HighlightNet: Highlighting Low-Light Potential Features for Real-Time UAV Tracking

In low-light conditions, crucial features hide in the shadow, even state-of-the-art trackers can hardly extract them. This situation, along with visual interference caused by illumination variation, partial occlusion, inevitable noise, and other reasons has made it more likely to damage the potential features of images, which might significantly degrade the credibility of tracking result. Apart from degradation on tracking performance, online object selection by human monitors in ground control station (GCS) also encounters difficulty due to low visibility. Therefore, this work proposed a novel enhancer, HighlightNet, to light up potential objects for both human operators and UAV trackers. By employing Transformer, HighlightNet can adjust enhancement parameters according to global features and is thus adaptive for illumination variation. Pixel-level range mask is introduced to make HighlightNet more focused on the enhancement of the tracking object and regions without light sources. Furthermore, a soft truncation mechanism is built to prevent background noise from being mistaken for crucial features. Experiments on image enhancement benchmark demonstrate HighlightNet has advantages in facilitating human perception.