SYSTEM OF IMAGE-BASED OBJECT CAPTURING AND FOLLOWING IN TWO DIMENSIONS

Abstract

This paper presents a system and software package created to examine control signals supplied to the motors of an unmanned aerial vehicle (UAV), as well as to implement automatic drone turn during photo-tracking of a moving target in order to increase the efficiency of control processes. A Q450 quadcopter equipped with D2212-920kv engines and 10×1.5 cm propellers was used as the hardware base. The system included the Pixhawk PX-4 autopilot with an open architecture and a Raspberry Pi 3 on-board computer with a Pi V2 camera (resolution 1080×720 pixels). To record the propeller rotation parameters, the device was suspended on a thin wire, orienting the propellers in a horizontal plane. 
Given setup also includes a digital oscilloscope ISDS205A to measure the duration of control pulses, and a laser tachometer UT-372 to determine the engine rotation speed. A software module has been developed that determines the coordinate of the center of the object image (in pixels along the X axis) and calculates the corresponding pulse durations transmitted to the fourth control channel of the autopilot. In the process of implementing the system, the DroneKit, datetime, errno, platform libraries were used to implement the flight logic, as well as OpenCV, Picamera and sys — for processing the video stream and computer vision. 
Analysis of the results of the experimental study of the functioning of the system and software allows us to conclude about the feasibility and effectiveness of the proposed approach for controlling an unmanned aerial vehicle in conditions of continuous receipt of a stream of frames from the camera during pursuit. Theoretical developments and practical recommendations made on the basis of the work performed have both independent theoretical value for prospective research in the field of photocapture and photopursuit of small and distant moving objects, and have high practical value in connection with the need for real detection and measurement of the parameters of the movement of unmanned aerial vehicles.