Research On Key Technology Of Gesture Recognition And Its Applications In Intelligent Laboratory

Laboratory,as an important place of aerospace research and development units,is the carrier of intelligent production.With the concepts of "Industrial 4.0","Internet +" and intelligent manufacturing-related have been proposed,intelligent laboratories will be the future key component of aerospace intelligent manufacturing infrastructure.In this paper,a set of intelligent laboratory gesture recognition system is developed to improve the level of intelligent laboratory,thus speeding up the laboratories' developing and the production efficiency with the integration of computer vision and machine learning technology as long as the background of the artificial intelligence has become the future development trend domestic and abroad.Gesture recognition can be treated as a way to understand the human body language,the use of gesture recognition technology can expand the interactive way from traditional human-computer interaction through the keyboard or mouse to through the camera remote identification gestures,which is undoubtedly significant to the development of human computer interaction.Therefore,in order to improve the accuracy and efficiency of dynamic gesture recognition of human-computer interaction in actual complex scenery.A hand gesture recognition system based on a novel method named Temporal Locality Sensitive Histograms of Oriented Gradients is proposed.The gesture recognition system has two main components:(1)accurately detecting the gesture area appears in the video frames in real time,by using the deep learning object detection model SSD(Single Shot Multi Box Detector);(2)the dynamic gesture recognition is realized accurately and rapidly by describing the temporal and spatial posture of gesture movement with the novel method of TLSHOG.The main steps of the TLSHOG algorithm are as follows: Firstly,a training set for machine learning is established by recording several videos of human hands using normal PC cameras.After that,a single-frame feature is proposed to describe the spatial feature of hands,and a Temporal Pyramid Method is further applied to describe the trajectories of dynamic hand gestures.At last,a multidimensional Support Vector Machine classifier is build,and all dynamic hand gestures in testing set is then correctly recognized.In this paper,a novel hand gesture recognition system is construct,which achieves the accuracy of 0.903 on the benchmark ASL-Finger Spelling dataset containing 24 categories.This algorithm not only ranks high comparing to other excellent ones,but also reaches a superior processing speed of more than 12 fps.In addition,an extra experiment is implemented on a real scene video dataset to verify the practically of the results,which achieves an accuracy of 0.893 and a processing speed of more than 14 fps,meeting the practical application requirements.To sum up,the following conclusions can be drawn: this new method is highly discriminative of dynamic hand gestures and robust to effects of complex backgrounds and illuminations.