Research On Barcode And Character Recognition Algorithm In Complex Environments

The barcodes and characters inside the solar photovoltaic modules are the identity information of the product,which needs to be recognized and read during the production process.Due to the influence of the production process,the surface of the barcodes and characters is covered with a layer of melted film and a layer of glass,making recognition extremely difficult.To address this issue,this thesis studies barcode and character recognition technology in complex environments,mainly including the following contents:(1)The comprehensive system solution is designed.In terms of hardware design,the relevant characteristics of industrial cameras and lenses are analyzed,and the selection of cameras and lenses is completed.The types of light sources and lighting methods are studied,and corresponding lighting and light source selection schemes are designed.In terms of software design,the system development language is selected,and the human-machine interaction interface is designed.(2)The image preprocessing algorithms are studied.Different enhancement and filtering denoising algorithms are compared,and a combination of median filtering and Gaussian filtering is used to suppress noise in the image.The grayscale stretching algorithm is selected to improve the contrast of barcodes and characters,and a segmented linear grayscale transformation is used for secondary enhancement of the character region.Three major segmentation algorithms for image segmentation are studied,and the maximum inter-class variance method is ultimately chosen as the foundation for subsequent barcode reconstruction and character recognition algorithms.Several edge operators are introduced,and the Sobel algorithm is chosen for edge detection in the image.(3)The localization and recognition algorithm of barcodes is studied.In the barcode positioning stage,a barcode positioning algorithm based on edge direction consistency is proposed.Firstly,the improved Sobel edge detection algorithm is used for edge detection and gradient calculation,and the edge map is divided into 16*16 small squares to obtain candidate blocks.Then,the candidate block detection and filtering algorithm is screened to obtain the candidate block area that truly contains the barcode image.Then merge the retained connected subdomains to locate the barcode area.Finally,the Hough transform is used to tilt the positioned barcode,and the barcode area is screened twice by using the aspect ratio of the barcode.In the stage of barcode reconstruction and recognition,a barcode reconstruction algorithm based on grayscale projection curve is proposed.Firstly,the segmented barcode area is projected vertically,and the gray average of each column pixel of the image is obtained.Then,each column of gray values is compared with the threshold of the maximum interclass variance method of the barcode area to obtain a new barcode binary image composed of 0 and 255.Finally,Halcon is used to identify the reconstructed barcode and output the coding information of the barcode.(4)The localization and recognition algorithm of characters is studied.Firstly,the corrected original image is horizontally segmented based on the known barcode information,and a combination of morphological operations and region feature extraction is used to locate the character area.Then,a prior knowledge-based equidistant character segmentation algorithm is proposed to segment the characters.Finally,an SVM classifier is used to train and recognize the segmented characters.After completing the above research,corresponding experimental tests are conducted using this system.Analysis of the test results revealed that the barcode recognition accuracy is98.4%,with an average recognition time of 47 ms.The character recognition accuracy is 98.67%,with an average recognition time of 32 ms.