Intelligent Image Processing Technology for Badminton Robot under Machine Vision of Internet of Things

The present work aims to promote the development of intelligent image processing technology for badminton robots and optimize the application effect of badminton robots in national fitness. Firstly, the problems and common needs of the badminton robot currently in use are investigated. Secondly, a shuttlecock aerodynamic model is established to simulate the effects of air resistance and gravity on the aerial flight of shuttlecock. Besides, the convolution neural network (CNN) is combined with traditional image processing technology to denoise and recognize the collected shuttlecock images. Finally, the badminton robot vision system is constructed and its performance is tested. The results demonstrate that the image denoising method based on CNN and the traditional image processing method can effectively process and denoise the captured moving image. Under the noise level of sigma = 25, the peak signal-to-noise ratio index of this method is better than the Gaussian Scale Model, k-Singular Value Decomposition, and Color Names methods, slightly better than that of the Multilayer Perceptron (MLP) network. In terms of the time consumed in processing the same number of pictures, the method reported here takes the least time. But when sigma > 50, the MLP method has a better denoising effect because the noise is overlarge and the features are not easy to learn. Moreover, the detection accuracy of the optimized Single Shot MultiBox Detector (SSD) method adopted here is 79.0%. This accuracy is 1.7% higher than that of the traditional SSD method, and 2.3% higher than that of Faster Region-Convolutional Neural Network based on Region Proposal Network. The optimized network structure reported here provides a certain idea for the software design of the badminton robot.