Application of Histogram of Oriented Gradients and Local Binary Pattern in Automatic Waste Management Bins

Existing waste segregation and management systems that use image processing predominantly work on stationery waste, focusing either on using conveyor belts or on industrial-scale processes. In this work, a household bin that opens, closes and handles waste automatically with minimum human intervention is designed. The goal here is to use image processing with machine learning (ML) algorithms to accurately detect the kind of waste being brought to the bin and trigger a subsequent appropriate set of protocols. This paper investigates the applicability of the following ML algorithms in image processing: Hu Moments, Histogram of Oriented Gradients (HOG) and Local Binary Pattern (LBP). All three are compared and reasons for selecting one, or more than one in combination, over the other are discussed.