Research On Gesture Recognition For Vehicle Air Conditioning Control

In this era of rapid development of intelligence,there are more and more auxiliary operations in the car,but the related operations are also becoming more and more complicated.In the case of changing the auxiliary operations in the car,the driver’s sight is inevitably far from the driving route,and traffic accidents are likely to occur.In order to simplify the auxiliary operation of the relevant system in the car and avoid the driver’s sight away from the driving route,this paper investigates the use of gesture recognition to control the relevant functions of the car air conditioner in the context of the car air conditioner auxiliary control.The main work of the paper is as follows:To address the problem that gesture recognition during driving can be affected by too much light and lead to a decrease in recognition detection rate or gestures cannot be recognized,the residual module is added to the Retinex-Net low-light image enhancement algorithm and the F-Conv Net network structure is designed to improve the learning ability of the network for information such as chromaticity,saturation,contrast and edge details in images,effectively suppressing the noise.The network can effectively suppress the generation of noise and improve the robustness of the algorithm to low-light scenes.To address the problems of missed false detection and low confidence in hand region detection,the CA attention mechanism is added to the YOLOv7 network model.This enables the model to focus more attention on the hand region,reducing the interference of background and some redundant information in the feature map,effectively improving the recognition accuracy of the YOLOv7 network model,and reducing the cases of missed and false detection of hand region detection,and increasing the confidence level.For the problem of large detection error of existing hand key point detection algorithm,a hand UNet network structure with fewer overall dimensions and number of parameters is designed,which makes the network model more lightweight and runs faster with less computation,while multi-layer feature extraction and multi-layer feature fusion are done on the hand UNet network model,which makes the network model more accurate and better detection effect.Based on the background of automotive air conditioning auxiliary control,the hand gesture command dataset for controlling car air conditioning is developed and the collected dataset is processed,and the fusion of BiGRU network and Capsule network is proposed for hand gesture command recognition,which makes it possible to extract both temporal features and spatial vector information of key points of hand gestures when extracting hand gesture features,thus making the extracted feature information more.This makes the extracted feature information more comprehensive and effectively improves the accuracy of gesture recognition.The experimental platform for gesture recognition in in-vehicle scenarios is built,and the upper computer and the lower computer are designed so that the upper computer can collect and recognize gesture commands in real time,and then send the recognition results to the lower computer according to the communication protocol to control the functions related to the car air conditioner simulated by the lower computer,realizing the real-time collection,recognition,transmission and control of gestures.