Research On Specific Target Recognition Technology In Industrial UAV Applications

The identification of technical vehicles is aimed at solving the safety problem in industrial transport: With the development of drone technology,the field of industrial transportation has undergone tremendous changes,and the progress of deep learning and the improvement of GPU hardware equipment are for technical vehicle categories accepted.Provides a powerful boost.In the industrial transport environment,the characteristics of technical vehicles are complex and video surveillance and manual identification are far removed from actual needs.The use of industrial UAV has led many people to see the dawn in many areas.By staging industrial UAV,people can capture information about technical vehicles in a timely manner and make operation more comfortable.Therefore,the deep learning method and the industrial UAV are applied to the identification of industrial transport vehicles that can classify and identify the construction vehicles well and play a role in the realization of the intelligibility of the industrial transportation.The identification algorithm of this document is based on the current industrial transport environment and the new model TD-CNN has been designed.The TD-CNN model is applied to the category identification of design vehicles.The main purpose is to improve the recognition accuracy.Based on the original CNN network,this document first improves the neural classification network and then integrates the regional extraction network to design a new network model TD-CNN.The present article focuses on the improvement of the structure of the neural network of classification folding.The specific model of the neural network classification is determined by determining the size of the convolution kernel,increasing the number of kernels,designing the mesh depth,adding special layers,and specifying the hypotheses for classifiers.To improve the recognition rate of the new model,various optimization strategies were used to correct and adjust the parameters,which proved that the addition of batch standardization is better than regularization optimization.Finally,the TD-CNN network structure is constructed by combining the region extraction network and then integrating the improved neural classification network.Secondly,due to the serious lack of aircraft records for technical vehicles and the difficulty of resource acquisition,this article compiles technical vehicle data sets based on the research laboratory’s industrial UAV that record videos and extract video frames.Earth car,water truck,cargo.In this document,the vehicle data set is tested with the new model TD-CNN,and finally the recognition accuracy for the classification of the development vehicle is determined.Finally,this paper combines the industrial background,using Caffe’s architecture and the GPU workstation to configure the experimental environment,performs test simulation studies on TD-CNN by the aircraft data data set,and then compares the experimental results of the recognition accuracy of various construction vehicles and the recognition accuracy.The final experimental results show that the overall recognition accuracy of the algorithm is 98%,which is better than the conventional algorithm.The use of the new TD-CNN model will have a number of implications for the identification of design vehicle categories.