基于Jetson TK1和深度卷积神经网络的行人检测

Pedestrian detection based on Jetson TK1 and convolutional neural network

近几年随着ADAS(高级驾驶辅助系统)以及无人驾驶等技术的逐渐兴起,作为ADAS和无人驾驶的关键技术,行人检测成为计算机视觉领域的热点问题。文中针对目前汽车电子系统大多是低功耗的嵌入式系统,提出采用低功耗嵌入式设备为平台配合深度卷积神经网络实现行人检测系统。实验结果表明该系统具有较高的识别率和较快的检测速度,基本满足实时性。

In recent years,with the gradual rise of ADAS（ advanced driving assistance system） and autonomous driving technology,as the key technology of ADAS and autonomous driving,pedestrian detection has become a hot issue in the field of computer vision. Due to the current automotive electronic systems are mostly low power embedded systems,this paper proposes a low power embedded device to realize the pedestrian detection system based on convolution neural network. The experimental results show that the system has higher recognition rate and faster detection speed,and it can meet the real-time performance.