Computer Vision System Toolbox
Object detection is the identification of an object in an image or video. Computer Vision System Toolbox supports several approaches to object detection, including the Viola-Jones algorithm, feature-based matching, blob analysis, foreground detection, and template matching.
The Viola-Jones object detector uses Haar-like features and a cascade of classifiers to detect pretrained objects. The system toolbox includes pretrained classifiers for the detection of faces, noses, eyes, and other body parts.
The system toolbox provides a pretrained SVM classifier that uses histograms of oriented gradients (HOG) features as inputs to detect upright humans. The pretrained human detector can be used in combination with an object tracker and other higher-level decision modules to form a video analytics system for activity analysis and video surveillance applications.
Blob analysis is used to identify objects of interest by computing the blob properties from the output of a segmentation algorithm such as background subtraction. The system toolbox provides blob analysis functions and a fast foreground detector that models background in a video stream using Gaussian Mixture Models (GMM). Combining blob analysis and foreground detection can be used to effectively detect moving objects.
Feature points are used for object detection by detecting a set of features in a reference image, extracting feature descriptors, and matching features between the reference image and an input. This method of object detection can detect reference objects despite scale and orientation changes and is robust to partial occlusions.