Detection Model And System Implementation Of Granary Storage Quantity Based On Deep Learning

Keeping instant track of gross grain reserves is greatly involved with the harmonious and stable development of our country.At this stage,the weighing measurement method used in China's total granary storage inspection has long been unable to meet the current needs of our country's modernization,due to its low efficiency and demanding workload.Therefore,to develop an intelligent granary storage detection system is of great significance for the modernization of granary management and the long-term stability of the country.With in-depth study of stored grain quantity detection and taking into consideration the characteristics of stored grain quantity detection,this paper proposes a new granary storage detection model based on deep learning.The main contents of this paper are as follows:(1)Put forward a kind of grain storage quantity detection method which combines deep restricted Boltzmann machine and support vector regression machine.This method takes the mean output values of the pressure sensor on the inward and outward circles as input,uses the deep restricted Boltzmann machine to reconstruct the sets of the input data,and then applies the support vector regression machine to visualize the sets.Through theory and practical application the method is proved to be feasible and effective.(2)In view of possibilities of abnormal value and accidental value from the pressure sensor output,the pauta criterion is introduced and a data preprocessing method is proposed to avoid the influence of abnormal value and accidental value on the detection result;the detection accuracy of the detection model is in this way further improved.(3)By the light of modular programming idea,a grain storage quantity detection system based on deep learning is designed and realized,which has many functions such as data processing,online detection,query,modeling and so on.The system is used to detect the quantity of stored grain of actual granaries and the results show that the system developed in this paper has high detection accuracy and can meet the national detection requirements.