Design of Video Acquisition System for Construction Transporter Self-driving on Deep Learning

Abstract

When using deep learning technology to achieve foreground detection for Unmanned Ground Vehicles (UGV), its visual real-time processing tasks need to be completed with customized embedded platforms. A large amount of reliable visual data for deep learning is provided to a video acquisition system. We present a video acquisition system for deep learning in construction transporter self-driving application to effectively shield electromagnetic interference in various frequency bands, cope with complex scenes and different types of light pollution, simplify the processing of original image data by the visual controller and the transmission mode of installation wiring, and provide a solution with high stability, high bandwidth, high reliability, long distance and low delay for image data transmission.