Human detection in a challenging situation

Although there have been a lot of studies on human detection in recent years, most of them have some particular requirements. However, some videos may have complicated scenes with low resolution and include both moving and standing people. The paper aims for individual detection under the challenging situation. An EM (Expectation-Maximum) based method has been developed to detect individuals in a common outdoor scenario. A new cluster model is employed in the method for human detection. The method can work well even for low-resolution images. Both moving and standing people can be detected as long as they show some occasional movements, which is satisfied in most cases. Some promising results have been presented and analyzed in a scene with around 100 people. ©2009 IEEE.link_to_subscribed_fulltex