Hardware And Software Co-implementation Of The Character Feature Extraction System Based On ZingRDK

Image feature extraction is the key of image recognition and machine vision application.It is the primary task of image processing to quickly and accurately extract image features.However,the traditional image processing technology is not perfect enough,existing some shortcomings such as low processing speed and low power consumption performance due to the huge amount of computation of the feature extraction algorithm.This thesis uses the method of software and hardware collaborative design to implement image feature extraction based on the zing RDK platform.The platform's core chip is Zynq that includes two parts: one of them is processing logic,which implements hardware acceleration of feature extraction algorithm;the other is programmable system,which drives the IP core.By applying Xilinx high-level synthesis tools to accelerate the implementation of the algorithm for improving the effect of the feature extraction.In the HLS platform,image preprocessing,corner detection,Sobel,and Laplacian edge detection IP cores are designed.And the custom IP core was connected to the main chip to integrate the hardware platform.At the same time,this thesis completes the transplant of the embedded Linux system,and implements cross-compiling tool chain construction,U-Boot transplantation,U-Image compilation and modification of the device tree.Finally,the design system is jointly debugged with software and hardware on the ZingRDK platform.The camera collects images then under the control of the PS part the PL part processes the images in real time and displays the processing results on the displayer.The test results show that the system not only has good effect on image preprocessing and image feature extraction but also has the feature of fast processing speed,low power consumption,so that the system has a good prospect for the application of machine vision.