Files
Abstract
Over the past decades, 3D optical measurement techniques, including passive and active approaches, have been extensively studied across diverse fields. Among these, fringe projection profilometry (FPP), an active triangulation-based 3D measurement technique, is widely recognized for its speed, accuracy, and adaptability. This dissertation aims to improve FPP’s measurement accuracy and completeness further and expand the data modality to broaden its applicability.
To improve accuracy on objects with sharp depth changes, a circular fringe projection technique is introduced, along with system calibration and 3D reconstruction. Its performance is compared to traditional linear FPP using both standard spheres and complex electronic samples, with results showing superior reconstruction accuracy. Factors affecting the calibration method’s accuracy are also explored through sensitivity and phase error analyses.
To mitigate the influence of occlusion, where abrupt geometric changes prevent projector illumination and result in incomplete 3D reconstructions, this dissertation proposes texture-guided phase-to-depth networks that integrate texture maps with phase information, reducing shadow-induced errors. Experimental results from simulations and real-world scans validate the networks' effectiveness.
Expanding FPP’s applicability, this dissertation introduces a 4D line-scan hyperspectral imaging system that captures both 3D spatial data and spectral information in the visible and near-infrared (VNIR) range, eliminating post-processing registration. Additionally, the work extends to a 4D Vis-SWIR line-scan hyperspectral imaging system, capturing 3D data and spectral images across the visible to shortwave infrared (Vis-SWIR) range. This dissertation proposes a line-scan homography method to align VNIR and SWIR hyperspectral images and a geometric transformation technique to register hyperspectral images with 3D data. Results confirm registration accuracy and demonstrate applicability in food quality evaluation.
In conclusion, this dissertation contributes to the field of 3D optical measurement by addressing critical challenges in FPP and introducing innovative techniques that enhance accuracy and expand the dimensionality of data acquisition. These advancements hold considerable potential for both academic research and industrial applications, providing deeper insights into a range of 3D optical measurement technologies.
To improve accuracy on objects with sharp depth changes, a circular fringe projection technique is introduced, along with system calibration and 3D reconstruction. Its performance is compared to traditional linear FPP using both standard spheres and complex electronic samples, with results showing superior reconstruction accuracy. Factors affecting the calibration method’s accuracy are also explored through sensitivity and phase error analyses.
To mitigate the influence of occlusion, where abrupt geometric changes prevent projector illumination and result in incomplete 3D reconstructions, this dissertation proposes texture-guided phase-to-depth networks that integrate texture maps with phase information, reducing shadow-induced errors. Experimental results from simulations and real-world scans validate the networks' effectiveness.
Expanding FPP’s applicability, this dissertation introduces a 4D line-scan hyperspectral imaging system that captures both 3D spatial data and spectral information in the visible and near-infrared (VNIR) range, eliminating post-processing registration. Additionally, the work extends to a 4D Vis-SWIR line-scan hyperspectral imaging system, capturing 3D data and spectral images across the visible to shortwave infrared (Vis-SWIR) range. This dissertation proposes a line-scan homography method to align VNIR and SWIR hyperspectral images and a geometric transformation technique to register hyperspectral images with 3D data. Results confirm registration accuracy and demonstrate applicability in food quality evaluation.
In conclusion, this dissertation contributes to the field of 3D optical measurement by addressing critical challenges in FPP and introducing innovative techniques that enhance accuracy and expand the dimensionality of data acquisition. These advancements hold considerable potential for both academic research and industrial applications, providing deeper insights into a range of 3D optical measurement technologies.