Digital image processing
Integrated Person Tracking Using Stereo, Color, and Pattern Detection
CVPR '98 Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
Human Detection using Geometrical Pixel Value Structures
FGR '02 Proceedings of the Fifth IEEE International Conference on Automatic Face and Gesture Recognition
Detecting Pedestrians Using Patterns of Motion and Appearance
ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Detecting Human Motion with Support Vector Machines
ICPR '04 Proceedings of the Pattern Recognition, 17th International Conference on (ICPR'04) Volume 2 - Volume 02
Histograms of Oriented Gradients for Human Detection
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Indoor Mobile Robotics at Grima, PUC
Journal of Intelligent and Robotic Systems
Hi-index | 0.00 |
In order to deploy mobile robots in social environments like indoor buildings, they need to be provided with perceptual abilities to detect people. In the computer vision literature the most typical solution to this problem is based on background subtraction techniques, however, in the case of a mobile robot this is not a viable solution. This paper shows an approach to robustly detect people in indoor environments using a mobile platform. The approach uses a stereo vision system that yields a stereo pair from which a disparity image is obtained. From this disparity image, interesting objects or blobs are segmented using a region growing algorithm. Afterwards, a color segmentation algorithm is performed on each blob, searching for human skin color areas. Finally, a probabilistic classifier provides information to decide if a given skin region corresponds to a human. We test the approach by mounting the resulting system on a mobile robot that navigates in an office type indoor building. We test the system under real time operation and different illumination conditions. The results indicate human detection accuracies over 90% in our test.