|
|
Milestones
|
|
|
|
- Image Segmentation and Object Extraction Module
We shall implement the Image Segmentation and Object Extraction Module based on our segmentation framework proposed in [5] (see 3.2.1). The framework proposed in [5] has shown that seemingly most of the existing segmentation methods belong to the special cases of our framework. That means segmentation, which is the most important part of the semantic video analysis process, can be systematically parameterized under our framework for different analysis tasks. We shall seek for optimal parameterization that is well suited for surveillance types of videos.
- Object Tracking and Recognition Module
With objects being segmented accurately in the first place, we can then implement various object tracking [2,6-9,30-34] and recognition [28-30, 25-38] algorithms for the further analysis. The collection of object tracking and recognition algorithms will be packaged as the Object Tracking and Recognition Module.
- Semantic Video Processing Module
With the object recognition and tracking methods in place, we can then implement the algorithms that process the object information (e.g. types of objects, trajectories, behavior) derived to determine the interactions among the objects. This corresponds to the Semantic Video Processing Module.
- Video Retrieval System with Semantic Video Analysis
Finally, all the tools developed will be integrated altogether to realize a Video Retrieval Prototyping System
|
|
|
|
|
|
|
