Research On Deep Learning Target Detection Algorithm In Mobile Terminal

With the enhancement of people’s awareness of environmental protection,the habitat of birds has gradually widened,but the frequent activities of birds have seriously affected the safe and stable operation of transmission lines.Traditional transmission line bird damage management methods have a good bird repellent effect in the early stage of application,but as time passes,birds gradually adapt to bird repellent devices,resulting in the failure of bird repellent functions.In order to improve the shortcomings of traditional bird repellent methods,this paper uses deep learning algorithm to detect birds near the transmission line,and then controls the start and stop of bird repellent devices according to the detection results,weakening the adaptability of birds to bird repellent devices,so as to achieve the purpose of real-time,effective.energy-saving and environmental protection.However,the deep learning model has a large amount of calculations and generally needs to run on a server with better performance.If the collected images are transmitted to a remote server,network delay will occur,and the detection results may be lost during the transmission,which will affect the drive Bird effect.Therefore,this paper proposes a deep learning target detection algorithm suitable for mobile terminals to solve the problem of bird damage management on transmission lines.The article first introduces existing deep learning object detection algorithms and deep learning model compression methods,and discusses the difficulties and solutions of analyzing deep learning object detection algorithms on mobile terminals.Then,the limitations of traditional target detection algorithms are studied and analyzed from the development process,algorithm idea,and network structure and model performance of related models.Then,based on the idea of deep separable convolution,the feature extraction network of Yolo V3 algorithm is improved,and the target detection algorithm suitable for mobile terminal is proposed,and the compression effect and feasibility of algorithm model are analyzed from the theoretical level.Finally,from the practical point of view,the algorithm is applied to the bird detection task of transmission lines.The experimental results show that the accuracy of the model can reach 83.67%,and the detection speed can reach 63fps,which can meet the accuracy requirements and real-time requirements of the bird detection task of transmission lines.It has a positive significance for bird pest control of transmission lines.The research results have a broad application prospect.