基于区域的卷积神经网络在空对地车辆检测中的应用

摘要

针对传统空对地车辆检测算法在光照变换、场景变化时检测效果不佳的问题,提出了基于区域卷积神经网络Faster RCNN模型的空对地车辆检测方法,介绍了Faster RCNN模型以及模型训练过程.实验结果表明,基于Faster RCNN的空对地车辆检测方法是可行的,对不同光照和场景下的车辆检测可以取得较好的效果.%Considering the deficiency of traditional airborne vehicle detection algorithms in illumination changing and scene variation,the airborne vehicle detection method based on region-convolutional neural network(Faster RCNN)is pro-posed.The faster RCNN model and its training procedure are described in this paper.The experiment results show that the airborne detection method based on Faster RCNN is feasible,good results can be achieved in different illumination and scene condition.