基于机器视觉的铲斗目标检测

Bucket Target Detection Based on Machine Vision

为实现挖掘机器人在挖掘工作中能够对挖掘轨迹进行自我认知,结合了机器视觉中比较常见的运动目标检测算法-帧差法和背景差分法,对挖掘机铲斗目标检测展开了研究。对帧差法中的二帧差法和三帧差法以及背景差分法中的混合高斯建模法进行了研究,基于这些算法对铲斗目标进行检测。基于Open CV进行程序设计,对挖掘机模型的铲斗进行检测,验证算法的可行性。通过实验结果对比分析,结合三帧差法以及高斯混合建模法能够更好地实现铲斗目标检测。

To Apply the robots to the excavation work, this paper refers the moving target detection algorithm and studies the frame difference method and background difference method in machine vision technology for the bucket target detection during the excava- tion. In particular, it also researches on two-frame difference and three-frame difference methods as well as Gaussian mixture modeling in the background difference method. Based on the results, its feasibility is verified through OpenCV program-based monitoring of the excavator bucket. The comparative analysis of the experimental results indicates that the three-frame difference method is used with Gaussian mixture modeling to allow the bucket target detection to be better.