Pedestrian road-crossing intention recognition method based on gaze detection and traffic scene recognition

Abstract

The invention relates to a pedestrian road-crossing intention recognition method based on gaze detection and traffic scene recognition. The method comprises the steps of firstly carrying out the pedestrian detection through a faster-RCNN, further searching an interested target through the motion information of a pedestrian, and extracting the motion sequence of the interested target and a local traffic scene sequence, secondly, detecting the head position in the interested target sequence through a head detector, judging the head orientation through a residual convolutional neural network, then detecting whether pedestrians watch coming vehicles or not, and meanwhile extracting head orientation features, then performing multi-label classification on a local traffic scene where the pedestrian is located through a convolutional neural network classifier, and extracting local traffic scene features, and finally, processing the spliced head orientation features and local traffic scene features by using a full connection layer, and identifying pedestrian crossing / non-crossing results. According to the method, a multi-task learning mode is used, and gaze / non-gaze recognition, traversal / non-traversal recognition and traffic scene classification can be completed in an end-to-end mode.

Description

Technical field[0001]The invention relates to the field of intelligent transportation technology, in particular to the field of pedestrian detection and analysis, and in particular to a method for identifying pedestrians' crossing road intentions based on gaze detection and traffic scene recognition.Background technique[0002]With the continuous development of computer vision, sensor technology and control theory, unmanned vehicles have received extensive attention and have bright application prospects. Unmanned vehicles can bring convenience to people's lives, but they need to guarantee the rights of other road users, especially the safety of pedestrians, during their smooth operation. In order to better protect pedestrians, unmanned vehicles need to understand the behavior of pedestrians and adjust their strategies accordingly. Among the many behaviors of pedestrians, pedestrians cross the road most frequently and are highly related to pedestrian safety. Human drivers can quickly d...

AI Technical Summary

Problems solved by technology

Due to the randomness of pedestrian behavior, neither machines nor humans can predict with complete accuracy whether pedestrians will cross the road

Images