Research On Depth Camera Gesture Recognition In Fire Environment

When carrying out various rescue operations in fires,the assistance of different equipment is often required,such as rescue robots to enter into hard-to-reach places,or to convey messages to other firefighters.In a dangerous and emergency cases such as fire,using a remote control to operate a robot can reduce work efficiency,and using voice to convey messages is often affected by the noise of the fire place.It is undoubtedly an efficient and feasible way to issue command message through gestures,so that fire operators can quickly and conveniently control auxiliary robots or convey messages to other firefighters in a timely and accurate manner.The traditional gesture recognition system based on Kinect depth camera often has poor accuracy due to the interference of flame infrared radiation,the large gap between light and dark of flame greeting and backlight surface,the change of distance between gesture and depth camera,and the protective gloves worn by the operator.Therefore,the research on gesture recognition based on depth camera in fires has certain reference significance in theory and is also of great reference significance in application.There are currently few studies in this area,in view of this series of problems,This paper designs and improves a gesture recognition algorithm that can overcome the influence of multiple factors in the fire environment,and the main research content is as follows:(1)In terms of hand segmentation,the conversion algorithm from traditional RGB color space to YCb Cr color space is improved,and the improved YCb Cr color space skin tone segmentation algorithm is fused with the depth image neighborhood segmentation algorithm.Through the analysis of the research content of the traditional RGB conversion YCb Cr algorithm,a conversion algorithm of RGB color space fusion combining linear and nonlinear conversion is proposed to improve the color space of YCb Cr.The improved YCb Cr color space is used for the skin color hand segmentation algorithm and finally fused with the neighborhood segmentation method based on depth image.Experimental data show that the improved fusion algorithm has a higher accuracy of hand segmentation under the influence of flame.(2)In terms of fingertip detection,the traditional K curvature fingertip detection algorithm is improved,and the convex hull algorithm is integrated.The palm width entering the lens is established with the K value size of the K curvature cosine fingertip detection algorithm.The K value is calculated by using the arc chord distance difference at the fingertip position and eventually the improved K cosine algorithm is fused with the convex hull algorithm.Finally,experimental data show that the improved fusion algorithm has advantages in the accuracy of fingertips when wearing various protective gloves and being far away from the depth camera.(3)Design an experiment to improve the accuracy of gesture recognition of the fusion algorithm in a simulated fire case.In this paper,the template matching method is used to detect gestures.First,define the gestures required in the gesture recognition experiment,second,with gloves and without gloves respectively,keep different distance with the depth camera in different fire cases with and without flame and repeat different gestures Finally,the statistical data is sorted out,which proves that the accuracy of gesture recognition in the fire place.The fused algorithm in this paper has been improved.In this paper,the gesture recognition rate of the proposed algorithm in fire is increased by 12.3% compared with the previous improvement,which has good reliability and can effectively improve the work efficiency of firefighters.