Verification of Feature Detection Through Thermal Imaging: An Extension of PiBase

Abstract

PiBase is a low-cost, Internet-of-Things-capable security systems. It offers users an integrated system of hardware and software in the form of a basic smart camera made from a Raspberry Pi and off-the-shelf parts, an Android app, and a backend built for Google Firebase. Using Haar-feature cascade classifiers and Linear Binary Pattern Histograms, it attempts to provide comprehensive detection and recognition of potential security threats. Being only a prototype there are some vulnerabilities in the initial design. This new design addresses one security threat, namely the possibility of mimicking an authorized user’s appearance. This is achieved through the integration of thermal imaging alongside the original camera used in the system. Some challenges with this approach include maintaining low-cost and part accessibility, working within limitations of the hardware, and choosing an effective method of integration. The proposed solution addresses each of these, in addition to the original issue, by transforming the output of a low-resolution thermal sensor array into a kind of clipping-mask to filter out non-human objects from the input image before performing its other operations.