INTELLIGENT IOT STREET LIGHTING USING LSTM FORECASTING AND DEEP VISION ALGORITHMS FOR ENERGY EFFICIENCY AND PREDICTIVE MAINTENANCE

Abstract

Public street illumination plays a crucial role in ensuring safety in cities and facilitating smooth transportation. However, conventional street lighting infrastructures usually depend on preset timing schedules or simple ambient light sensors, leading to significant energy waste during periods of low traffic [1]. To address this issue, this paper presents an intelligent IoT-based street lighting solution that combines sensor-based environmental monitoring with deep learning–driven object detection. The proposed system links low-power components, including Passive Infrared (PIR) motion sensors and Light Dependent Resistors (LDR), with a YOLOv8 computer vision model to detect pedestrians, vehicles, and animals in real time [3], [4]. Video feeds captured through a webcam are processed by the YOLOv8 model in a Python environment, and the detected objects are transmitted wirelessly over Wi-Fi to an ESP32 microcontroller [5]. Based on both the measured ambient light intensity and the detected entities, the ESP32 dynamically controls streetlight operation via a relay module. Experimental evaluation indicates that this approach significantly reduces unnecessary energy consumption while maintaining safety through context-aware lighting. The system thus provides a scalable and cost-effective approach for deploying smart lighting within urban infrastructure [1].