Implementation and evaluation of an open source stereo camera

Abstract

The aim of this thesis is to make the field of applied machine vision easier to approach by proposing a fully open source FPGA based hardware system of a stereo camera. The importance of this task lays to the fact that the area of machine vision sensors is very proprietary which in turn discourages newcomers and hinders the progress in related areas and applications. The main processing capability of this module is to detect corners in images using the Harris corner detector, along with other minor features such as applying contrast compensation before the processing. To do that we create a pipeline of low level filters (Sobel and Gaussian). This type of basic filtering is computationally heavy, so by implementing it in hardware makes the whole system faster and more efficient. Every concept exploited in the process of creating this design is presented in the Theoretical Background Chapter and every hardware component designed is analyzed through block diagrams in the chapter proceeding the theory one. The cameras and FPGAs we used are also presented in a distinct chapter. The same goes for the experimental evaluation process we followed. Remarks on future work and feature expansion on this project are also made.