A Color-Based Real-Time Target Detection Using OpenCV for DJI Tello Drone

The utilization of drones has spread into a variety of fields due to the continuous development progress of their technologies. Their applications across military aviation fields into civilian life, where they are performing various tasks such as disaster management, search and rescue operations, and surveillance. Their applications in those operations are possible with the ability to detect the target of the operations. Hence, object detection gained attention in the research community of drones. One of the methods of object detection is color-based detection. The application of the method is not only limited to drones' technology. However, most of the applied color-based methods are designed with a focus on recognizing and labeling the name of the detected colors, but not to detect and recognize the custom target for tracking purposes in the operations. Thus, a color-based detection method without utilizing any advanced machine learning algorithm is proposed to detect the desired target of a drone's operation. A real-time color-based target detection is developed using the OpenCV library. The proposed method is simulated with a webcam and tested with a DJI Tello camera. The results show the target is successfully detected when tested in the same environment as trained. However, the method is sensitive to light intensity and unable to differentiate the target and other objects with the same color. Hence, several improvements to be made are suggested. In the future, for the applications in the drone's target-following operations, the improved algorithm of the proposed method will be put to comparison with a detection using YOLO.