摘要 |
An object is detected at the VGA resolution of 30 frames per second in real time. A preprocessor performs a run length encoding (RLE) and generates a summed area table (SAT) of images. The RLE and the SAT identify candidate objects and are used to repetitively refine their boundary. A histogram of a gradient (HoG) and a support vector machine (SVM) trustworthily classify objects afterward. The method can be a part of an advanced driver assistance system (ADAS). |