Evolutionary algorithms for a mixed stereovision uncalibrated 3D reconstruction

This paper proposes an original 3D shape reconstruction which is a mixture of the passive and active stereovision systems. Similarly to the passive stereovision systems, two cameras are used to acquire the images. As for the active stereovision methods, the detection of the points of interest (POIs) and the matching problem are solved by using a structured-light pattern projected onto the analysed object. An encoding is proposed to ease the matching procedure. Then, Evolutionary Algorithms (EAs) are designed to calculate the depth of the detected POIs. Numerous experiments are conducted to validate the different steps of the proposed method.