Josep Forest

A review of recent range image registration methods with accuracy evaluation

The three-dimensional reconstruction of real objects is an important topic in computer vision. Most of the acquisition systems are limited to reconstruct a partial view of the object obtaining in blind areas and occlusions, while in most applications a full reconstruction is required. Many authors have proposed techniques to fuse 3D surfaces by determining the motion between the different views. The first problem is related to obtaining a rough registration when such motion is not available. The second one is focused on obtaining a fine registration from an initial approximation. In this paper, a survey of the most common techniques is presented. Furthermore, a sample of the techniques has been programmed and experimental results are reported to determine the best method in the presence of noise and outliers, providing a useful guide for an interested reader including a Matlab toolbox available at the webpage of the authors.

Optimised De Bruijn patterns for one-shot shape acquisition

Coded structured light is an optical technique based on active stereovision which allows shape acquisition. By projecting a suitable set of light patterns onto the surface of an object and capturing images with a camera, a large number of correspondences can be found and 3D points can be reconstructed by means of triangulation. One-shot techniques are based on projecting an unique pattern so that moving objects can be measured. A major group of techniques in this field define coloured multi-slit or stripe patterns in order to obtain dense reconstructions. The former type of patterns is suitable for locating intensity peaks in the image while the latter is aimed to locate edges. In this paper, we present a new way to design coloured stripe patterns so that both intensity peaks and edges can be located without loss of accuracy and reducing the number of hue levels included in the pattern. The results obtained by the new pattern are quantitatively and qualitatively compared to similar techniques. These results also contribute to a comparison between the peak-based and edge-based reconstruction strategies.

Registration of moving surfaces by means of one-shot laser projection

The acquisition of three-dimensional models of a given surface is a very interesting subject in computer vision. Most of techniques are based on the use of laser range finders coupled to a mechanical system that scans the surface. These techniques lacks of accuracy in the presence of vibrations or non-controlled surface motion because of the misalignments between the acquired images. In this paper, we propose a new one-shot pattern which benefits from the use of registration techniques to recover a whole surface in the presence of non-controlled motion.

Sensorial and Navigation Systems for a Mobile Robot (Roger)

Abstract This paper presents basically the sensorial system, the navigation systems and the data communication of the ROGER autonomous mobile robot. This robot is capable to operate in structured indoor environments and in certain outdoor environments. The robot can place itself independently in its workspace, and can perform some basic manipulation operations. These features make possible the use this mobile robot in a wide range of applications (industrial, social, etc.)

Image Ranging System for Underwater Applications

Abstract This paper presents a computer vision and a slit laser beam which have been integrated in order to build up a fully operative 3D vision measurement system. The paper explains the laser modelling, the extrinsic and intrinsic camera calibration using a linear model, and some experimental results of 3D reconstruction. A method for automatically calibrating the laser system with few restrictions of position and orientation with respect to the camera is explained in detail. This system has been developped to improve the autonomy of the ROV GARBI, increasing its computer vision capabilities, incorporating 3D sensing features. Moreover, this paper addresses the constraints related to underwater imaging by using such a sensor, i.e. swaying and drifting of the vehicle due to marine currents or turbulences, water scattering and high light attenuation.

An Experimental Benchmark for Point Set Coarse Matching

Coarse Matching of point clouds is a fundamental problem in a variety of computer vision applications. While many algorithms have been developed in recent years to address its different aspects, the lack of unified measures and commonly agreed upon data hampers algorithm performances comparison. Additionally, a large number of contributions are tested only with synthetic or processed data. This is a problem as the resulting scenario is somewhat less challenging and does not always conform to practical application conditions. In this paper, we present a new, publicly available database that aims at overcoming the existing problems, provide researchers with a useful tool to compare new contributions to existing ones and represent a step towards standardization. The database contains both processed and unprocessed data with attention to specially challenging datasets. It also includes information on correct solution, presence of noise, overlap percentages and additional information that will allow researchers to focus only on specific parts of the matching pipeline.

Underwater 3D Laser Scanners: The Deformation of the Plane

Development of underwater 3D perception is necessary for autonomous manipulation and mapping. Using a mirror-galvanometer system to steer a laser plane and using triangulation, it is possible to produce full 3D perception without the need of moving the sensor. If the sensor does not meet certain hardware requirements, the laser plane is distorted when it passes through the different media (air–viewport–water). However, the deformation of this plane has not been studied. In this work a ray-tracing model is presented to study the deformation of the laser plane. To validate it, two types of datasets have been used, one synthetically generated using the model presented below, and another one using real data gathered underwater with an actual laser scanner. For both datasets an elliptic cone is fitted on the data and compared to a plane fit (the surface commonly used for triangulation). In the two experiments, the elliptic cone proved to be a better fit than the plane.

A study on the robustness of shape descriptors to common scanning artifacts

Registration is a fundamental problem in a myriad of applications ranging from heritage reconstruction to industrial applications. Descriptors are an important part of the registration pipeline as well as a very active research field. However, the sets used to illustrate descriptor performance have often undergone several preprocessing steps such as noise filtering, hole filling or outlier removal. These steps simplify the problem but are not readily available in many applications. In this paper we compare the performances of 4 state of the art shape descriptors: SHOT [1], Spin Image [2], FPFH [3] and 3DSC [4]. Experiments were carried out with real as well as synthetic data paying special attention to issues commonly present in real data (noise, outliers and low overlap). The method obtaining a best result overall is SHOT, based mostly on the results with synthetic data. Experiments with real data showed how state of the art descriptors are not yet able to produce optimal results in the most challenging scenarios.

3D Laser Scanner for Underwater Manipulation

Nowadays, research in autonomous underwater manipulation has demonstrated simple applications like picking an object from the sea floor, turning a valve or plugging and unplugging a connector. These are fairly simple tasks compared with those already demonstrated by the mobile robotics community, which include, among others, safe arm motion within areas populated with a priori unknown obstacles or the recognition and location of objects based on their 3D model to grasp them. Kinect-like 3D sensors have contributed significantly to the advance of mobile manipulation providing 3D sensing capabilities in real-time at low cost. Unfortunately, the underwater robotics community is lacking a 3D sensor with similar capabilities to provide rich 3D information of the work space. In this paper, we present a new underwater 3D laser scanner and demonstrate its capabilities for underwater manipulation. In order to use this sensor in conjunction with manipulators, a calibration method to find the relative position between the manipulator and the 3D laser scanner is presented. Then, two different advanced underwater manipulation tasks beyond the state of the art are demonstrated using two different manipulation systems. First, an eight Degrees of Freedom (DoF) fixed-base manipulator system is used to demonstrate arm motion within a work space populated with a priori unknown fixed obstacles. Next, an eight DoF free floating Underwater Vehicle-Manipulator System (UVMS) is used to autonomously grasp an object from the bottom of a water tank.

Specialized processor for 3D real-time tracking of moving objects using the knowledge of their chromatic components

Tracking in real-time is a complex problem handled by researchers in may different manners. In a general way, a tracking system can be organized in some pipelined modules: the first one provides sensorial capabilities using some different technologies such as radar, ultrasonic, IR or CCD cameras, etc. The second one works as a pre-processing device, usually in real time, and supplies filtered information. The third step usually deals with algorithms, supplying information to the last and most important phase: scene understanding and taking decisions. This paper describes a whole computer vision system for tracking moving objects in real-time. First, a high-speed image processing hardware for computing 2D positions is presented. This hardware is the basic sensor of a higher-level system where the integration of several views of the scene allows the computation of 3D positions. Then, the mathematical formulation to solve the trinocular problem is pointed out.