PROJECT TITLE :

A Vision-Based Precipitation Sensor for Detection and Classification of Hydrometeors

ABSTRACT:

Measuring precipitation is a vital half of ground observations of the world’s atmosphere. Existing systems for this task focus mainly on the hydrometeors’ micro-structure (e.g., form, size, and velocity), however seldom consider to classify them. This paper proposes a replacement vision-based mostly system for precipitation observation and type recognition (POTR) comprising one camera and different commercially accessible components. The system is efficient in terms of energy use and memory necessities by having the ability to switch between periodic and continuous monitoring as required, based mostly on a quick detection algorithm for precipitation particles (FDAP). FDAP uses a background model and thresholding strategy to section precipitation particles. In explicit, it applies an space rule and a part rule to eliminate the influence of noise and external interference (e.g., flying insects) on the sphere observations. We describe precipitation particles employing a composite representation that includes geometric options, Fourier descriptors, and Hu moment invariants, and conjointly adopt gradient boosting trees to classify the kind of precipitation. The experimental analysis of POTR on knowledge sets collected on-site in Beijing from August 2014 to February 2015 shows that FDAP has an accuracy of a lot of than ninety six% which sort recognition has an accuracy more than ninety%.