Based On Quaternion Attention Selection And Pulse Coupled Neural Network Detection In Football

Soccer is one of the most popular sports all over the world. Automated soccer detection plays an important role in automatic video tracking, video enhancement and so on. It is also very helpful to soccer referee in judgment. There were many target tracking methods used for object detection. However, players and soccer move fast and they will overlap each other. These mentioned methods can’t achieve ideal identification rate in the specific scene.We do a lot research to solve the existing problems. This paper proposes a new soccer detection method combining PQFT (Phase Spectrum of Quaternion Fourier Transform) based attention selection and PCNN (Pulse Coupled Neural Network). First, using image preprocessing techniques removes the regions outside the soccer field. Then, using PQFT extracts the Region of Interest (ROI). Next, using PCNN model and shape information detects the soccer. If there is only one soccer candidate, it is caught as the target. Otherwise using Kalman filter predicts the location of the soccer. The main contributions are as follows:(1) Attention selection method is introduced into algorithm of soccer detection. Firstly we take preprocessing to remove the regions those are not the field. Then, we use PQFT to extract the regions of interest. The target will be in one of regions of interest. It will reduce the searching range and raise efficiency.(2) A simplified PCNN color matching model is proposed in this thesis. Due to the propagation characteristics of the PCNN, neurons in similar regions of the image will remain fire in the same cycle and some isolated matching points can be removed.(3) In the thesis, we introduce a soccer detection algorithm based on PCNN hole-filter. Using PCNN hole-filter processes quaternion saliency map which based on color differences, brightness and prior knowledge. It can detect the soccer directly in most case.