Research And Implement Of Gesture Recognition Technology In Virtual Maintenance

As a modern information technology,virtual reality is widely concerned by various industries.The virtual maintenance realized by this technology has the advantages of high fidelity and free from hardware constraints in the field of engineering training,which is a hot topic today.As a key problem in virtual maintenance,human-computer interaction has a great impact on the improvement of user experience and the final effect of virtual maintenance.Therefore,it is an important development direction in the future to seek a natural interaction mode in the field of virtual maintenance.In recent years,with the continuous upgrading of the somatosensory devices and the rise of artificial intelligence,gesture recognition embodies its natural interactive characteristics,and gradually becomes an ideal virtual maintenance interactive technology.In the virtual maintenance scenario,the gesture library can be established according to the required maintenance actions and interactive instructions,and the corresponding gestures can be identified accurately and quickly to complete the interaction and maintenance actions.Therefore,on the basis of virtual maintenance,this paper has carried out in-depth research on gesture recognition technology,focusing on the establishment of an accurate and efficient gesture recognition model and its implementation in the virtual maintenance system.In this paper,Leap Motion is used as the body sensing equipment to build a virtual maintenance system based on gesture recognition for aircraft landing gear.The main work includes the following aspects:(1)The key technology of virtual maintenance is mainly explained by human-computer interaction,and the development status and basic principles of natural interaction technology of gesture recognition are studied.Based on the analysis of the general framework and principle of SVM,Ada Boost,BPNN and other classification and recognition algorithms,particle swarm optimization is used to jointly optimize the initial weight matrix and the number of neurons in the hidden layer of BPNN algorithm.Based on the form of Laplace kernel function,the particle fitness value and the number of iterations are integrated into the inertia weight in the process of particle swarm optimization,so that it can adapt to the process of improving the recognition accuracy and get better value.The simulation results are verified by the MATLAB platform.(2)A set of gesture libraries containing static instructions and dynamic maintenance operations was designed for the virtual maintenance process.The Leap Motion somatosensory device is used to collect relevant gesture information and extract features.After processing by PCA,a feature sequence is formed for training.According to the test results of recognition accuracy and efficiency,the gesture recognition model applied to virtual maintenance was designed and implemented based on the.Net machine learning class library under Visual Studio platform.(3)Taking the virtual maintenance of aircraft landing gear as an example,a virtual maintenance system is designed based on gesture recognition.The gesture recognition model is implemented in the internal operation and maintenance process of the system.3ds Max is used to build the 3D solid model and import it into the development platform unity3D and complete the coordinate transformation.Using TCP protocol as the client and classification algorithm as the server,the real-time and reliability of gesture recognition are guaranteed by asynchronous communication.C# language is used to realize the functions of scene connection and gesture recognition interactive trigger,and the realization effect of the system is shown.