engleski

Real time detection of humans and small objects from UAV images using CNN

Detection of small objects and humans using aerial images taken with the unmanned aerial vehicles is a challenging task. Different kinds of environments, small numbers of pixels that represent the target, variable shapes and colors as well as lack of adequate image databases makes this research important and demanding. Applications of possible solutions can used for border surveillance, search and rescue missions, firefighter support, environment protection etc. During last years, our research group investigated several procedures and algorithms mainly for detecting the lost people for search and rescue missions. Some of the important topics are: image segmentation and filtering, feature detection, convolutional neural networks, image databases. In this presentation we will show some of the most successful methods and obtained results (precision, accuracy, speed, ...) that were developed and applied. Also, ongoing research of the intelligent system for search and rescue will be presented and discussed as well as the ideas for the future improvement of current algorithms and detection results.

object detection, UAV, CNN

nije evidentirano