Speed Estimation of Video Target Based on Siamese Convolutional Network and Kalman Filtering

In recent years, computer vision-based moving target state detection and motion parameter estimation have become a hot topic. Aiming at the problem that the accurate speed of moving target cannot be calculated from the target tracking results, in this paper, a speed estimation method using Siamese convolutional network and Kalman filtering is proposed and it is validated by single pendulum experimental data. Firstly, an improved Siamese convolutional network is integrated to detect the position of a moving ball in the single pendulum video. Then, these position coordinates are integrated to estimate the speed of the moving ball by using the information from the previous and current frame through a Kalman filter. The experimental results show that this method can achieve a sound on the speed estimation of moving objects in videos.