Detecting and/or Removing Shadow from Video Stream

By using the Physical Linguistic Vision Technologies, researchers in Yang’s Scientific Research Institute, LLC., USA., developed an automatic shadow detecting system for many on-going objects tracking and recognizing projects including: vehicle tracking, intelligent traffic monitoring, pedestrian monitoring, visual security systems, vehicle classification and many more. Many existing shadow detecting methods are subjects to critical drawbacks when meet challenges imposed by real-life working conditions. For an intelligent traffic control system, it will be impossible to allow interferences from staff personals for many monitoring video cameras. These cameras must learn the environment all by themselves. Many existing shadow detecting algorithms based heavily on prescribed parameters and ad hoc hypotheses which are very far from the truth of the real-life working conditions of a real monitoring camera, especially a working traffic camera. The following challenges were systematically considered by the researchers and field engineers from the strategic partners of Yang’s when Physical Linguistic Vision Technologies were used to design shadow detectors:
1. Dramatic changes of the lightening conditions during a day-night cycle in a typical summer day with a few clouds;
2. The headlight glaring at night;
3. The transition during sunsets;
4. Properties and controls of the video cameras;
5. Rain, fog and snow;
6. No traffic camera is used for the sole purpose of removing shadows;
7. Seasons.
The backbone behind our algorithm is the Physical Linguistic Vision Technologies for dynamic scene understanding. (October 16, 2005, Tucson, Arizona).

Shadow removal SDK is available. [Demo video].
Now a part of TrafGo Intelligent Transportation Systems Developer’s C\C++ kit.