Object Detection & Tracking

Monkeys are important to many areas of science and ecology. The study of monkeys and their welfare are important components requiring complex observational studies. This work is therefore concerned with the development of computer vision techniques for the purposes of detecting and tracking monkeys with the ultimate aim to help in such studies. Monkeys are complex creatures for the purposes of tracking because of complex deformations. This complexity is further compounded by an in the wild setting where forest conditions result in frequent occlusions and changes in lighting. Despite these complexities monkeys present some interesting features that can make detection and tracking possible: their bottoms and faces. A system is thus described consisting of detectors trained to detect faces and bottoms of monkeys which are used within a tracking framework to initialise a system of tracklet construction. Steps are also described to enable disparate but coincident tracklets to be merged thus enabling longer run analysis of individual monkey movements. Experiments are performed using image data taken from video footage of Crested Black Macaques in natural forest surroundings. Results demonstrate relatively successful detection of monkey bottoms where the correspondence analysis and tracking process helps to reduce false positives. 

Automatic bootstrapping and tracking of object contours

John Chiverton, Xianghua Xie, Majid Mirmehdi

Paper in IEEE TIP 2012, https://doi.org/10.1109/TIP.2011.2167343

A new fully automatic object tracking and segmentation framework is proposed. The framework consists of a motion-based bootstrapping algorithm concurrent to a shape-based active contour. The shape-based active contour uses finite shape memory that is automatically and continuously built from both the bootstrap process and the active-contour object tracker. A scheme is proposed to ensure that the finite shape memory is continuously updated but forgets unnecessary information. Two new ways of automatically extracting shape information from image data given a region of interest are also proposed. Results demonstrate that the bootstrapping stage provides important motion and shape information to the object tracker. This information is found to be essential for good (fully automatic) initialization of the active contour. Further results also demonstrate convergence properties of the content of the finite shape memory and similar object tracking performance in comparison with an object tracker with unlimited shape memory. Tests with an active contour using a fixed-shape prior also demonstrate superior performance for the proposed bootstrapped finite-shape-memory framework and similar performance when compared with a recently proposed active contour that uses an alternative online learning model.