A Pyramid For Multi-Task Feature Selection And Its Application In Electric Power Equipment Detection

With the rapid development of society,the number of electrical equipment has increased greatly,and the demand for power supply security of power system is also increasing day by day.As an important part of the power system,the smooth operation of power equipment is very important for the power system.The state detection method of power equipment based on digital image processing technology provides an effective way for the safe operation of power grid.Based on a detailed review of target detection methods for power equipment,this paper proposes a multi-task feature selection pyramid model and applies it to target detection of power equipment.Its greatest advantage is that through multi-task learning,feature selection pyramid and deformable convolution,it improves the ability of neural network to mine image information,and has different forms of power equipment.Strong adaptability.The main achievements of this paper are as follows:(1)Aiming at the problem that single-task learning is commonly used in power equipment detection and the ability of featurachieve e mining is weak,a multi-task learning method is introduced in power equipment detection,and the balance of convergence and loss between multi-tasks is improved to the goal of maintaining the balance between tasks in the process of multi-task learning.The experimental results show that the multi-task learning method is effective.It is verified that the detection model obtained by multi-task learning is better than that of single-task learning.(2)Aiming at the problem that the existing network structure lacks redundant information for feature selection in multi-scale detection of power equipment,a feature selection pyramid structure is proposed.By adding feature selection module to the feature pyramid,the function of enhancing the feature pyramid's ability to screen feature redundant information is realized.The experimental results verify the feature selection pyramid's ability in multi-scale detection of power equipment.Improve the detection capability of the network.(3)Aiming at the problems of diversified forms of power equipment,single detection objects of existing detection network and poor adaptability of various power equiprment,the technology of fusing the first two chapters is proposed,and the pyramid network of multi-task feature selection based on deformable convolution is introduced,which improves the adaptability of neural network to various fonns of-power equipment.The experimental results verify the multi-task characteristics.The universality of pyramid selection in power equipment detection is obviously improved than that of ordinary network..