Development of a Stereo Vision System Using ESP32-CAM for Object Distance Measurement and Small Robot Arm Control

Abstract

This research presents the development of a stereo vision system using ESP32-CAM for object distance measurement and small robot arm control to automatically grasp or manipulate objects. The system consists of a dual camera mounted on a baseline-adjustable structure. It uses disparity mapping from the left and right images to estimate the distance in real-time using AI via an OpenCV library. Furthermore, the system incorporates object detection using a deep learning model with Mask R-CNN-LSTM techniques to identify the position and boundaries of the target object before calculating the distance. Linear regression is used to minimize measurement errors. Experimental results show that the AI model has an object detection accuracy of over 80% for small objects and over 90% for medium-sized objects. The system can measure the distance of objects in the range of 20–70 centimeters with a relative error of less than 5% for small objects and less than 3% for medium-sized objects. The distance and position data obtained from the stereo cameras are coordinate-transformed into the coordinate system of a custom-developed small robotic arm using a 3D printer. The Arduino Uno-controlled stepper motor system incorporates calibration and a rotation matrix for precise object gripping. This proposed system is useful for applications in small industrial robots, service robots, and automation in assembly or parts sorting requiring high precision.