Object detection

摘要

Objects are detected in real-time at full VGA 30 frame per second resolution. A preprocessor performs run-length encoding (RLE) and generates a summed area table (SAT) of an image. The RLE and SAT are used to identify candidate objects and to iteratively refine their boundaries. A histogram of gradients (HoG) and support vector machine (SVM) then reliably classify the object. The method may be part of an advanced driver assistance system (ADAS).

主权项

1. A method of detecting an object of interest in a first image comprising:
receiving the first image; performing run-length encoding on the first image to identify edges for a shape bounding a candidate object; generating a hypothesis that the candidate object is the object of interest; calculating a histogram of gradients for the candidate object; classifying the candidate object as the object of interest using a support vector machine method; computing a summed area table for a second image, the second image containing the candidate object of the first image, the candidate object having associated with it a shape bounding the candidate object; and dividing the shape into discrete regions and using the summed area table to test for symmetry between regions.