Method and device for accurately automatically identifying driver unsafe behavior based on convolutional neural network

Abstract

The embodiment of the invention discloses a method and device for accurately automatically identifying a driver unsafe behavior based on a convolutional neural network, and relates to the fields of image recognition, pattern recognition and automation. In the case of the problem of unsafe behavior detection of important cargo transportation or the driver of a passenger car, the method of DriverBeCog is provided; the method comprises the steps of using a capsule neural network and the convolutional neural network to extract real-time image characteristics of the driver; respectively performingtwo classifications on various behaviors in parallel; using a monitoring device back screen for real-time monitoring, issuing a warning for the unsafe behavior, and recording relevant information intoa database. The method adopts multiple models concurrent processing, and proposes the convolutional neural network model with few network layers, few parameters, small calculation amount and easy practicality; the recognition rate of the unsafe behavior exceeds existing achievements, which is beneficial to practical use; a weight parameter adjustment method solves the problem that a positive andnegative data volume gap is too large.

Description

technical field [0001] The invention relates to the fields of pattern recognition and safe driving behavior monitoring, in particular to a method and device for automatically identifying driver's unsafe behavior. Background technique [0002] As of the end of 2016, the number of motor vehicles in my country reached 290 million, including 194 million cars, 360 million motor vehicle drivers, and more than 310 million car drivers. A large number of cars have caused a large number of traffic accidents, and many of these traffic accidents are caused by uncivilized and unsafe traffic behaviors, such as smoking with one hand or making a phone call, and controlling the car with one hand; Insufficient awareness, there are still a large number of cases of not wearing seat belts. For important cargo transportation or passenger vehicles, the harm caused by accidents is huge. Therefore, society has a strong demand for systems that can automatically identify driver violations and issue ...

AI Technical Summary

Problems solved by technology

However, a single neural network structure cannot handle complex actual situations, and its accuracy rate is unsatisfactory, requiring a more complex and advanced algorithm and model structure to complete

Images