Joint Detection and Tracking with Movable Camera and Its Application to a Drilling Robot in Underground Coal Mine

摘要

The movable camera usually has two states, namely static and moving, which makes it difficult to detect or track the object. The following three problems may rise. First, when the camera and the object are relatively stationary, the stationary object may possibly be misjudged as the background or even be gradually absorbed by the background, due to the complexity and limitation of the background model construction. Second, when the camera and the object are relatively moving, the object may be blurred due to camera vibration since the relative movement between them is irregular, which affects the tracking effect. Third, the background is constantly changing when the camera is moving, which may cause confusion between the object and the background. In order to solve the above problems, this paper proposes a joint detection and tracking method with moving camera. The relative status between the camera and the object is judged in real time through the Lucas-Kanade (LK) algorithm at first. For relative stationary state, object detection can be conducted with foreground segmentation method. For relative moving state, object tracking can be implemented with color based adaptive tracking method. Then, the objective can be simultaneously detected and tracked with moving camera, by distinguishing relative stationary and moving states. Finally, to verify the effectiveness of the proposed method, it is applied to borehole detection and tracking for a drilling robot in underground coal mine.