CNN-LSTM Recognition Model of Vehicle Load State with Dynamic Focusing Loss
In order to monitor the operation of the vehicle in real time,the CNN-LSTM recognition model of vehicle load state with dynamic focus loss is introduced.The selected original vehicle feature data and k-step differential features were used to form the preliminary feature vector matrix,and the convolutional neural network(CNN)and long short-term memory neural network(LSTM)were used to extract the local and global features of the preliminary feature vector matrix to form the final feature vector,and the fully connected network was further used to identify the final feature vector as loading,unload-ing and operation.The dynamic focused loss function is introduced into the model training to balance the sample loss weight and increase the loss weight of the decision boundary.Experimental results show that the accuracy of the CNN-LSTM recognition model is improved by 17.74%,3.97%and 2.81%respectively compared with the support vector machine(SVM),convolutional neural network and long short-term memory neural network recognition model.
vehicle loadconvolutional neural networkslong short-term memorytime series analysisdynamic focal loss function
NETL
NSTL
万方数据
