| The five-axis CNC machine tool generates dynamic contour errors due to the servosystem lagging during operation.In the process of measuring the error using the vision system,the high-speed movement of the machine tool will cause the captured image to produce blur,affecting the accuracy of the extraction of positioning information.Therefore,it is necessary to conduct deblurring research on motion blur images.Based on the analysis of the current linear motion blur kernel estimation methods and image restoration algorithms,the measurement principle of the vision measurement system,the importance of image deblurring in the visual measurement process,and the influence of motion blur on the measurement results are explained.This paper combines the prior information of the vision measurement system and proposes an image deblurring algorithm that is suitable for the target image to be measured.The main research content of this article is as follows:(1)In order to obtain a low accuracy for the linear motion blur kernel,this paper uses the camera frame rate,exposure time and inter-frame relationship to solve the blur angle and blur scale of each frame image.This paper proposes a method to solve the motion blur kernel based on the prior information of the more accurate visual measurement system.(2)Different images have different image characteristics for different scenes,and the image blurring process has different effects on these image characteristics.This paper analyzes the image characteristics of the measured target image,and learns that the blurring process attenuates the sparsity of grayscale and the image gradient obeys the Laplacian distribution.Therefore,the objective function model of the image non-blind restoration algorithm is established and solved.Compared with the nine non-blind restoration algorithms that have outstanding contributions in the field of image deblurring,both the natural scene grayscale image and the captured measurement target image have good image restoration quality.(3)Because the current image deblurring has not been applied to the measurement field,there is no suitable image restoration quality evaluation method in the measurement field.This paper combines the improvement of the positioning accuracy of the machine tool with the measurement target image,the structural numerical information of the captured image itself and the image quality evaluation method of the traditional structural similarity,and establishes an evaluation method of the image deblurring processing in the measurement field.Then the paper validates the proposed evaluation method through experiments.The experiments of image deblurring algorithm and evaluation method proposed in this paper have been fully experimented.Experimental results show that the average error of the vision measurement of blurred sequence images is 56.7 ? m,while the average error of deblurring sequence images is 37.0 ? m.And the evaluation function proposed increases from0.8459 to 0.8811.It can effectively reduce the impact of motion blur in the process of measuring the dynamic contouring error of machine tools by vision system. |
