High-speed Real-time Infrared Image Target Detection Method Based on Intelligent Platform

With the rapid development of infrared detectors, the high-speed detection data has grown exponentially, which bring great challenges to the real-time target detection processed by back-end. Aiming at this problem, we complete the target fast detection experiment in the sequence oriented infrared image based on YOLOv5 and intelligent platform, which is composed of cpu and smart chip. According to the experimental results, for 8K×8K 16bit quantized high-speed detection image can basically meet the real-time processing capability of Gb/s level, and for conventional images less than 2K×2K 16bit quantized high-speed detection image can basically realize the ability of simultaneous processing of less than ten detection images. According to the results of our experiment, the high-speed real-time infrared image target detection method based on intelligent platform will have a certain application in the actual space-based detection and the application processing.