图像LK光流金字塔与帧间差分法的机器人视觉避障

Visual obstacle avoidance analysis of inspection robot based on pyramid LK light flow method

为了提高巡检机器人视觉避障能力,构建一种综合运用LK光流金字塔与帧间差分法(TD)的优化算法。通过帧间差分运算处理图像视频序列,确定运动对象所在区域,再选择金字塔光流运算,有效消除光照环境引起的检测结果变化。研究结果表明:通过帧间差分运算处理图像视频序列,确定运动对象所在区域,再选择金字塔光流运算,可以有效消除光照环境引起的检测结果变化,并且也能够对快速运动物体进行检测。环境噪声对障碍物区域产生的影响比较小,因实际测试效果受到帧间间隔的影响明显,导致有些区域无法检测。采用LK光流金字塔进行检测时,由于受到环境因素的影响明显,存在大量无关区域,此算法几乎不会受到环境因素的干扰。采用本文补偿模型形成了明显的边缘,与人眼感受深度相近的景物深度,能够有效满足计算得到的深度图数据误差补偿要求。

In order to improve the visual obstacle avoidance ability of inspection robot,an optimization algorithm based on pyramid LK optical flow method and inter-frame difference method(TD)was constructed.Image and video sequences are processed by inter-frame difference operation to determine the region where the moving object is located,and then pyramid optical flow operation is selected to effectively eliminate the change of detection results caused by the illumination environment.The results show that the image and video sequence can be processed by inter-frame difference operation,the region where the moving object is located can be determined,and then the pyramid optical flow operation can effectively eliminate the change of detection results caused by the illumination environment,and the fast-moving object can be detected.The impact of environmental noise on the obstacle area is relatively small,and the actual test effect is significantly affected by the interval between frames,resulting in some areas cannot be detected.The detection by pyramid LK optical flow method is obviously affected by environmental factors,and there are a lot of irrelevant areas.The algorithm is almost immune to environmental factors.By using the compensation model in this paper,the obvious edge is formed,and the scene depth is close to the perceived depth of human eyes,which can effectively meet the error compensation requirements of the calculated depth map data.