Driver distraction detection method

Abstract

The invention discloses a driver distraction detection method. Converting each frame of driver image into a grayscale image; the method comprises the following steps: firstly, extracting gray scale images corresponding to training samples, performing normalization processing and preprocessing in sequence, inputting one of the training samples into an initialized convolutional neural network, performing batch regularization processing on HOG features extracted from the gray scale images corresponding to the training samples, and then performing full connection layer connection to obtain HOG feature vectors; and finally, carrying out global mean pooling on an output result of each convolution layer to obtain a total feature vector composed of the feature vector and the HOG feature vector, and carrying out full connection layer and Softmax classification in the convolutional neural network in sequence to obtain a global mean pooling result of each convolution layer. Obtaining the actual action category of the driver so as to update the parameters in the convolutional neural network; and updating the convolutional neural network in sequence by adopting the same method. And finally, obtaining an action category corresponding to the driver image in the test set. The detection result of the detection method is more accurate, and the network structure has fewer network parameters.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a driver distraction detection method. Background technique [0002] In recent years, as the number of private cars continues to increase, traffic accidents are also increasing, a large part of which is caused by driver distraction, such as: drivers answering calls, drinking water, taking things and other distracting actions during driving prone to traffic accidents. Therefore, it is necessary to detect the driver's actions in real time and remind the distracted driver in time so as to effectively avoid the occurrence of safety accidents. [0003] For example, a Chinese invention patent with the application number CN201910532626.X (the application publication number is CN110363093A) discloses a driver action recognition method and device, including: by acquiring the image of the current driver; In the product neural network and the three-dimensional convolutional neural network,...

AI Technical Summary

Problems solved by technology

However, the huge amount of network parameters of the convolutional neural network makes its training model difficult to use for real-time detection, and the existing convolutional neural network model only focuses on the output of the last layer of the network and fails to make full use of the output characteristics of the middle layer. Upper middle layer features contain a lot of useful information

Images