Camera view from crane payload: video stabilization

Tower crane operators often have limited visibility during lifting operations in the construction of high-rise buildings. Providing the operator with visual feedback from the perspective of the payload has potential to reduce collisions. However, a camera feed from this perspective is susceptible to disorienting swinging motions. This paper presents an algorithm to digitally stabilize this low-frequency, high-amplitude swinging motion in 3D. A stabilized virtual camera feed is created to give the operator an intuitive perspective. The Euclidean transform between the actual and virtual cameras is related to an image transform. This is then used to warp the image to achieve the desired stabilization effect. Experimental validation with a robotic arm to simulate the crane dynamics demonstrates the effectiveness of stabilization.