Smart Window Transform とエッジベース識別器に基づく人検出
- 精密工学会秋季学会学術講演会, 2011
Download: PDF (Japanese)
Almost human detection systems are based on classification of rectangular sub-windows, which are enumerated from 2D image. Therefore, the detection rate of such systems is affected by image distortion and titled orientation of the human in images. To overcome this problem, a new method called Smart Window Transform is proposed. This method creates a Smart Window for an observed object in 3D space, which is paralleled to the observed camera image plan, then, transforms Smart Window into 2D space, using camera intrinsic and extrinsic parameters and finally projects to a rectangular sub- window, specifed by a Classifer. As experiments results, human detection rates using Smart Window Transform and Edge-based Classifer were successfully improved compared with the one without Smart Window Transform.