Automated Lamp-Type Identification for City-Wide Outdoor Lighting Infrastructures

Abstract

As cities ramp up the efforts to convert their aging lighting infrastructure to connected and energy-efficient Light-Emitting Diodes (LEDs), they are confounded by the lack of reliable information about their existing outdoor lighting bases. In this paper, we propose a vehicle-mounted spectrom etry-based approach to scalably audit the roadway lamp types by driving across the city, thereby quickly and efficiently providing the basis for planning and executing LED conversion projects. LambdaSeek, a mobile sensing system that can be mounted on a vehicle, is developed to reliably capture the Spectral Power Distributions (SPDs) of the light emitted by the luminaires on the light poles by driving around the city. The on-board illuminance sensor and the global positioning system receiver helps to localize the SPDs, which are then classified into the corresponding lamp types using a k-Nearest Neighbor classification algorithm. Validation experiments across four field trials are presented: the most commonly found High-Pressure Sodium, Mercury Vapor, Metal Halide and LED lamps were classified correctly with a recall rate of more than 95%.