Driver key sub-region identification and positioning method based on improved Yolov3

Abstract

The invention discloses a driver key sub-region identification and positioning method based on improved Yolov3, and the method comprises the following steps: obtaining driver images in different scenes, and carrying out the segmentation of a driver from a background; inputting the segmented driver image into a convolutional neural network, establishing a thermal imaging graph of the driver throughweak learning of the weak supervised convolutional neural network, and obtaining a key sub-region of the driver through the thermal imaging graph; constructing a driver key subarea positioning modelbased on an improved Yolov3 algorithm, and training the driver key subarea positioning model through the driver key subarea; and acquiring a driver image in real time, inputting the segmented driver image into the convolutional neural network to obtain a key subarea of the driver, and inputting the key subarea of the driver into the trained driver key subarea positioning model to obtain a key subarea image of the driver. According to the invention, the key subarea of the driver can be accurately identified and positioned in real time.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to an improved Yolov3 driver key sub-region recognition and positioning method. Background technique [0002] With the rapid development of social economy, the interconnection of highways has been basically realized between cities, and there are also many expressways in cities, such as closed roads, elevated roads, etc. Motor vehicles traveling on these expressways must not only limit their maximum speed , and to limit its minimum speed, that is, the motor vehicle must maintain a certain speed, because the speed of the motor vehicle is very fast when driving on the expressway, a slight mistake is prone to major traffic accidents, resulting in car crashes, death, speeding, fatigue, etc. Driving is a major factor in road traffic accidents, especially in the driving process of school buses, shuttle buses, long-distance buses, long-distance trucks, etc., traffic accidents caus...

AI Technical Summary

Problems solved by technology

Different from other image classification tasks, the behaviors of different categories in driver behavior recognition often depend on a small action, such a small difference between classes greatly increases the difficulty of recognition. The impact of classification varies greatly. For example, the driver is holding a mobile phone. Although the body is upright, the result is still abnormal driving state. Fast and accurate identification of abnormal driving behavior

Identifying and locating areas that have a large impact on abnormal driving recognition in advance can effectively improve the recognition accuracy of abnormal driving behavior. Real-time processing of data

Images