础上进行快速障碍物目标深度信息检测的方法。首先,在双目视觉系统得到的场景图像对中,通过迭代的多层次最佳阈值方法对图像进行分割,归一化后得到场景的目标区域;然后对目标区域进行快速的立体匹配工作,以得到目标的深度信息,进行障碍物检测。实验结果表明,该算法具有良好的效果和实用价值,不仅能应用于灰度一致性的道路,而且对于具有丰富纹理的非道路避障也有很好的效果。
For solving the visual systems problems of the mobile robot and automatic drive, an obstacle depth information re ̄cognition method based on the image segmentation was presented. At first, use the multi level best threshold method to segment the images getting from visual system and make up the segmented areas. Thus, we are comparatively easy to process the stereo matching of binocular images and get the targets depth information. The experimental results showed this algorithm is effective and has practical value. The obstacle avoiding method not only used for images of the roads with consistent gray, but also used for surroundings images with rich texture.
出处
《计算机应用研究》
CSCD
北大核心
2005年第4期249-251,共3页
Application Research of Computers
基金
国防基础研究资助项目(J1500C002)
关键词
图像分割
计算机视觉
障碍物检测
视差
Image Segmentation
Computer Stereo
Obstacle Detection
Disparity