Research And Implementation Of Auxiliary Sliding Distance Prediction System For Freestyle Skiing

In 2022,the 24th Winter Olympic Games will be held in China for the first time,which highlights the improvement of China’s national strength and will better promote the improvement of the level of ice and snow sports in China.In today’s sports events,science and technology have been closely linked with sports,and the research of science and technology has promoted the rapid development of sports competition and constantly breakthrough the limit.It also encourages researchers to better serve the Olympic Movement.The freestyle skiing aerials in the Beijing Winter Olympic Games are highly appreciated and loved by people because of their "high appreciation value" and "strong stimulation".At the same time,due to the main features of this project focuses on the techniques and flexibility,very adapt to the characteristics of sports of Chinese athletes and body state,China’s athletes are expected to get the project results in the 2022 winter Olympics breakthrough,for freestyle skiing aerial skills project,has accomplished the key lies in the choice of athletes and coaches to help sliding distance.But at present,this process is usually judged by experience.In complex circumstances,it is a very difficult task for athletes and coaches,so it is urgent to describe the distance of assisted sliding by quantitative method.So in this paper,based on ARM embedded technology and artificial intelligence algorithm,a set of real-time data collection,fusion of a variety of sensor information recognition,intelligent calculation of freestyle skiing distance prediction system is designed.The system automatically collects the natural environment parameters through a variety of sensors,and provides the prediction data reference for the sliding distance for the coaches and athletes,which reduces the pressure of the athletes and coaches who estimate the real-time environmental resistance completely from their own experience and lack of data reference.In this paper,through the acquisition and processing of the sensor data,machine learning and deep learning algorithms are used to achieve the prediction of the auxiliary sliding distance of free skiing aerials.The main tasks in this paper include the following:(1)Overall hardware system construction.By comparing with the current mainstream technologies,this paper uses modular embedded devices to achieve the portability and lightweight of the devices to the greatest extent.Among them,RaspberryPi serves as the control host and WIFI serves as the auxiliary communication mode.Through computer interface technology,external temperature and humidity,wind speed and direction sensor and snow surface friction coefficient acquisition device to realize the collection of natural environment parameters of the ski track.(2)Multi-sensor data acquisition system.In this paper,four kinds of acquisition devices are used to collect the natural environment data of snow field in real time.The multi-sensor data acquisition system is mainly composed of snow surface friction coefficient acquisition system control module,wind speed and direction control module,temperature and humidity control module and data synchronization module.At the same time,aiming at the acquisition task of multi-sensor,a parallel data acquisition system is designed to realize the parallel data acquisition and synchronization process of multi-sensor.(3)Data preprocessing of the original signal.In this paper,the collected sensor data are analyzed in frequency domain.The extraction process of some key features of noise interference exists in the signal.Aiming at the noise,wavelet transform is used to filter and reduce the noise.At the same time,in order to realize the segmentation of the data collected by the snow surface friction coefficient sensor and the cutting of the data of the contact deceleration movement segment,two common methods were compared and analyzed.Finally,the slope intercept method was adopted to realize the detection of the starting and ending points of different activity segments.(4)Feature extraction.In this paper,multiple groups of different environmental parameters were extracted for feature extraction and feature analysis.In order to distinguish different environmental states more accurately and make the regression algorithm perform better,this paper extended the extracted basic features and used polynomial regression method to construct the high-order features.The features with high weight were selected as typical features to participate in the construction of the regression model.(5)Regression algorithm comparison and experimental verification.In this paper,a variety of machine learning and deep learning regression algorithms were used for comparative analysis,and an optimal regression model was established to predict the sliding distance assisted by landslide.By adjusting the super parameters of the regression model,the optimal model among different regression methods was selected.Finally test through the experiment,the choice of performance in the experiment,the optimal regression model based on the depth of the neural network of AdaBoost integration regression model as the final model of this system,the model for the prediction RMS error of this system and the mean absolute percentage error values,and able to capture the data on the largest amount of information,The usability of the system is verified.