The Image Recognition System Based On Dsp + Fpga Design And Implementation

The image processing and recognition technologies have seen rapid growth in the recent years. People are fully aware that image processing and recognition technologies are important means in understanding and transforming our world. Currently, their presence has been seen in many fields, showing a promising future in medical care, public security, transportation as well as industries.This paper studies a real-time image recognition system based on DSP+FPGA structure. The system has DSP as its core of image recognition module, which enables the image recognition algorithm and FPGA as its core of image acquisition module, which enables image acquisition and preprocessing. The control of image recognition algorithm is complex and requires high computing capacity, posing a high demand over system performance. DSP ideally meets the requirements of the system due to its unique structure and excellent performance. Due to its high speed and flexibility, FPGA enables the real-time system and simplifies peripheral circuits, thus resulting in easier system design.The recognition algorithm is mainly based on template matching and neural network, with numbers 0-9 as its targets. Template matching generally applies to the recognition of small character set (especially the character set of the same font), such as standardized numbers and characters. With simpler structure and high processing capacity, template matching contributes to higher recognition speed and accuracy, resulting in great effectiveness. The disturbed storage, high fault tolerant and the capacity of self-organized and self-learning of the neural network shows its advantages in image recognition.Study indicates that image recognition system based on DSP+FPGA structure is more flexible and applicable to general purposes, and implements modular design, thus increasing the overall efficiency of the algorithm. By fully leveraging the combined advantages of DSP and FPGA, the system could enable accurate and fast image recognition. With flexible configuration of software and hardware, the system could have most image processing functions and applicable in the fields of industrial visual inspection and automobile license plate recognition etc.