Vision-based pedestrian taxi calling behavior identification method

Abstract

The invention discloses a pedestrian taxi calling behavior recognition method based on vision. The pedestrian taxi calling behavior recognition method comprises the following steps of image preprocessing and intention reasoning. According to the method, a computer vision method is adopted to accurately and efficiently identify pedestrians with taxi calling behaviors from the images, passengers can be more efficiently found in the automatic driving taxi, the use efficiency of the automatic driving taxi is improved, and the travel efficiency of the passengers is also improved. According to the method, reasoning of the passenger taxi calling behaviors is achieved through the spatial reasoning network, dependence on time dimension information is reduced, compared with a traditional behavior recognition algorithm, the time feature extraction process is reduced, the network can be simplified, and the real-time performance of behavior reasoning can be improved. According to the method, a set of logically explainable fusion rules is adopted, fusion of the random forest and the graph convolutional network is realized, the logically explainable characteristics can improve the environmental adaptability of the algorithm and the accuracy of behavior recognition, and more stable and accurate reasoning of the fusion algorithm on the passenger taxi calling intention is realized.

Description

technical field [0001] The invention belongs to the field of vehicle intelligence, and in particular relates to a method for identifying pedestrian behavior intentions by a self-driving taxi. Background technique [0002] The behavior of vehicles recognizing pedestrians in traffic scenes belongs to the category of vehicle intelligence. Accurate and effective identification of pedestrians' intention to call a car can help self-driving taxis quickly find pedestrians with the intention of calling a car on the road, which is important for improving the travel efficiency of pedestrians and the use efficiency of self-driving taxis, and avoiding traffic congestion significance. [0003] Pedestrian calling behavior recognition refers to the use of computer vision methods to analyze pedestrians in traffic scenes and find pedestrians with the intention to call a car. Traffic scenes are highly complex, and the number and types of traffic participants (including pedestrians, vehicles,...

AI Technical Summary

Problems solved by technology

Based on the above two characteristics, traditional behavior recognition algorithms based on 3DCNN (3D Convolutional Neural Network) and LSTM (Long Short Term Memory Network) cannot be applied to reasoning of car-hailing intentions with instantaneous characteristics.

Pedestrian gestures are the key information to express pedestrian intentions, and most current gesture recognition algorithms are mainly used in indoor scenes, and vision-based gesture recognition algorithms require high resolution of hand contours in images, but smart cars equipped with Vehicle-mounted cameras cannot produce such high-quality images in complex traffic scenes

Images