基于ARM处理器的单目视觉测距定位系统

Monocular Vision Measurement System Based on ARM Microprocessor

针对在多机器人协作和服务机器人目标追踪的过程中,需要对目标物体或者同伴进行准确的目标定位这个问题,设计了一个以ARM9微处理器为处理模块、CMOS摄像头为采集模块的一个嵌入式视觉系统,实现了单目视觉测距定位。该系统设计大体分为目标识别和定位两部分,目标标识使用一个纯色的色标块,通过对摄像头采集的图像中检测色标块区域来实现目标物体定位;对色标块的图像分割采用基于RGB三个颜色分量的阈值分割算法,单目测距采用P4P方法,其中,摄像机的内参数标定则借助于Matlab工具箱来完成。同时,采用RANSAC方法确定色标块的四条边界直线,以四条直线交点作为色标块的4个角点,从而提高角点检测的精度。实验结果表明,该测距定位系统定位精度较高,系统鲁棒性好,具有良好的应用前景。

The target location problem of multi-robot cooperation and service robot tracking is discussed.A monocular vision measurement system with the ARM9 microprocessor as the main processing module and the CMOS camera as the image capturing module is presented. The system design includes two parts,i.e.,object recognition and positioning.A pure color block is served as the goal object, which can be detected by searching the color block in the captured image.The color image threshold segmentation method based on RG...