TY  - THES
AB  - Nowadays, time-of-flight (TOF) cameras have become more popular in many practical applications, e.g., robot navigation, 3D reconstruction. Typically, they produce depth map of the entire scene through a low-cost and high-frame-rate system. Nevertheless, there exist many TOF problems, e.g., linearity errors, ambiguity range, multipath interferences (MPIs). Especially, MPIs which are usually caused by transparent object imaging or broaden illumination influence negatively on depth reconstruction results of the traditional phase-stepping method. For this reason, this thesis aims to resolve the MPI problem by carrying out multiple-frequency TOF (MFT) acquisition. According to the compressed sensing (CS) theory, since the amount of MPIs in real-life scenes is small, only a few MFT measurements are required to estimate the sparse time profile of the MPIs. However, this CS-MFT model suffers from hardware design limitation. To be concrete, under the Rayleigh resolution theorem, the low-frequency modulation of a commercial TOF camera leads to poor depth accuracy and low range resolution of the CS-MFT model. Whereas, increasing the modulation frequency is a significantly complicated task. Thus, our solution approach is to construct a super-resolution CS-MFT model with a large refinement factor. From this model, super-resolution CS techniques can reduce mismatch model errors but simultaneously cause poor sparse reconstruction performance with a highly coherent sensing matrix. This thesis introduces a variety of CS techniques to improve these reconstruction results as well as to maintain high-processing speed. They include exploring new CS reconstruction algorithms and optimizing the superresolution CS-MFT sensing matrix structure. Besides, an alternative relaxed metric with a tolerance offset is introduced for gauging the quality of spike recovery in a more accurate way. The results achieved through numerical and practical experiments show a significant improvement in accuracy and resolution of the MPI time profile reconstruction.
AU  - Nguyen, Xuan Vinh
DA  - 2017
KW  - Komprimierte Abtastung
KW  - Mehrweginterferenzen
KW  - Super-Resolution
KW  - Compressed sensing
KW  - Time-of-flight
KW  - Multipath interferences
KW  - Super-resolution
KW  - Reconstruction algorithm
LA  - eng
PY  - 2017
TI  - Super-resolution compressed sensing for resolving time-of-flight multipath interferences
UR  - https://nbn-resolving.org/urn:nbn:de:hbz:467-13890
Y2  - 2024-11-24T19:34:20
ER  -