Grain Condition Monitoring System Based On Multi-Sensor Data Fusion

Food security is crucial to maintaining social stability,and grain storage is one of the most important ways to ensure food security in our country.With the increasing annual grain output in our country,the problem of grain safe storage is also becoming more and more serious.Every year,our country loses about 5% in grain storage.The loss of grain storage is mainly caused by the unsuitable environment in the granary,which makes the grain mildew,germinate,breed bacteria,etc.In order to solve the above problems,this thesis constructs a grain condition monitoring system based on multi-sensor data fusion.On the one hand,it adopts wired and wireless dual communication methods to collect the important parameters affecting the storage status of the grain in real time and make the environment suitable by controlling the external equipment.On the other hand,it establishes a two-level data fusion model,which can fuse the environmental information in the granary and make comprehensive decisions.This thesis first analyzes the current development status of domestic and foreign granaries.According to the requirements of different granaries,it designs a wired communication solution combining RS485 and Ethernet and a wireless communication solution combining ZigBee and LoRa.The grain condition environment monitoring system is mainly composed of three parts: data acquisition layer,central processing layer,and remote management layer.The data acquisition layer mainly collects the environmental parameter information of different areas in the granary in real time.The central processing layer mainly receives,displays and controls the collected environmental parameter information.The remote management layer is responsible for displaying and storing the data from multiple regional center processing layers,and remotely controlling center processing layer intervene in the environment.Secondly,this thesis designs a two-layer data fusion model.The first layer of similar sensor data fusion uses the box plot method in statistics to identify outliers in the sensor data,and then uses the radial basis neural network model to correct the outliers instead of simply removing them to ensure the integrity of the data.In order to solve the problem that the traditional RBF neural network model training method is easy to fall into local optimum,this thesis designs a particle swarm optimization algorithm to optimize the neural network,and uses adaptive weighted average method to calculate the local fusion results of similar sensors.In the second layer of heterogeneous sensor data fusion,this thesis uses D-S evidence theory as the decision level fusion method,and uses fuzzy sets to construct the basic probability distribution function.Aiming at the problem that the generated evidence will produce conflicts,this thesis designs a method based on the angle cosine to judge the conflict source and the distance entropy to remeasure the basic probability distribution function to solve the problem of evidence conflicts,and obtain the global fusion result.Finally,this thesis tests the function of each part of the system.The test results show that the grain condition monitoring system designed in this thesis meets the actual needs,has good stability and reliability,and can meet different needs of granaries.