Research On Workpiece Recognition Based On Binary Descriptor

With the rapid development of image processisng,workpiece recognition based oncomputer vision technology has also attracted much attention.Workpiece recognition is an important part of industrial automation production,which is mainly through the combination of computer vision technology and industrial robots to achieve industrial production line on the target workpiece automatic recognition and classification.Workpiece recognition based on computer vision has made great progress and applied in industry relevant scenes,including automobile assembling,device installing and so on,which has a great industrial application value and bright prospect.This thesis focuses on workpiece recognition based on binary descriptor.The major works of this thesis are as follows:(1)To improve the instability of rotational invariant taken from the using of dominant orientation in SURF descriptor,a binary recognition method of rotation invariant descriptor and SURF feature detection algorithm is presented.Binary descriptor adopts the idea of rotating sampling and is unnecessary to calculate the dominant orientation,the sampling area and the eigenvector are rotation invariance,avoiding the wrong match caused by the wrong dominant orientation.The experimental results demonstrate that the method can improve the poor rotational robustness caused by the dominant orientation technique of SURF descriptor.(2)To improve the low recognition rate of FREAK algorithm under scale transformation and rotation transformation,an improved workpiece recognition method based on improved FREAK descriptor and Fast Hessian feature detection operator is given.Fast Hessian feature detection operator enhances the recognition ability of FREAK descriptor under scale transformation.A new circular sampling pattern is proposed by reducing the overlapping area of the sampling point smoothing range.The sub-region of the image is divided by the gray-level sorting idea,and the rotation invariance is independent from the direction of the feature point.When sampling in a sub-region,by setting the threshold of the gray value,the influence of the background is weakened,and the performance of the descriptor is improved.In the description of sub-matching sub-region cascade matching method,and then use RANSAC algorithm to eliminate the mismatch,both to improve the matching rate and the accuracy of matching.The experimental results demonstrate that the improved method enhances the distinguishing ability of FREAK descriptor and improves the recognition performance of the FREAK descriptor to the workpiece.