Study And Implementation Of Vision Based Express Sheet Information Extraction Algorithm

With the concept of intelligent logistics first proposed by IBM,improving the intelligence and automation of logistics system has become the research focus in the field of logistics.At present,the distribution centers of small and medium-sized express companies have a low level of standardization and informatization,and there are still situations where customer information is addressed manually.At the same time,the distribution centers of express giants represented by SF and "Three Links and One Reach" are relatively intelligent,but their original automated sorting solutions can no longer meet new business needs.Based on the above background,a vision based express sheet information extraction algorithm is proposed in this paper,consisting of a vision based express sheet segmentation algorithm,and a barcode localization and barcode decoding algorithm.To be concrete,the express sheet segmentation algorithm includes extraction of mask on express sheet based on Mask R-CNN framework,the enhancement of image details based on the adaptive gray-scale scaling algorithm,the first straightening of the express sheet based on the minimum external rectangle algorithm and perspective transformation together with the second straightening of the express sheet based on the VGG-16 classification network and rotation transformation.The barcode localization algorithm first generates salient patches of the express sheet,then establishes a set of suspected barcode areas searched by breadth first search algorithm,and filters out false barcodes discriminated by the covariance matrix,thereby obtaining all the barcodes on the express sheet.The barcode decoding algorithm uses distance measurement of similar edge method to decode main barcode on the express sheet based on Code 128 symbology.Finally,the running time and memory usage of the whole algorithm,the performance of express sheet segmentation,and the accuracy of barcode localization as well as barcode decoding are tested according to the business needs of the research project.It can be seen from the experimental results that the project objectives are well achieved by the algorithm proposed in this paper.