IMPLEMENTATION OF A VISUAL-BASED PARKING SLOT OCCUPANCY DETECTION SYSTEM USING ESP32 CAM AND CONVOLUTIONAL NEURAL NETWORK

• Main Author: Dimas Yoga Pratama, Moh
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/53877
• Institution: Institut Teknologi Bandung

This final project describes the design and implementation of a visual-based car parking slot occupancy detection system. Visual-based car parking slot occupancy detection has several advantages over other methods of detecting the occupancy status of a slot, one of which is the wider monitoring range for each sensor so that it can reduce the number of sensors that must be installed and maintained. The system uses a camera built with Ai Thinker's ESP32 CAM to take an image from a parking lot. This camera is designed to work in a temperature range of 18 ° -33 ° C and is rated IP55 for protection against existing environmental conditions. In addition, the camera is also designed to weigh no more than 1 kilogram and a size of less than 16 cm x 16 cm x 16 cm to make installation easier. For parking slot status classification, this system uses the mAlexNet convolutional neural network which classifies the parking slot status to be empty or filled. This system has several cameras installed in each parking lot and has one computer that functions as the central processing unit to process all images from the camera every 10 seconds. From the test results, the system detection accuracy can reach 98% accuracy.