A Moving Target Recognition Algorithm Based on Improved Mixture Gaussian Background Model
Zhang Yongmei, Li Ma, Liu Mengmeng, Sun Haiyan · 2017
Many mutant factors in the video such as noise and illumination can easily lead to the moving target recognition error. The paper proposes a moving target recognition algorithm based on improved mixture Gaussian background model for ships and vehicles. The algorithm on account of mixture Gaussian background model and three-frame difference can obtain the potential target regions with less background under the condition of motion disturbance and light mutation in the background, extract the straight line, shape factor and Zernike moment features from the potential regions, and construct the least square support vector machine to identify the ships and vehicles. The experiment results show the algorithm can accurately identify the ships and vehicles.