Design And Implementation Of The Cheque's Multi-Spectral Image Capture And Recognition Software System

Computer technology is increasingly advanced and more widely used in nowadays.More and more simple daily financial businesses processed by manual are changing into automated electronic processing.However,the promotion degree of digital business of cheque used in financial transactions between companies is still very low.Most banks are still refer to using manual processing and recognition due to the lack of a mature cheque image capture and recognition commercial system.And the image capture equipment(camera or document scanner)used to archive cheque images only store cheque's white light images rather than multi-spectral images.Thus,a computerized collection and recognition software system for automatic collection,archiving,identification and pseudocounterfeiting of cheque's multi-spectral images is very important for improving the working efficiency of banks as well as enhancing the payment security.According to above situations,this paper proposes a general multi-spectral image collection and recognition software system designed for cheque.CIS is controlled by FPGA to capture cheque images contain various spectrums including white light,infrared light and ultra violet.Then FPGA wraps the image data and transports it to the upper computer through the USB3.0.After receiving the image packets,the computer software resolves the wrapped data and generates the original image according to the agreed protocol.Image correction,segmentation,standardization algorithms are designed based on the characteristic of CIS and cheque collection software interface is designed based on the user requirements when checking and archiving images.Furthermore,we design a character segmentation algorithm based on gray projection according to the characteristics of infrared reflection cheque image and the character identification is realized by the skeleton extracting algorithm.At last,an improved LBP algorithm based on Gaussian model for texture detection is proposed to distinguish the real cheque from fake cheque.Experiments are realized on cheque to test the effectiveness of the designed multispectral image collection and recognition system.Experimental results demonstrate that our system can capture various cheque multi-spectral image images.The image we acquire is in higher quality and our system works stably with a sound recognition accuracy.The system described in this paper has been delivered to the “SHENZHEN INFOGEM TECHNOLOGIES CO”.