Research On Gesture Recognition Based On Thermal Imager

Nowadays,with the rapid development of intelligence,human-computer interaction has been widely used.As one of the most natural and intuitive ways of human communication,gesture plays an important role in human-computer interaction.At present,the realization methods of gesture recognition mainly include gesture recognition based on vision sensor and gesture recognition based on wearable sensor.In the research of gesture recognition based on vision sensor,the main method is using monocular or multi-camera to complete gesture recognition,but this method has the problem of being greatly affected by environmental light.Because thermal imager can solve the problem that ordinary camera is affected by the change of ambient light,a static gesture recognition system based on thermal imager is proposed.The main contents of the system are as follows:In order to solve the problem that the gesture part is not prominent in complex environment,the method of gray scale expansion of infrared image is adopted.By converting infrared data into image pixels through logarithmic conversion and stretching the linear gray scale,infrared images with high contrast and prominent gestures can be obtained in various complex environments,which is conducive to image edge detection.Aiming at the problem that the traditional Canny edge detection algorithm needs to set threshold manually,which makes it unable to adapt to different environments,an adaptive threshold Canny edge detection algorithm is designed.The algorithm adaptively adjusts the input threshold of edge detection by using the proportion of the number of edge pixels to the number of all pixels in the whole image as feedback information,keeps the threshold reasonable and obtains the edge of the whole image,and achieves the correct image segmentation according to the changing environment.In order to meet the real-time requirement of the system,the gesture recognition system adopts the idea of hierarchical structure,and classifies by setting feature threshold.The hierarchical structure is designed by using the geometric structure features of gestures,and the hierarchical arrangement is completed in the order of small to large according to the calculation amount of extracted features.The recognition rate of ten gestures in simple background and complex background is98.8% and 85.7% respectively.Compared with the method without gray scale expansion,the recognition rate of this algorithm is improved by 2.61% and 21.53% respectively.The recognition rate is 96%,99.8% and 97.6% when the corresponding ambient temperature is 5,20 and 35,respectively.The system is not affected by temperature change and has good stability.At the same time,the system can still realize normal recognition without visible light.In addition,the recognition rate of wearing gloves is 97.4%,which indicates that the system can realize the recognition of wearing gloves.