Method and device for identifying distraction behavior of driver, terminal and storage medium

Abstract

The invention discloses a method and a device for identifying the distraction behavior of a driver, a terminal and a storage medium. Image enhancement of a driving distraction data set is carried out through a deep learning method, employing convolution, an activation function and an attention mechanism, the feature expression of an image region of interest is enhanced through the information interaction between learning channels, and as the input of the neural network, channel redundancy is reduced, and the feature expression ability is enhanced. Human body features are extracted from the images passing through the image enhancement network, expression of important features is enhanced, feature extraction of the Transform model is easier, the recognition result is more accurate, finally, wrong tag images in samples are screened through confidence learning, the recognition accuracy is improved, and the method and the device can be compatible with various automatic driving systems to realize end-to-end driving distraction behavior recognition.

Description

technical field [0001] The present invention relates to the fields of deep learning and computer vision, and in particular to a method, device, terminal and storage medium for identifying driver distraction behavior. Background technique [0002] At present, the highest level of self-driving vehicles also needs to complete automatic driving with the assistance of humans, so it is necessary to monitor the driver's attention status to ensure driving safety. Driving distraction behavior refers to: in the state of non-alcohol, drug-free or fatigue-free driving, the driver diverts his attention from the main task of driving, resulting in a decrease in the ability to perceive the environment, thereby threatening driving safety. Driving distraction behavior recognition is difficult to effectively capture driving distraction behavior due to the large differences in driving environment and driver characteristics. In recent years, image classification and behavior recognition technol...

AI Technical Summary

Problems solved by technology

[0004] Embodiments of the present invention provide a method, device, terminal, and storage medium for identifying driver distraction behavior, which solves the problem of large data interference, discarding information during feature extraction, resulting in data loss, little difference between each category, and data loss. The problem of lack of correlation on time series

Images