Design And Implementation Of Image Preprocessing System For Handwriting Practice

In the problem of deep learning image preprocessing,the orientation and perspective of image data in some application scenarios may affect the prediction results,and cluttered backgrounds may also cause interference,such as in the scene of automatic review of pictures of calligraphy practice assignments.In this case,in the process of preprocessing the data set,the perspective optimization of the pictures taken by the user may be useful for subsequent training,recognition,and other picture purposes.According to the development needs of the calligraphy job intelligent review system for online education platform to optimize the perspective and background of the job image,open CV and Tensor Flow Serving were used as the basic framework to detect and analyze the area of interest and projection transformation of the image,and established the automatic optimization of the calligraphy job image system.The principles of edge detection,target detection and projection transformation related algorithms are described.The function and performance requirements of the calligraphy job image optimization system are analyzed,and the processing flow of the optimization system is designed.First,canny edge detection is performed on the image,including filtering and denoising,calculating image gradient values,non-maximum suppression,double threshold filtering,and delayed edge tracking.A multi-level stitching straight line extraction method is proposed and applied.Then,target detection is performed.The common sense of the maximum extremum stable area algorithm is used to determine the area range of the target text box,and the credibility judgment is made,and whether to use the Center Net target detection algorithm based on deep learning for further target detection is determined.If a reliable localization of the region of interest can be obtained,projection transformation is performed according to the corner points of each text frame as key points,and the unnecessary background areas are removed.Finally,the centos system based on docker is used to implement the calligraphy job image optimization system.The system was tested and used on the calligraphy assignment module on the children’s online course platform,and the results achieved good results.