农业机器人视觉导航中多分辨率路径识别

针对非结构化的农田自然环境，提出农业机器人视觉导航时的多分辨率路径识别算法。首先探讨了用于路径识别的适宜的彩色特征，然后基于小波变换分析了多分辨率边缘检测过程，最后结合跟踪路径的特点融合了多分辨率检测结果。油菜地图像的处理结果表明了该算法的有效性。

Multi-resolution Road Recognition for Vision Navigation

Agricultural robot navigated by machine vision is important for the sustainable development of agriculture, but the road recognition, a key technology for the navigation, is a difficult problem when used in non structured field environment. In this paper, an algorithm of multi resolution road recognition is introduced to provide the solution to above problem. At first, a suitable color feature for the road recognition was discussed, based on the wavelet transform then the multi resolution edge detection is analyzed carefully in the image of non structured field. Finally, under the premise of line road model in the view field of the machine vision, the results of the multi resolution edge detection were fused through the least square error. The treatment of the actual color picture of the cole field showed the efficiency of this algorithm.