Human tracking in crowded public places.
- The paper [1] presents a novel method of Differential Images based Target Tracking.
- Complete system of Human Detection, Target Person Selection, Target Tracking and Robot’s Steering towards the target person is presented.
- Comparison of presented tracking method is done with conventional tracking methods.
Limitations of existing methods
Limitations of existing methods
Limitations of Cam Shift based tracking
Limitations of Cam Shift based tracking
- Good for tracking of small objects.
- False tracking in case target's histogram distribution is similar to any other object's distribution.
- Average response time = 1.4 ms
Limitations of LK Optical Flow based tracking
Limitations of LK Optical Flow based tracking
- False tracking under background movement.
- Cannot detect large motion.
- Tracking is based on the movement of featured points only.
- Average response time = 0.8 ms
Limitations of Particle filter based tracking
Limitations of Particle filter based tracking
- Based on Gaussian sampling from images.
- False tracking in case sample distribution matches the distribution of any other region.
- Average response time = 24 ms
[1] Badar Ali, Ahmed Hussain Qureshi, Khawaja Fahad Iqbal, Yasar Ayaz, Syed Omer Gilani, Mohsin Jamil, Naveed Muhammad, Faizan Ahmed, Mannan Saeed Muhammad, Whoi-Yul Kim, Moonsoo Ra, "Human tracking by a mobile robot using 3D features", Proceedings of IEEE-RAS International Conference on Robotics and Biomimetics (ROBIO), pp. 2464-2469, Shenzhen, China, 2013.