| Workpiece recognition is a typical application of machine vision technology in the industry. During the automated production process, in order to complete the task of workpiece assembling and sorting, we need to obtain the type and the position information of the workpiece. In the industrial production line or bench, the position of workpiece is not fixed. The occlusion between multiple workpieces is often occurred, which brings a great challenge to the workpiece recognition. The research of occluded workpiece recognition has great research value and practical significance.This thesis focuses on the problem of occlusion in workpiece recognition. The main works are as following:(1) This thesis use edge feature matching method to solve the problem of occlusion. Gradient direction of the edge points are used as the matching feature in calculating the similarity, which has good robustness to illumination variation and partial occlusion. Meanwhile, stopping criteria and image pyramid are employed to accelerate the matching rate. In the stopping criteria aspect, separate edge points are selected as the test points to solve the problem of fault judgment caused by single stopping criteria.(2) This thesis proposed a fast and robust binary feature matching method, and applied it to recognize occluded workpiece. By combining the advantages of random sampling pattern and fixed-point sampling pattern, this method can obtain global optimum point-pair nearby local optimum point-pair. Meanwhile, during the process of offline training, a more reasonable point-pair identification model is built by taking both the mean distance and the points relevant into account. What’s more, we use BRISK feature point detection method to obtain scale invariant feature points. In order to estimate the direction of the feature points, we use the simple and efficient gray centroid method.Experimental results show that the algorithm introduced in this thesis can recognize the occluded workpiece effectively. |
PDF Full Text Request