Driver attitude recognition method based on depth image and virtual data

Abstract

The invention discloses a driver's posture recognition method based on depth images and virtual data. The driver's joint point detection: the driver's depth image and joint point labels are used as the input of the deep learning framework Caffe to train a deep convolutional neural network model; Use the deep learning model to detect the joint point position of the driver in the image. Construction of virtual driver's head posture data set: use the modeling function Modeling of the 3D modeling software to construct the driver model; use the animation function Animation to set the driver's head movement; use the rendering function Rendering to render the driver images in batches; Carry out head segmentation processing and construct a virtual driver's head pose data set. Driver's head pose analysis: use the virtual driver's head pose data obtained during the construction process of the virtual driver's head pose dataset, and use the method of transfer learning to train the deep transfer learning model; according to the deep transfer learning model, estimate the target domain The head posture of the driver in the image is detected, and the position of the facial feature points is detected.

Description

technical field [0001] The present invention relates to the fields of computer vision and vehicle assisted driving, and more specifically, relates to a driver gesture recognition method based on depth images and virtual data. Background technique [0002] Driver gesture recognition is an important research topic in the field of vehicle assisted driving, and it is the application of human gesture recognition technology in real scenes. Image-based driver pose recognition refers to detecting the driver's upper body joints, estimating the head pose, and detecting facial feature points in a given driver image. The computer can analyze the driver's posture, and finally realize the warning of the driver's dangerous behavior, and achieve the purpose of reminding the driver. [0003] The input of image-based driver gesture recognition is usually a visible light image. Extract the features of various parts of the human body from the image, such as color, edge, contour, shape, etc., ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problems in the above-mentioned driver gesture recognition technology, such as manual extraction of image features is time-consuming and inefficient, shallow learning methods cannot learn image multi-level structural features, and the characteristics of virtual data are quite different from real application scenarios. Driver attitude recognition method based on depth image and virtual data

Images