Vehicle Identification Of Hazardous Chemicals Based On Deep Learning

Hazardous chemicals vehicles are a dangerous source of transportation.There are great safety hazards in highway transportation.According to statistics,77% of hazardous chemicals accidents occur during transportation.In recent years,deep learning has been widely used in the field of image classification and target detection,providing a new technical basis for the monitoring and management of hazardous chemicals road transportation.Therefore,deep learning technology can be used to identify hazardous chemicals vehicles on important road sections,monitor the transportation of hazardous chemicals vehicles,improve the safety of transportation of hazardous chemicals vehicles,and ensure the safety of life and property during road transportation.This paper focuses on the vehicle identification technology and application of hazardous chemicals by monitoring the images of hazardous chemicals vehicles captured in video images.We first analyzed the feature extraction algorithm and classification method of current vehicle identification in this paper.The Histogram of Oriented Gradient and Scale-invariant feature transform commonly used in traditional image feature extraction algorithms are also analyzed,and we summed up their feature extraction methods and features as well.In addition,we gave an overview of the most useful classification method for the SVM.We also explained the way of the structure and training process on convolutional neural networks.The Histogram of Oriented Gradient and Scale-invariant feature transform which are used as feature extraction operators in traditional features are compared,and the SVM is used to identify hazardous chemicals vehicles.Finally,we choose AlexNet and VGGNet based on deep learning to achieve the vehicle identification of hazardous chemicals.We collected a total of 15000 images to do the experiments to verify the characteristic of the deep learning network on vehicle recognition.And we also used traditional method to classify the same image set.During the experiments,the accuracy of VGGNet vehicle recognition is higher than that of AlexNet,achieved 95.4% accuracy.In traditional method,SIFT+SVM achieved 68.7% accuracy,which is higher than the HOG+SVM model.Therefore,we can come to a conclusion that deep learning can be used to do the hazardous chemicals vehicles recognition.