Computer Vision

 In CLIR we conduct extensive research in computer vision techniques aimed to solve   challenging problems in robotics, optical metrology and the processing of digital images.


 

Unmanned Aerial System for Remote Data Acquisition and Photogrammetric Sensing

Environmental problems have an obvious impact on virtually every aspect of our daily lives. In particular, the impact of deforestation, erosion and climate change have significantly altered the species population and distribution and have changed the coastal geography. To understand this problem and propose solutions, sensing tools that could enable professionals from various disciplines to have reliable and easily accessible data for the development of models to better understand the problem at hand, are needed. We are interested in developing methodologies for acquiring and merging data from multiple sensors to generate topographical information, with the particular interest in developing the cooperative aspect of the procurement process information across multiple agents.

 


 

Fringe pattern analysis

Fringe analysis techniques are very popular to estimate with reasonable accuracy physical quantities such as shape of objects, deformation, refractive index and temperature fields. They achieve these goals by recovering the local phase from one or a collection of interference fringe pattern images. The mathematical model of a fringe pattern is described by the equation

u=a+bcos(ψ+φ)

where a is the background illumination, b is the amplitude modulation, ψ is the spatial carrier frequency and φ is the phase map to be recovered. The problem of recovering not only φ but also a and b from the above equation is an ill-posed problem. Recently, variational techniques that aim to reduce uncertainty of the solution by introducing more information into the model by means of regularization of the unknown variables have proved to deliver a feasible solution to this problem.

 


 

Imaging problems

Imaging is the process of creating visual representation of objects. In the last years, advances in hardware and software have prompted a very fast growth in the field of imaging science. For instance, in the medical imaging field, high resolution images and new imaging techniques demand new algorithms and mathematical models to tackle new challenges in old processes such as image segmentation, image registration, image denoising and image inpaiting to name a few.