Research on human-vehicle gesture interaction technology based on computer visionbility

With the development of human-computer interaction technology, gesture recognition is becoming more and more important. At the same time, due to the rapid development of automotive intelligence, the introduction of human-computer interaction technology into intelligent vehicles has increasingly become an important work. Aiming at the problems of low accuracy, low recognition efficiency and weak anti-interference ability of previous gesture recognition applications in driving scenes. This paper presents an improved yolov5 algorithm. By adding the improved and optimized k-means++ clustering optimization algorithm, the problems of unstable clustering effect of K-means clustering algorithm in yolov5 model and slow convergence to large-scale data are solved. In addition, by combining the C3 module in the backbone network with the attention mechanism (CBAM), the effect of target gesture recognition under complex background is improved. Finally, the latest optimization method of loss function (EIOU) is added to the algorithm model to improve the accuracy of training convergence. The average recognition accuracy of the algorithm proposed in this paper is 4.8% higher than 88.19% of the original yolov5s algorithm when the intersection to union ratio threshold is 0.5 to 0.95. The practical availability of the improved gesture recognition algorithm is verified by the simulation scene based on ROS (robot operating system) and unity.