A state of the art in structured light patterns for surface profilometry

Abstract

Shape reconstruction using coded structured light is considered one of the most reliable techniques to recover object surfaces. Having a calibrated projector–camera pair, a light pattern is projected onto the scene and imaged by the camera. Correspondences between projected and recovered patterns are found and used to extract 3D surface information. This paper presents an up-to-date review and a new classification of the existing techniques. Some of these techniques have been implemented and compared, obtaining both qualitative and quantitative results. The advantages and drawbacks of the different patterns and their potentials are discussed.

Introduction

Three-dimensional measurement constitutes an important topic in computer vision, having different applications such as range sensoring, industrial inspection of manufactured parts, reverse engineering (digitization of complex, free-form surfaces), object recognition, 3D map building, biometrics, clothing design and others. The developed solutions are traditionally categorized into contact and non-contact techniques. Contact measurement techniques have been used for a long time in reverse engineering and industrial inspections. The main problems of contact techniques are their slow performance and the high cost of using mechanically calibrated passive arms [1]. Besides, the fact of touching the object is not feasible for many applications. Non-contact techniques were developed to cope with these problems, and have been widely studied. Non-contact techniques can be classified into two different categories: active and passive. In passive approaches, the scene is first imaged by video cameras from two or more points of view and correspondences between the images are found. It is important to mention that the cameras have to be previously calibrated [2]. The main problem experimented when using this approach is a sparse reconstruction since density is directly related to the texture of the object. This complicates the process of finding correspondences in the presence of textureless surfaces [3]. Therefore, passive reconstruction is rather limited to reconstruct dense 3D surfaces, due to the problem of finding correspondences [4]. Methods based on structured light (active techniques) came to cope with this issue, creating correspondences and giving specific codewords to every unitary position in the image. In this approach one of the cameras is substituted by an active device (a projector), which projects a structured light pattern onto the scene. This active device is modeled as an inverse camera, being the calibration step a similar procedure to the one used in a classical stereo-vision system [5]. The projected pattern imposes the illusion of texture onto an object, increasing the number of correspondences [6]. Therefore, surface reconstruction is possible when looking for differences between projected and recorded patterns.

In this paper, an exhaustive analysis of the different coding strategies used in active structured light is done, focusing on the advancements presented in the last years. A new classification regarding the strategy used to create the pattern is proposed, comparing some common characteristics between them. Feasibility and accuracy are analysed, giving both qualitative and quantitative results for the implemented techniques. This paper is structured as follows: Section 2 presents a classification of the different techniques. Discrete pattern based codification is studied in Section 3, while Section 4 deals with the continuous ones. The results of implementing some of the most relevant techniques are showed in Section 5, comparing their pros and cons. Finally, Section 6 concludes with a discussion of the surveyed methods, pointing out advantages and disadvantages of the most relevant ones.