Image Segmentation: Algorithms and their Applications to Visual Size Measurement

This talk focuses on our recent research in developing image segmentation algorithms for identifying and measuring the sizes of objects in static images or a video sequence. Such a capability is essential in a variety of applications including medical imaging, remote sensing and, in our case, monitoring the industrial process of ore-crushing. Specifically, this talk will overview our novel solutions to detecting steam/smoke in video, segmentation through pixel classification, texture-based background subtraction, and object delineating with image slicing and shape discrimination. The presentation will begin with the basic approaches to image segmentation, outline some of the technical challenges in this general problem, and move to high-level descriptions of our image segmentation algorithms and demonstrations of their potential in meeting the needs of the mining industry in Alberta.