基于光照无关图的农业机器人视觉导航算法

完成沿作物行的行走作业是农业机器人视觉导航系统的一个基础功能，但是由于田间环境的复杂性，比如阴影的存在和天气的恶劣变化等外界因素的影响使导航参数的提取变得困难。该研究针对农业机器人视觉导航中存在的阴影干扰问题，采用基于光照无关图的方法去除导航图像中的阴影，然后采用增强的最大类间方差法进行图像分割和优化的Hough变换提取作物行中心线，最终通过坐标转换获得导航参数。最后，通过作物行跟踪试验表明，基于光照无关图的阴影去除方法不仅满足了导航实时性的要求，而且使农业机器人在光照变化的情况下导航参数提取的鲁棒性有了更大的提高。

Agricultural robot vision navigation algorithm based on illumination invariant image

Operation along crop row is a basic function of vision navigation for agricultural robot. The complex external factors of working environment in the field, such as existed shadow and weather change etc., make the navigation parameters extracting become more difficult. Aiming at shadow influence problem in vision navigation process of agricultural robot, illumination invariant image based on shadow removal algorithm was introduced. An enhanced OSTU image segmentation method was used, and an optimized Hough transformation algorithm was used to extract crop line, and coordinate transform was applied to obtain navigation parameters. Finally, tests of following crop row showed that shadow removal method based on illumination invariant image satisfied real-time requirement of system and improved the robustness of extracting navigation parameters under variable daylight illuminants.