Research Of Embedded Image Processing End Based On SoC

Embedded image processing system is widely used in daily life, commodities circulation, and industrial production. As an embedded system it is easy to use and low power consumption; as an image processing system it is applied in fingerprint recognition, barcode recognition, text recognition, face recognition and other fields, among those applications barcode recognition is a main discussion point of this paper. A variety of one-dimensional, two-dimensional or even multi-dimensional barcodes generated with development of barcode technology appears on various commodities, and mobile payment terminal, etc., and becomes the core of web of things and the major application field of machine vision. The most widely used of two-dimensional barcodes in our country is QR Code (Quick Response Code) invented by Japanese company "Denso Wave". What discussed in this paper is QR Code recognition using image processing, and the content includes constructing video capture terminal based on a SoC (System on Chip), developing QR Code decoding algorithm, implementing QR Code recognition and displaying recognition result of QR Code image on the client.The SoC plateform chose in this paper is DM8127, and it is introduced that this plateform, development of embedded software under Linux operate system, and methods of building development environment for the platform. Meanwhile, design DM8127 data acquisition process based on IPNC_RDK V3.8 Development Kit, using LINK to achieve a cascade to obtain YUV data from the front end MT9J003 CMOS, to process data in the follow-up sequence in intra different cores or inter cores. DM8127, as the main compute unit of the system, is a multi-core chip, and the main core Cortex-A8 of DM8127 runs Linux system controlling VPSS M3 to capture video, and Video M3 to encode video as H.264 format which is broadcasted through network via Cortex-A8. In order to control process and show the result, develop a client based on QT IDE using FFmpeg to encode and decode, and finally realize controlling the data process by self-defined protocol and network image data displayment. And then design the QR code recognition algorithm for collected images which contain the QR Code, transplante algorithm to the SoC platform to achieve QR Code recognition and displayment in real time.In the acquisition and recognition process of QR code, the collected real images generally have flaws, including deformation, tilt, uneven illumination, etc. Therefore, this paper on the basis of analyzing the structure and encoding method of the QR code designes QR code image processing method, realizes an accurate segmentation of image containing multiple QR codes, and makes correction for defect images ensuring the accuracy of decoding, and finally finish a complete recognition and decoding process of QR Code image on the SoC embedded system.