PART BASED PEDESTRIAN DETECTION BY FEATURE SYNTHESIS
MetadataShow full item record
Detecting humans in images or video sequences is a challenging task owing to their variable appearance and the wide range of poses that they can adopt. The first and most important factor in achieving satisfactory detection result is robust feature set that allows the human form to be discriminated clearly, even in cluttered background under difficult illumination. Many researchers endeavored to find a robust and light descriptor for human body for over twenty years. Some of them have presented enlightening approaches and promising performance, such as histogram of oriented gradients , purposed in 2006 and speeded fast robust feature  purposed in 2008 are both outstanding holistic features that give a nice description of pedestrian. Recently a rising trend in part-based approaches pedestrian classification aims to deal with occlusion occasions happen frequently in real world urban street scenarios. By dividing the pedestrian’s body image into several parts relaxes the detector’s emphasis on the whole pedestrian body. Even when part of pedestrian’s body cannot be seen in the image, the rest part of the body will still contribute to the classification.