Application Research Of Pattern Recognition Algorithm Based On Motion Imagination In UAV Environment Monitoring And Control

With the development of modern industrial civilization,people are not only enjoying modernization,but also being harmed by environmental pollution.Therefore,it is extremely urgent to improve environmental monitoring technology to monitor and prevent environmental problems and improve the natural environment on which we live.In recent years,UAV has developed rapidly in the field of environmental monitoring due to its advantages of hover,easy operation,low operating cost and flexible control.However,unmanned aerial vehicles on the market are commonly controlled by humans and cannot free their hands.As a new type of human-computer interaction system,motion imagination brain-computer interface can combine with unmanned aerial vehicle technology to realize "mind control" UAV in a real sense.The brain-computer interactive system can realize the control of multiple unmanned aerial vehicles for different tasks by a single person,which means that the monitoring range is larger and the working efficiency is higher than that of a single unmanned aerial vehicle.Aiming at the problems of inaccurate flight control and low classification accuracy of pattern recognition algorithm in brain-controlled UAVs,this paper optimizes the neural network and pattern recognition algorithm of long and short term memory unit in order to improve the accuracy of flight movement control of UAVs by brain and complete the task of precise positioning and monitoring of UAVs.This paper studies how to realize the goal of accurately recognizing the consciousness of the operator in a brain-controlled UAV.Brain control of unmanned aerial vehicle concrete realization way is: the motion picture signal event related synchronization/synchronization characteristics,extract the operator imagine UAV motion characteristics of the up and down or so as input of classifier,and the flight control of unmanned aerial vehicle in the brain command into classification results output,the output through wireless communication module to transmit to the UAV flight control system to control the movement of the unmanned aerial vehicle,finally realizes the brain remote mind control of unmanned aerial vehicle.In order to achieve the above research objectives,this paper analyzes the characteristics,preprocessing,feature extraction algorithm and classification and recognition algorithm of motion imagination signals,and focuses on the research of four different classification and recognition algorithms of motion imagination signals: LDA,SVM,CNN and LSTM.The linear discrimination and support vector machine classification algorithms were used to classify and recognize the extracted frequency band energy features and CSP features,and the highest classification accuracy was 85%.The experimental results showed that the LDA and SVM with single features had poor classification performance.Therefore,a multi-feature fusion method is proposed in this paper.The extracted frequency band energy features and CSP features are fused,and then the LDA and SVM algorithms are respectively used for classification and recognition,and the average classification accuracy of 90% is finally obtained.By comparing the results of multifeature fusion and single feature classification,the average classification accuracy is increased by about 5%,that is,the classification effect of feature fusion is obviously better than that of single feature.LDA and SVM classification algorithms only realize the left and right lateral flight control of UAV.In order to realize the upper and lower left and right flight control of UAV,convolutional neural network and long and Shortterm memory unit model are adopted in this paper.By comparing and analyzing the performance indexes of CNN model,the structure of CNN model was improved by adding weight decay,dropout,Bath normalization and LRN.The improved convolutional neural network model not only improved the generalization ability,but also achieved 88% classification accuracy.In order to accurate control of unmanned aerial vehicle flight,improve the classification accuracy of experiment using LSTM model on the structure characteristics of sequence classification,through the analysis model and comparing the improved CNN model and the LSTM model classification result shows that the combination of characteristics of tectonic sequence the LSTM model classification method of classification accuracy is higher than the CNN model,can reach more than 90%.Therefore,in this paper,the classification method of constructing characteristic sequences combined with LSTM model is finally determined to be applied in the EEG signal processing system of BRAIN-controlled UAVs.