Method for segmentation and recognition of object e.g. cyclist around vehicle, involves forming classifiers for classification of objects based on the frequency statistics and object models

摘要

The method involves segmenting image (B) outside free space (F) into segments (S1-Sn) of the same width. A single set of code (C) with image-based features (M1-Mx) is generated for several object classes (K1-Km), during learning phase of training data so as to detect objects (O1-O4) within the segments. The frequency statistics, and object models (KL1-KLm) are formed with respect to object classes. The classifiers for classification of objects are formed based on the frequency statistics, and models.