Image processing method and system based on convolutional neural network FPGA acceleration

The invention provides an image processing method and system based on convolutional neural network FPGA acceleration, and relates to the technical field of image processing, initial image data are acquired, and after the initial image data are preprocessed, initial data FPGA loading of a convolutional neural network is carried out; accelerated calculation of initial data on an FPGA is realized through a convolutional neural network, and on-chip storage management of images is realized; wherein the parallel computing comprises the following steps: establishing an annular buffer region between different convolutional layers, enabling the annular buffer region to correspond to on-chip image storage of an FPGA, enabling image data flow between the convolutional layers not to pass through an external storage device, carrying out parallel computing on the convolutional layers from the direction of an output channel of a convolutional neural network, and changing an image data flow sequence of the convolutional laye