Artur Wilkowski

Efficient generation of 3D surfel maps using RGB–D sensors

Abstract The article focuses on the problem of building dense 3D occupancy maps using commercial RGB-D sensors and the SLAM approach. In particular, it addresses the problem of 3D map representations, which must be able both to store millions of points and to offer efficient update mechanisms. The proposed solution consists of two such key elements, visual odometry and surfel-based mapping, but it contains substantial improvements: storing the surfel maps in octree form and utilizing a frustum culling-based method to accelerate the map update step. The performed experiments verify the usefulness and efficiency of the developed system.

A HMM-based system for real-time gesture recognition in movie sequences

In the paper there is presented an efficient system for dynamic gesture recognition in movie sequences based on hidden Markov models. The system uses colour-based image segmentation methods and introduces high-dimensional feature vectors to more accurately describe hand shape in the picture. It also utilizes a-priori knowledge on gestures construction in order to allow effective dimensionality reduction, hand posture classification and detection schemes. The system has demonstrated a reliable recognition properties both for static hand postures and dynamic gestures in movie sequences.

Point-Based Object Recognition in RGB-D Images

To operate autonomously a robot system needs among others to perceive the environment and to recognize the scene objects. In particular, nowadays an RGB-D sensor can be applied for vision-based perception. In this paper, two data-driven RGB-D image analysis steps, required for a reliable 3D object recognition process, are studied and appropriate algorithmic solutions are proposed. Clusters of 3D point features are detected in order to represent 3D object hypotheses. Particular clusters act as initial rough object hypotheses, allowing to constrain the subsequent model-based search for more distinctive object features in the image, like surface patches, textures and edges. In parallel, a 3D surface-based occupancy map is created, that delivers surface segments for the object recognition process. Test results are reported on various approaches to point feature detection and description, and point cloud processing.

Detection and Recognition of Compound 3D Models by Hypothesis Generation

In the paper there is proposed an integrated object detection and recognition system, based on object description given in semantic form [5]. The objects are described in a generic way in terms of parts and relations between them. The Bayesian inference system is utilized, so each object detection and recognition score has probabilistic interpretation. There are designed basic 3D models founded on the inference framework. Object instances are then detected and recognized in real-world Kinect RGBD images.

Printed steganography applied for the authentication of identity photos in face verification

Steganography methods are proposed for the authentication of the holders photo in an ICAO-consistent (travel) document. The embedded message is heavily influenced by the print-scan process, as the electronic image is first printed to be included into the document (or identity card) and is scanned next to constitute the reference template in an automatic face verification procedure. Two sufficiently robust steganography methods are designed, modifications of the “Fujitsu method” and the “triangle net” method. A third method, a commercial Digimarc tool is also applied. The methods are tested w.r.t. to face image authentication ability in a face verification procedure, using two commercial biometric SDK-s. Test results demonstrate the feasibility in biometric verification and high authentication quality of proposed approach.

Hand Gesture Modeling Using Dynamic Bayesian Networks and Deformable Templates

The paper presents a stochastic approach to articulated hand (palm shape) tracking in images. The gesture model is given in terms of a Dynamic Bayesian network that incorporates a Hidden Markov Model in order to utilize prior information on gesture structure in the tracking task. The Deformable Templates methodology is applied for hand shape modeling. Experimental evaluation of articulated hand tracking in cluttered environment using particle filtering is provided. A comparison of this method with a typical tracking approach, that makes no use of temporal gesture information, is also given.

A constraint satisfaction framework with Bayesian inference for model-based object recognition

A general (application independent) framework for the recognition of partially hidden 3-D objects in images is presented. It views the model-to-image matching as a constraint satisfaction problem (CSP) supported by Bayesian net-based evaluation of partial variable assignments. A modified incremental search for CSP is designed that allows partial solutions and calls for stochastic inference in order to provide judgments of partial states. Hence the detection of partial occlusion of objects is handled consistently with Bayesian inference over evidence and hidden variables. A particular problem of passing different objects to a machine by a human hand is solved while applying the general framework. The conducted experiments deal with the recognition of three objects: a simple cube, a Rubik cube and a tea cup.

Low-Cost Canoe Counting System for Application in a Natural Environment

This paper presents low-cost system for counting canoes and canoeists to control cannoning tourist routes. The created system was implemented on Raspberry Pi 2 and the total cost of the tracking device is less than 200

Integrating Data- and Model-Driven Analysis of RGB-D Images

. The proposed algorithm uses background subtraction and Support Vector Machines to track vessels and recognize canoes among them. The obtained results are rewarding as for low-cost solution. Depending on considered group of objects the accuracy of the algorithm reaches 84, 89.5, and 96 % for canoes, vessels, and all objects respectively.

Steganographic Authentication Method for Electronic IDs

There is a growing use of RGB-D sensors in vision-based robot perception. A reliable 3D object recognition requires the integration of image-driven and model-based analysis. Only then the low-level image-like representation can be successfully transformed into a symbolic description with equivalent semantics, considered by the ontology-level representation of an autonomous robot system. An RGB-D image analysis approach is proposed that consists of a data-driven hypothesis generation step and a generic model-based object recognition step. Initially point clusters are created assuming to represent 3D object hypotheses. In parallel, 3D surface patches are estimated, 2D image textures and shapes are classified, building multi-modal image segmentation data. In the model-driven step, a built-in knowledge about basic solids, shapes and textures is used to verify the point clusters in terms of meaningful volume-like aggregates, and to create (or to recognize) generic 3D object models.