一种改进SURF特征匹配的装配工件快速识别方法

A Fast Recognition Method for Assembly Workpiece Based on Improved SURF Feature Matching

在工件装配过程中,针对基于SURF算法的工件匹配识别过程速度慢、实时性较差的问题,通过对K-Means聚类算法以及机器学习中文本检索方法进行研究,推导了基于聚类算法的特征点种类词袋模型,对特征点类别进行统计。提出反向标记法生成目标物体描述向量。在物体匹配识别阶段,对待检测物体的描述向量分段进行匹配并选取满足阈值条件的子向量作为最终匹配向量来实现目标物体的识别。实验结果表明改进算法在一定尺度、光照、旋转条件影响下实现了工件的准确识别,并且提高了识别速度。

To solve the problem of slow speed and poor real-time in object matching and recognition based on SURF algorithm during the process of workpiece assembly,through the research of K-means clustering algorithm and text retrieval method in machine learning,the feature point category bag model based on clustering algorithm is deduced,and the feature point category is counted.The reverse labeling method is used to generate the description vector of the target object.In the phase of object matching recognition,the description vectors of the detected objects are matched in segments and the sub-vectors satisfying the threshold conditions are selected as the final matching vectors to realize the target object recognition.The result of the experiment shows that the improved algorithm achieves the accurate recognition of workpiece under the influence of certain scale,illumination and rotation conditions,and improves the recognition speed.