Circularity Parameter Utilized Vision-Based Navigation for Quadcopter’s Robotic Arm

This paper presents a new Vision-Based Navigation (VBN) scheme, referred to as the Detect Classify Guide and Grab (DCGG) scheme, for a depth camera-mounted two-degree-of-freedom robotic arm integrated with a logistics quadcopter. The DCGG scheme comprises a depth camera, Scale-Invariant Feature Transform (SIFT) algorithm, Sum of Squared Differences (SSD) algorithm, Random Sample Consensus (RANSAC) algorithm, Shape Detection using Circularity Parameter (SDCP) algorithm and Closed-Loop Servo Motor (CLSM) control systems. When the quadcopter reaches the payload, the SIFT detects the desired features of the image captured by the camera. Then, the SSD identifies the matched features by comparing the features from SIFT with the features of the reference image associated with the payload. The RANSAC algorithm concurrently filters unmatched features. The payload orientation (θob) is determined using matched features and similarity transformation. At the same time, the SDCP algorithm identifies the shape of the payload in the image along with its optimal grabbing points and grabbing angle (θg) for the robotic arm’s claw. The grabbing angle and payload orientation are summed to find the desired rotational angle for the robotic arm’s claw. The robotic arm’s claw achieves the desired rotation (θd) using a CLSM system comprising a PD controller. At the same time, the relative distance between the robotic arm and the payload is obtained through the depth camera mounted on the robotic arm. The robotic arm is moved towards the payload, reducing the relative distance to zero, using another CLSM control system. The effectiveness of the algorithm is validated through numerical simulations.