The Extraction Studies of Self-driving Awareness Video Semantics Based on Convolution Neural Networks
Shan Wang, Junfeng Wang, Donghong Cai · 2022
We propose an extraction method on video object, combined with self-driving automobile camera. Firstly, we use the frame differencing to extract key frame, the modified Video Segmentation Algorithm reduce the computer resource requirements, increase the speed of extracting prospective object. After that, establish an automobile external environment applied for collecting self-driving-aware video to realize the intelligent analysis. Intelligent extract the key frame of prospective object, then use the Convolution Neural Networks to extract feature and analysis behavior on the key frames. To build modeling and realize the detection of self-driving obstacle, road and pedestrian, traffic lights and automobiles etc. processes the ability of learning decision-making behavior of self-driving or auxiliary driving.