Research On Multi-view Motion Deblurring Method

In the field of modern manufacturing,machine vision detection and recognition technology is the key to realize automatic and intelligent?In the practical applications,it is usually adopt spatial multi-view stereo imaging system to get the spatial information of the object.At the same time,in order to improve the production efficiency,it have to make objects still in the state of high speed movement.Due to the restrictions of exposure time,the obtained image will appear motion blur which cannot produce enough target characteristics,and it will seriously affect the accuracy of detection and recognition.Then it need to deblur the blurred image to restore the target characteristics.Existing deblur technology cannot meet the requirements of actual production application very well,\because the same feature points of the multi-view images only mapped to the same position in three dimensional space,and the existing deblur method has destroyed the space mapping relationship.To solve these problems,a space multi-view moving target image deblur method is proposes in this paper,it ensure the space consistency of the same feature points in the multi-view the image while improve the blurred image clarity.According to the practical application scenarios of industrial product testing,a multi-view image acquisition system is designed and implemented in this paper.In order to guarantee the quality of the image,a set of high resolution industrial cameras are chosen in this system.Based on the camera interface provided by the official,a set of imaging system control software is develop to realize the real-time debugging for the image acquisition process,and a stable development environment is build for the subsequent image processing.For the different blurred imaging that captured from each imaging plane,If it want to ensure corresponding relation of the same feature points in the each viewpoints image with the same space consistency,it must build the internal relations of blur kernel m multiple viewpoints.Firstly,estimate the blur kernel of some viewpoints of blurred image(based on the two viewpoints for example),then use Zhang camera calibration method to get the parameter of the camera model.With estimated blur kernel of partly viewpoints and camera model,the relationship function of multiple viewpoints blurred image blur kernel could be build.Then using leastsquare error criterion to ensure the correspondence of multiple viewpoints blur kernel and optimize blur kernel for each viewpoints of image accurately.Machine vision detection system has strict requirements on real-time for deblur algorithm.this paper analysis the characteristics of linear motion blurred image,and derived the process that convert two-dimensional blur model into process of one-dimensional blur model,and proposed a fast deblur method based on linear blur kernel.