Pedestrian joint point detection algorithm based on improved OpenPose
Aiming at the problem that target pedestrian is small,carrying information is less,and the pedestrian joint point cannot be accurately detected that in-vehicle image has,a pedestrian joint point detection algorithm based on improved OpenPose is proposed.Firstly,by increasing the resolution of the input image and increasing the image scaling ratio,the feature information of the pedestrian joint point under the vehicle perspective is captured.Secondly,the network structure is improved,convolution kernel size and using depthwise separable convolution instead of standard convolution to reduce the quantities of parameters and computation of the network model.The experimental results show that the improved network improves the accuracy of pedestrian joint point detection by 6％,and the quantities of parameters and calculation of the improved network model is reduced by 69％ and 39％,respectively,compared with the original OpenPose,detecting speed of pedestrian joint point is improved.
convolutional neural networkOpenPosedepth separable convolutionjoint point detection
NETL
NSTL
万方数据
