Sunlight Compensation for Vision Based Drone Detection

Computer vision based object detection can be applied in security and monitoring scenarios, such as detecting and tracking drone intrusions using cameras. However, its effectiveness is dependent on environmental conditions. For example, under bright sunlight and clear sky conditions, the sunlight reflecting off a target could cause it to blend into the sky and prevent detection. In this paper, an algorithm to compensation for the effects of sunlight on object detection was proposed. The algorithm applied a localised contrast increase to the sky through RGB-HSV conversion and image extraction techniques, which avoided the generation of false positives among the treeline. Preliminary tests with prerecorded videos showed that the algorithm improves detection under bright sunlight conditions but the contrast gain had to be manually tuned. Methods to dynamically tune the gain, and field tests to determine the algorithm's real time effectiveness, are slated for future work.