Active chromatic depth from defocus for industrial inspection

Abstract

In this paper we propose a new concept for a compact 3D sensor dedicated to industrial inspection, combining chromatic Depth From Defocus (DFD) and structured illumination. Depth is estimated from a single image using local estimation of the defocus blur. As industrial objects usually show poor texture information, which is crucial for DFD, we rely on structured illumination. In contrast with state of the art approaches for active DFD, which project sparse patterns on the scene, our method exploits a dense textured pattern and provides dense depth maps of the scene. Besides, to overcome depth ambiguity and dead zone of DFD with a classical camera, we use an unconventional lens with chromatic aberration, providing spectrally varying defocus blur in the camera color channels. We provide comparisons of depth estimation performance for several projected patterns at various scales based on simulation and real experiments. The proposed method is then qualitatively evaluated on a real industrial object. Finally we discuss the perspectives of this work especially in terms of co-design of an 3D active sensor using DFD.

Learning scene and blur model for active chromatic depth from defocus

“…In this paper we first propose a new single image DFD algorithm. As [4][5][6][7][8], our blur estimation method is based on the selection of a depth among a set of potential depths using a criterion derived from a maximum likelihood approach in a Bayesian framework. But here, we propose to directly learn the image model used to define the likelihood from calibration acquisitions.…”

“…Several methods for active DFD have been proposed in the literature, yet except from references [8,14], most of them use a sparse projection patterns such as evenly spaced vertical lines [13,[15][16][17][18] or evenly spaced dots [19,20]. Blur is then estimated by measuring the spread or the brightness of each elements of the pattern in the image.…”

“…Nayar et al [14] propose to use a dense illumination pattern in the form of a black and white grid with optimized grid parameters to improve depth estimation performance, in a two-frames DFD framework. In Buat et al [8] comparison of several projected patterns for active DFD is conducted for dense depth map estimation using a state of the art single image DFD algorithm proposed by Trouvé et al [7]. This algorithm, based on a generic structured scene covariance model, will be referred to as SC-DFD in this paper.…”

“…Acquisition is then easier and provides a larger field of applications than DFF, but the processing is more difficult, specially in the Single Image case (SIDFD) as both scene and blur are unknown. In the literature of local blur estimation [4][5][6][7][8], a common approach is to define a scene density of probability model while the relation between depth and blur is obtained through a calibration process of a finite set of potential PSFs. However, both approximations limit the accuracy of the depth estimation.…”

“…However it relies on the parallax between images of the camera and projected pattern onto the scene, which induces the same limitations as passive stereoscopy. To tackle the case of low textured object, active Depth-From-Defocus methods have been proposed [8,[13][14][15][16][17][18]. A projector is used to add an artificial texture onto the scene and hence increase defocus blur cues.…”