机器人视觉中的障碍物优化识别方法研究

Research on Optimization Recognition Method for the Obstacle in Robot Vision

针对机器人视觉中障碍物识别的准确率和速度问题,机器人对障碍物的识别主要依据障碍物以背景在颜色、纹理、形状上的差异特征完成识别的,但是,当前的单目识别过于依赖差异特征为主,一旦障碍物与背景的差异化特征较小,会造成障碍物特征的淡化。传统算法需要进行多次校对识别,降低了识别的准确率和效率。提出采用Hough变换的机器人视觉中障碍物的识别方法。通过机器人头部的超声波传感器获得机器人与障碍物的距离,利用三角形成像原理建立障碍物的识别模型,采用Hough变换的方法确定障碍物在机器人视觉采集的图像中的直线边缘,利用约束条件获得障碍物的最大高度和最大宽度,完成了障碍物的准确识别。仿真结果表明,改进算法能够提高机器人视觉识别的准确率和速度。

In view of the accuracy and speed of obstacle recognition in robot vision, the realization of obstacle recognition of robot is based on the difference characteristics of the background in color, texture and shape. Howev- er, current single target recognition is too much dependent on the difference characteristics, once the difference char- acteristics of the obstacle and the background are small, it will cause the dilution of the characteristics of the obsta- cle. Traditional algorithms need to be checked and identified for many times to reduce the accuracy and efficiency of the identification. An identification method for the obstacles in robot vision based on Hough transform is proposed. The distance between robot and obstacle is obtained by ultrasonic sensor in robot head. The recognition model of ob- stacle is established by using the triangle imaging principle. Hough transform method is used to determine the linear edge of obstacles in the image of robot vision acquisition. The maximum height and the maximum width of the obsta- cle are obtained by using the constraint condition to compete the accurate recognition of obstacle. The simulation re- suits show that the proposed algorithm can improve the accuracy and speed of the robot visual identification.