International Doctorate School in Information and Communication Technologies DIT - University of Trento SMART CMOS IMAGE SENSOR FOR 3 D MEASUREMENT

3D measurements are concerned with extracting visual information from the geometry of visible surfaces and interpreting the 3D coordinate data thus obtained, to detect or track the position or reconstruct the profile of an object, often in real time. These systems necessitate image sensors with high accuracy of position estimation and high frame rate of data processing for handling large volumes of data. A standard imager cannot address the requirements of fast image acquisition and processing, which are the two figures of merit for 3D measurements. Hence, dedicated VLSI imager architectures are indispensable for designing these high performance sensors. CMOS imaging technology provides potential to integrate image processing algorithms on the focal plane of the device, resulting in smart image sensors, capable of achieving better processing features in handling massive image data. The objective of this thesis is to present a new architecture of smart CMOS image sensor for real time 3D measurement using the sheet-beam projection methods based on active triangulation. Proposing the vision sensor as an ensemble of linear sensor arrays, all working in parallel and processing the entire image in slices, the complexity of the image-processing task shifts from O (N 2) to O (N). Inherent also in the design is the high level of parallelism to achieve massive parallel processing at high frame rate, required in 3D computation problems. This work demonstrates a prototype of the smart linear sensor incorporating full testability features to test and debug both at device and system levels. The salient features of this work are the asynchronous position to pulse stream conversion, multiple images binarization, high parallelism and modular architecture resulting in frame rate and sub-pixel resolution suitable for real time 3D measurements.