Pedestrian Detection in Low Resolution Videos
Consider a video taken by a flying vehicle such as a drone. People in these images are very small (maybe 20 pixels in height). That makes automatic detection of people challenging. However, it is easy for us to look at a video like this and identify people. How do we do it? The key thing we use is MOTION … we can recognize the characteristic appearance of a person walking.
My graduate student (Hisham Sager) and I developed an algorithm that detects pedestrians (i.e., walking people) from low resolution videos. The method extracts gradient features from a spatiotemporal volume, consisting of a short sequence of images (about one second in duration). The additional information provided by the motion of the person compensates for the loss of resolution. On standard datasets we show a significant improvement in performance over single-frame detection.
For complete details, see our journal paper: H. Sager and W. Hoff. "Pedestrian Detection in Low Resolution Videos Using A Multi-Frame Hog-Based Detector." International Research Journal of Computer Science (IRJCS), Issue 03, Volume 6, pp. 55-71, 2019. (pdf)