基于网格运动统计的工件识别改进算法

Improved Algorithm for Workpiece Recognition Based on Grid-based Motion Statistics

为提高快速鲁棒特征(SURF)算法在复杂环境下进行工件识别的匹配精度及匹配效率,提出一种基于网格运动统计(GMS)的工件识别改进算法。首先,用加速分割检测(FAST)算法代替SURF特征点检测进行工件图像特征点的快速提取,并对特征点建立64维SURF描述符,使加快关键点检测速度的同时得到具有旋转尺度鲁棒性的特征点;其次,对特征点进行快速近似最近邻(FLANN)匹配得到粗匹配集;最后,采用改进GMS算法对图像进行网格化处理,根据运动平滑性进行内点与离群点的区分,保留正确匹配特征点对,实现误匹配点的剔除,准确找到待识别工件的位置。实验结果表明:改进算法比传统的SIFT、SURF和ORB等算法在工件图像受到旋转、平移、尺度、亮度等变化及影响时能够更高效、更准确识别工件,提高了复杂环境下工件识别的效率。

In order to improve the matching accuracy and efficiency of SURF algorithm for workpiece recognition in complex environments,an improved algorithm for workpiece recognition based on GMS is proposed.Firstly,uses the FAST algorithm to replace the SURF feature point detection to quickly extract the feature points of the workpiece image,and establish a 64-dimensional SURF descriptor for the feature points,so that the key point detection speed is accelerated and the feature points are robust to rotation and scale;secondly,FLANN matching is performed on the feature points to obtain a rough matching set;finally,the improved GMS algorithm is used to grid the image,and the interior points and outliers are distinguished according to the smoothness of motion,and keep the correct matching feature point pairs,realize the elimination of mismatched points,and accurately find the position of the workpiece to be identified.The experimental results show that the improved algorithm is more efficient and more accurate in identifying the workpiece than the traditional SIFT,SURF and ORB algorithms when the workpiece image is affected by changes and effects such as rotation,translation,scale,and brightness,and improves the efficiency of workpiece recognition in complex environment.