Real-time object recognition algorithm based on deep convolutional neural network
Lihong Yang, Liewei Wang, Shuo Wu · 2018
Radar detection of moving objects is vulnerable to external environment. By introducing the object confirmation algorithm, the intelligent radar perimeter security system's false alarm rate can be reduced significantly. The object confirmation algorithm is essentially an object detection algorithm. Due to the poor generalization of artificial feature extraction algorithms, we use the deep convolution neural network to extract deep features automatically for object confirmation. In order to meet the real-time requirement in engineering practices, our algorithm use the YOLOv2 system as a basis, and selects anchor boxes which meet object scales of our training data set by k-means++ clustering. To improve the YOLOv2 network structure, low-layer deep features which denote the texture information and high-layer deep features which denote the semantic information are combined layer by layer to make object detection more accurate. The experimental results show that the false alarm rate of the intelligent radar perimeter security system is further reduced by introducing the object confirmation algorithm. Especially for extreme weather, false alarms of radar detection greatly increase. But most of them are eliminated after running object confirmation algorithm. Therefore the warning accuracy of the entire system can be guaranteed. The detection speed of the object confirmation algorithm is 33FPS, which meets the real-time requirement of engineering practices.