John Willian Branch

A genetic algorithm to segment range image by edge detection

The following article presents a segmentation method of range images. This method is based on edge map detection by calculations of depth gradients and orientation gradients and a genetic algorithm. The objective is to delimit the planar patches contained in images to facilitate the labelling of each region. The genetic algorithm is guided by depth gradients and orientation gradients in order to find the edge map of the image

A Method for Automatic Surface Inspection Using a Model-Based 3D Descriptor

Automatic visual inspection allows for the identification of surface defects in manufactured parts. Nevertheless, when defects are on a sub-millimeter scale, detection and recognition are a challenge. This is particularly true when the defect generates topological deformations that are not shown with strong contrast in the 2D image. In this paper, we present a method for recognizing surface defects in 3D point clouds. Firstly, we propose a novel 3D local descriptor called the Model Point Feature Histogram (MPFH) for defect detection. Our descriptor is inspired from earlier descriptors such as the Point Feature Histogram (PFH). To construct the MPFH descriptor, the models that best fit the local surface and their normal vectors are estimated. For each surface model, its contribution weight to the formation of the surface region is calculated and from the relative difference between models of the same region a histogram is generated representing the underlying surface changes. Secondly, through a classification stage, the points on the surface are labeled according to five types of primitives and the defect is detected. Thirdly, the connected components of primitives are projected to a plane, forming a 2D image. Finally, 2D geometrical features are extracted and by a support vector machine, the defects are recognized. The database used is composed of 3D simulated surfaces and 3D reconstructions of defects in welding, artificial teeth, indentations in materials, ceramics and 3D models of defects. The quantitative and qualitative results showed that the proposed method of description is robust to noise and the scale factor, and it is sufficiently discriminative for detecting some surface defects. The performance evaluation of the proposed method was performed for a classification task of the 3D point cloud in primitives, reporting an accuracy of 95%, which is higher than for other state-of-art descriptors. The rate of recognition of defects was close to 94%.

Improving Representation of the Positive Class in Imbalanced Multiple-Instance Learning

In standard supervised learning, the problem of learning from imbalanced data has been addressed to improve the performance of learning algorithms in the presence of underrepresented data. However, in Multiple-Instance Learning (MIL), where the imbalance problem is more complex, there is little discussion about it. Motivated by the need of further studies, we discuss the multiple-instance imbalance problem and propose a method to improve the representation of the positive class. Our approach looks for the target concept in positive bags and tries to strength it using an oversampling technique while removes the borderline (ambiguous) instances in positive and negative bags. We evaluate our method on several standard MIL benchmarking data sets in order to show its ability to get an enhanced representation of the positive class.

A Metric for Automatic Hole Characterization

The correct repair of three-dimensional models is still an open research problem, since acquiring processes (methods and technology) still have limitations. Although a wide range of approaches have been proposed, the main limitation is that user intervention is required to decide which regions of the surface should be corrected. We propose an automatic method for hole characterization enabling the classification of real and false anomalies without user intervention by using an irregularity measure based on two geometrical estimations: the torsion contour’s estimation uncertainty, and an approximation of geometrical shape measure surrounding the hole.

A Bag Oversampling Approach for Class Imbalance in Multiple Instance Learning

Multiple Instance Learning (MIL) is a relatively new learning paradigm which allows to train a classifier with weakly labelled data. In spite that the community has been developing different methods to learn from this kind of data, there is little discussion on how to proceed when there is an imbalanced representation of the classes. The class imbalance problem in MIL is more complex compared with their counterpart in single-instance learning because it may occur at instance and/or bag level, or at both. Here, we propose an oversampling approach at bag level in order to improve the representation of the minority class. Experiments in nine benchmark data sets are conducted to evaluate the proposed approach.

Long term three dimensional tracking of orthodontic patients using registered cone beam CT and photogrammetry

The measurements from registered images obtained from Cone Beam Computed Tomography (CBCT) and a photogrammetric sensor are used to track three-dimensional shape variations of orthodontic patients before and after their treatments. The methodology consists of five main steps: (1) the patients bone and skin shapes are measured in 3D using the fusion of images from a CBCT and a photogrammetric sensor. (2) The bone shape is extracted from the CBCT data using a standard marching cube algorithm. (3) The bone and skin shape measurements are registered using titanium targets located on the head of the patient. (4) Using a manual segmentation technique the head and lower jaw geometry are extracted separately to deal with jaw motion at the different record visits. (5) Using natural features of the upper head the two datasets are then registered with each other and then compared to evaluate bone, teeth, and skin displacements before and after treatments. This procedure is now used at the University of Alberta orthodontic clinic.

Automatic visual inspection

HighlightsAn approach for quality inspection with multi-instance learning is proposed.Using weakly labeled images reduces the labeling effort in quality inspection.Experiments show that the approach can be effectively used in real-world applications. One of the industrial applications of computer vision is automatic visual inspection. In the last decade, standard supervised learning methods have been used to detect defects in different kind of products. These methods are trained with a set of images where every image has to be manually segmented and labeled by experts in the application domain. These manual segmentations require a large amount of high quality delineations (on pixels), which can be time consuming and often a difficult task. Multi-instance learning (MIL), in contrast to standard supervised classifiers, avoids this task and can, therefore, be trained with weakly labeled images. In this paper, we propose an approach for the automatic visual inspection that uses MIL for defect detection. The approach has been tested with data from three artificial benchmark datasets and three real-world industrial scenarios: inspection of artificial teeth, weld defect detection and fishbone detection. Results show that the proposed approach can be used with weakly labeled images for defect detection on automatic visual inspection systems. This approach is able to increase the area under the receiver-operating characteristic curve (AUC) up to 6.3% compared with the nave MIL approach of propagating the bag labels.

A preliminary application of mobile agents to intrusion detection

This work presents an application of mobile agents to the topic of intrusion detection. The presented solution implements a distributed system based on agents with the capability of mobility, in such a way that when an abnormal event is detected, the agent located at the host where detection was performed moves across the network modifying the firewall rules of the involved hosts to implement a kind of prevention of the probably on going attack.

Modelo de un Personaje en un Entorno Virtual Inteligente

In this paper, a character reference model, based on its perception and reasoning, with the objective of achieving visual realism in an Intelligent Virtual Environment (IVE) is proposed. The necessary steps for obtaining the model are: i) to identify the characteristics of the objects and of the virtual character of the IVE; ii) design of the IVE model; iii) specification of the Artificial Intelligence technique (AI); iv) defining the physical level and semantic system for the integration of the character with the environment; v) validation of the proposed model in a prototype; and vi) evaluation of the results. The proposed model allows generating an appropriate dynamics between the character en the elements of the virtual environment, which is required in several applications.

Robust three-dimensional registration of range images using a new genetic algorithm

Given two approximately aligned range images of a real object, it is possible to carry out the registration of those images using numerous algorithms such as ICP. Registration is a fundamental stage in a 3D reconstruction process. Basically the task is to match two or more images taken in different times, from different sensors, or from different viewpoints. In this paper, we discuss a number of possible approaches to the registration problem and propose a new method based on the manual pre-alignment of the images followed by an automatic registration process using a novel genetic optimization algorithm. Results for real range data are presented. This procedure focuses, on the problem of obtaining the best correspondence between points through a robust search method between partially overlapped images.