Research on ZYNQ Image Acquisition and Processing Method for Target Recognition

When CNN and other computer vision algorithms process high-resolution images, they often need to adapt to specific input specifications. Therefore, real-time and efficient image scaling has become an important part that supports the operation of the algorithms. Based on ZYNQ7100 heterogeneous FPGA platform, by integrating the MIPI CSI-2 interface of OV5640 with the programmable logic (PL) and ARM processing system (PS) of ZYNQ7100, this paper optimizes the bilinear interpolation algorithm using a hardware parallel processing architecture and designs a dedicated image scaling IP core using Verilog language. The experimental results show that the design requirements have been met.