Further Research On Character Recognition Based On Zernike Moment Image

Optical character recognition is the key technology to realize artificial intelligence,whose application in the manufacturing industry is becoming more and more widely.The existing algorithms of character recognition can't be recognized accurately and efficiently.It is necessary to further study the recognition algorithm,improve the accuracy of character and speed up the automation process of industrial production.In this thesis,character region location based on edge detection,character segmentation based on row column scan and character recognition method based on Zernike moments are proposed by literature research,contrast analysis and experiment simulation,taking images in complex industrial environment as research objects.The effectiveness and practicality of proposed method in this thesis are verified.The main contents of the thesis are as follows:Character area location and character segmentation based on edge detection.First of all,the recognition of the character image region location,the image grayage and Gauss filter processing to weaken the influence of noise,the improved ant colony algorithm edge detection for the processed images,and the corresponding optimization research on the problem of local optimal and low efficiency of ant colony algorithm in edge detection..In the edge detection,the gray gradient of the image is used as the weight coefficient of the ant colony transfer probability to guide the ants to select the edge points to avoid the local optimal problem of the ant colony algorithm.At the same time,the ant threshold is introduced in the pheromone updating,which can make the ant gather to the edge part faster and improve the edge detection effect.Then,based on the image edge detection information,the region is located,and the character after the location is segmented by row and column scanning and character prior condition.Based on Zernike moment character recognition.For the two value after segmentation,the character Zernike moment modulus is used as the feature vector and the Zernike moment has a great influence on the edge of the character image.On the basis of the proposed outer circle mapping method,the invariant distance calculation frame is improved on the basis of the proposed outer circle mapping method for the Zernike moment optimization,and the Zernike moments due to the characteristics of the Zernike sampling and quantization.The normalization is processed,and the Euclidean distance is matched with the character library template character to output the result of character recognition.Finally,in order to verify the effectiveness of the algorithm,the container picture is used to represent the box number characters.The algorithm is applied to the scene system,and the box number is recognized and verified.The experimental results show that the recognition rate of box number is 96%.It shows that the algorithm can identify the characters in the manufacturing environment accurately and efficiently,and speed up the construction of industrial automation process.