基于深度学习的行人过街意图中人脸检测和姿态估计分析

随着计算机视觉技术的发展,在人工智能领域也开展了很多以行人为目标的研究。本文基于深度学习,对行人进行人脸检测和姿态估计,为行人过街意图的进一步研究分析打下基础。本研究利用TensorFlow-SSD进行行人目标检测,分为两部分内容。一是检测行人目标,进行姿态估计分析动作,二是检测行人脸部,用来配合姿态估计对行人运动方向进行分析。采集数据后上传至服务平台后端,其调用OpenCV读取图片,通过TensorFlow提供的api读取pb文件,传递给训练好的检测模型,然后进行人脸检测和人体姿态检测与估计。

Analysis of Face Detection and Pose Estimation When Pedestrians are Crossing the Street Based on Deep Learning

With the development of computer vision technology,many researches targeting pedestrians have been carried out in the field of artificial intelligence.Based on deep learning,this paper conducts face detection and pose estimation of pedestrians,laying a foundation for further research and analysis of pedestrians’intention to cross the road.This research uses Tensor Flow-SSD for pedestrian target detection,which is divided into two parts.One is to detect pedestrian targets and perform pose estimation and analysis actions,and the other is to detect pedestrian faces to analyze the direction of pedestrian movement in conjunction with pose estimation.The collected data is uploaded to the back end of the service platform,which calls Open CV to read the picture,reads the pb file through the api provided by Tensor Flow,and passes it to the trained detection model,and then performs face detection and human pose detection and estimation.