Low-Level Vision Algorithms for Localization, Classification, and Tracking

Camera networks can provide images of detected objects that vary in perspective and level of obstruction. To improve the understanding of visual events, vision algorithms are implemented in a wireless sensor network. Methods were developed to fuse data from multiple cameras to improve object identification and location in the presence of obstructions. Training sets of images allow classification of objects into familiar categories. Feature-based object correspondence is used to track multiple objects throughout a sequence of images