Design and implementation of a computer Vision-based Object Tracking Robot using an Uncalibrated Camera

Abstract Object detection and tracking are challenging tasks in computer vision applications, such as autonomous robot navigation, vehicle navigation, human-computer interaction, augmented reality and defence surveillance systems. Detecting and tracking an object, especially for humans and vehicles, is a challenging research topic in computer vision. This paper represents an autonomous object-tracking robot using a computer vision object detection and tracking algorithm with a novel robot controller. The speeded-up-robust-feature (SURF) based object detection technique has been employed for object detection, and Kanade-Lucas-Tomasi (KLT) has been used for object tracking. The stereo vision-based depth measurement technique measures objects' depth. Based on the object's location in 3D space, the robot controller navigated the robot to track the object. For this purpose, the robot controller is designed with Arduino and L293D motor driver IC. The robot's speed was controlled using a pulse width modulation (PWM) speed control algorithm. The simulated results show that the proposed techniques efficiently track objects in 3-D space.