Uav in fog identification based on simplified model
Kai Luo, Guibin Zhu, Xuanni Liu, Hao Liu, He Zhang · 2019
The identification of low-altitude and low-speed the unmanned aerial vehicle (Uav) is a popular problem in the field of computer vision. The traditional identification method can solve most problems, but there are blind spots for low-altitude Uav. By contrast, the deep learning can better solve this problem, but the effect of CNN and other methods is poor in the case of fog. In order to solve this problem, we will apply CNN to dehaze the image. At the same time, for the sake of our subsequent recognition model is more accurate, more real-time, we put forward on the basis of the residual model, using the different sizes of convolution kernels, combining the Inception block and MobileNet[1] network. Our proposed model maintain the better identify accurate rate, increasing our recognition speed. Therefore, this paper aims to train multi-scale convolution model, which can effectively identify low-altitude and slow-speed Uav in fog. Experimental results show that this method is feasible.