Semi-supervised Extreme Learning Machinebased Method for Moving Cast Shadow Detection
Guanglei Kan, Chengduan Wang · 2021
The moving cast shadows often have a negative effect on computer vision and image processing. In this paper, we propose a moving shadow detection method based on semi-supervised extreme learning machine to effectively distinguish shadows from the foreground. Firstly, pixel-level features and region-level features are extracted from the foreground. Secondly, S-ELM (semi-supervised extreme learning machine, S-ELM) is trained by using the features extracted above to classify the object and shadow pixels. Finally, in order to overcome the influence of noises and improve the accuracy of the results, the post-processing procedures are carried out. A substantial number of experiments on two common datasets show that the proposed method has good performance and is superior to some existing well-known methods.