An intelligent optoelectronic information processing system and method based on accelerated processing

Abstract

The invention relates to the field of target detection and tracking, in particular to an intelligent photoelectric information processing system and method based on accelerated processing. The image sequence of the target model is collected by the infrared detector, and the image information is transmitted to the x86 host through the WLAN network port, and then the communication with the FPGA board is realized through the PCIe interface, and the improved YOLO v3 deep learning network calculation based on the FPGA is realized. Acceleration, and position pre-push based on particle filter algorithm, and ensure the computing power and low delay of the system, to achieve real-time detection and tracking of infrared targets.

Description

technical field [0001] The invention belongs to the field of image target recognition, and in particular relates to a deep learning target model recognition method based on FPGA board, YOLO v3 algorithm, particle filter algorithm and hardware acceleration. Background technique [0002] Traditional infrared image target recognition algorithms use DSP processors to realize infrared target detection and tracking based on traditional image algorithms such as Gaussian filtering and threshold segmentation. In the face of complex scenes and the presence of various infrared occlusions and interferences, it is difficult to achieve the target effective detection and stable tracking. [0003] Due to its powerful feature representation ability, deep convolutional neural network has been successfully applied in the field of target detection and recognition, and achieved good results. However, there is currently no network model specifically designed for infrared image target recognition...

AI Technical Summary

Problems solved by technology

However, there is currently no network model specifically designed for infrared image target recognition, and the convolutional neural network needs strong computing power to support image processing. In order to ensure the accuracy of image recognition, the number of parameters of the deep neural network can reach tens of millions or even billions. , the computing power of the existing on-chip processor DSP is far from meeting the requirements

Existing convolutional neural networks usually use GPUs to achieve network acceleration. However, GPU power consumption is relatively high. For example, the power consumption of 1080ti boards commonly used in deep learning is usually around 250W, which is not suitable for intelligent photoelectric information processing systems. Low power consumption requirements, therefore, in order to achieve real-time detection and recognition of infrared air targets, it is necessary to develop an intelligent photoelectric information processing system based on real-time target recognition of embedded convolutional neural networks

Images