Reconstrução tridimensional de feridas utilizando descritores de pontos: um estudo comparativo

Abstract

Ulcer is the generic name given to any lesion in the skin tissue or mucous. These lesions culminate in the rupture of the epithelium, resulting in the exposure of deeper tissues. The complete problem to be solved by the project in which this work is inserted refers to the development of accurate and efficient computational tools aimed at monitoring of treating chronic wounds. This fundamental follow-up importance for the determination of the evolution picture in the treatment of patient. Thus, in this work, a quantitative evaluation of the three-dimensional reconstructions obtained using Structure from Motion with the aid of 6 different point descriptors. The specific problem tackled is the to determine which point descriptor(s) are most efficient and accurate for the three-dimensional reconstruction of chronic wounds, having been chosen the descriptors SIFT, SURF, ORB, BRIEF, FREAK and DRINK. The results achieved assume that the measurement of chronic wound areas can be obtained through the use of a smartphone through the methodology addressed. Regarding processing time, descriptors based on floating points, SIFT and SURF were the ones with the highest cost computation. In calculating the area of wound surfaces, the descriptors obtained average errors of 2.61% when using the SIFT, 3.36% for the SURF, 10.03% for the BRIEF, 6.33% for the ORB, 6.27% for the FREAK and 3.74% for the DRINK, when using a configuration with 8 images.