Research area
Stereo 3D Vision
Dual-camera stereo imaging, image processing and graphics for recovering 3D information, from classical geometry to Transformer models.
Cameras see in two dimensions, but robots, inspection systems and interactive devices need to know how far away things are. Recovering depth reliably from ordinary cameras, under changing light and with limited computing power, remains an open engineering problem.
We work with dual-camera stereo setups that capture the same scene from two viewpoints and compute 3D information from the disparity between them. The pipeline covers camera calibration, image rectification, stereo matching and depth estimation, and computer-graphics methods are used to display and check the reconstructed 3D data. Classical image processing and geometry, prototyped in MATLAB, sit alongside deep-learning vision models, including Transformers.
The lab’s image-processing work goes back a long way: earlier papers measure wafer probe marks from images and detect wafer microcracks. Depth from this direction also supports robotic arm manipulation, and the lab’s embedded tradition pushes the models toward small, affordable hardware rather than large servers.
Related publications
-
Application of image processing to wafer probe mark area calculation
2010 5th IEEE Conference on Industrial Electronics and Applications (ICIEA), pp. 414–419
Measures the area of the marks that test needles leave on wafer pads from high-magnification microscope images, replacing the manual practice of drawing an oval around each mark; the method is faster, more accurate and does not vary from person to person.
Conference DOI
Interested? Visit the lab or write to us.