Low-Cost Stereo Vision System Using ESP32-CAM and YOLO

Today, distance estimation algorithm (DEA) is very important to enable machines to map the world and interact with it, independent of human controls. These algorithms are beneficial for various usages, such as the ADAS or Humanoid Robots, to control the movements of each devices. In order to implement DEA, we can integrate the concept of stereo camera with ESP32-CAM connectivity capability to implement distance estimation algorithm in three-dimensional space, by simulating how human eyes work. The objectives of this experiment are to build a Stereo Camera with ESP32-CAM module, design a Distance Estimation Algorithm (DEA), implement DEA to estimate the distance of objects in three-dimensional space, and measure the error and processing delay generated by DEA. The stereo camera implementation stage involves designing the device with ESP32-CAM, uploading the camera web server program, and calculating camera parameters. Distance Estimation Algorithm (DEA) algorithm is executed using Python and YOLO to analyze images from both camera eyes. DEA provides optimal performance for objects with distances below 7 meters after calibration, the calculation of object distances above 7 meters produces significant errors. The DEA implementation also shows limitations in real-time, taking 0.1 - 2 seconds from ESP32-CAM video recording to object calculation results due to delays from the internet network and computer processing limitations.