Dimensional metrology is of prime importance in the industrial and scientific domains, particularly in the field of quality control of manufactured products. In robotics too: without dimensional sensors, robots would be nothing else more than automatons, going through the same repetitive tasks again and again in a carefully controlled environment. This thesis deals with a noncontact measurement technique involving active vision, called triangulation under laser plane lighting. This optomechatronic method consists in projecting a laser sheet on an object or a surface under test, and analyzing the intersecting curve on an image taken by a camera. It allows making a wide variety of dimensional, noncontact and nondestructive, measurements (length, area, volume, diameter, curvature, reverse engineering,...). The original approach of the work is that the focus is brought on the determination of specific, restricted dimensional information on objects of diverse, but a priori known, shapes with the objective of achieving metrological performances in agreement with the industrial requirements. Furthermore, ambulatory instruments­i.e. devices that are at least portable, or even handheld­are exclusively aimed, using low-cost components. Another objective is to obtain systems for which an industrial transposition to innovative instrumental products is feasible. The text is divided in two distinct parts, both strongly correlated. The first one deals with all the theoretical aspects of the method: camera model, passage from 2-D image to 3-D scene, image processing, calibration, accuracy analysis... The performances of the developed models are also studied, in terms of robustness and repeatability. The second part describes four innovative applications of our own: the diameter measurement of cylindrical and of spherical objects, dimensional measurements in the building sector and the determination of the road surface microtexture. The achieved accuracies are globally of about 1%.

Dimensional metrology is of prime importance in the industrial and scientific domains, particularly in the field of quality control of manufactured products. In robotics too: without dimensional sensors, robots would be nothing else more than automatons, going through the same repetitive tasks again and again in a carefully controlled environment. This thesis deals with a noncontact measurement technique involving active vision, called triangulation under laser plane lighting. This optomechatronic method consists in projecting a laser sheet on an object or a surface under test, and analyzing the intersecting curve on an image taken by a camera. It allows making a wide variety of dimensional, noncontact and nondestructive, measurements (length, area, volume, diameter, curvature, reverse engineering,...). The original approach of the work is that the focus is brought on the determination of specific, restricted dimensional information on objects of diverse, but a priori known, shapes with the objective of achieving metrological performances in agreement with the industrial requirements. Furthermore, ambulatory instruments­i.e. devices that are at least portable, or even handheld­are exclusively aimed, using low-cost components. Another objective is to obtain systems for which an industrial transposition to innovative instrumental products is feasible. The text is divided in two distinct parts, both strongly correlated. The first one deals with all the theoretical aspects of the method: camera model, passage from 2-D image to 3-D scene, image processing, calibration, accuracy analysis... The performances of the developed models are also studied, in terms of robustness and repeatability. The second part describes four innovative applications of our own: the diameter measurement of cylindrical and of spherical objects, dimensional measurements in the building sector and the determination of the road surface microtexture. The achieved accuracies are globally of about 1%.

This thesis describes a project to create a hardware based 3D interior scanner. This was based on a previous project that created a scanner optimised for interior conditions, using structured light triangulation. The original project referred to as the Mark-I scanner, performed its control and processing on a PC and the primary goal of this project was to re-implement this system using hardware, making the scanner more portable and simpler to use. The Mark-I system required a specialised camera which had an unusually high noise associated with it, so a secondary goal was to investigate whether this camera could be replaced with a superior model or this noise corrected. A Mark-II scanner system was created using FPGA processing and control implemented in the VHDL language. This read from a CMOS camera, controlled the system's motor and laser, generated 3D points and communicated with users. A suitable camera was not found and the Mark-I scanners camera was found to have been damaged and become unusable, so a simulation environment was constructed that simulated the operation of the scanner, created 3D images for it to process, and tested its results. Chapter 1 of this thesis outlines the goals of this pro ject and describes the Mark-I system. Chapter 2 describes the theory and properties of the Mark-I system, and chapter 3 describes the work undertaken to replace the scanner's sensor. Chapter 4 describes the system created to interface to CMOS sensors, and chapter 5 outlines the theory involved in calculating 3D points using structured light triangulation. The final hardware scanner, and the simulation system used to test it, are then described in chapter 6.