Approach For Object Recognition Research Articles

Object Recognition and Localization: The Role of Tactile Sensors

Tactile sensors, because of their intrinsic insensitivity to lighting conditions and water turbidity, provide promising opportunities for augmenting the capabilities of vision sensors in applications involving object recognition and localization. This paper presents two approaches for haptic object recognition and localization for ground and underwater environments. The first approach called Batch Ransac and Iterative Closest Point augmented Particle Filter (BRICPPF) is based on an innovative combination of particle filters, Iterative-Closest-Point algorithm, and a feature-based Random Sampling and Consensus (RANSAC) algorithm for database matching. It can handle a large database of 3D-objects of complex shapes and performs a complete six-degree-of-freedom localization of static objects. The algorithms are validated by experimentation in ground and underwater environments using real hardware. To our knowledge this is the first instance of haptic object recognition and localization in underwater environments. The second approach is biologically inspired, and provides a close integration between exploration and recognition. An edge following exploration strategy is developed that receives feedback from the current state of recognition. A recognition by parts approach is developed which uses the BRICPPF for object sub-part recognition. Object exploration is either directed to explore a part until it is successfully recognized, or is directed towards new parts to endorse the current recognition belief. This approach is validated by simulation experiments.

Ontology-based inference system for adaptive object recognition

This paper presents a statistical ontology approach for adaptive object recognition in a situation-variant environment. We propose a context model based on statistical ontology that is concentrated on object recognition. Due to the effects of illumination on a supreme obstinate designing context-sensitive recognition system, we focused on designing a context-variant system using statistical ontology. Ontology, a collection of concepts and their interrelationships, provides an abstract view of an application domain. Researchers produce ontologies in order to understand and explain underlying principles and environmental factors. In this paper, we propose an ontology-based inference system for adaptive object recognition. The proposed method utilizes context ontology, context modeling, context adaptation, and context categorization to design the ontology based on illumination criteria for surveillance. After selecting the proper ontology domain, a set of actions is selected that produces better performance in that domain. We also carried out extensive experiments on these concepts in the area of object recognition in a dynamic changing environment, achieving enormous success that will enable us to proceed with our basic concepts.

A feature construction method for general object recognition

This paper presents a novel approach for object detection using a feature construction method called Evolution-COnstructed (ECO) features. Most other object recognition approaches rely on human experts to construct features. ECO features are automatically constructed by uniquely employing a standard genetic algorithm to discover series of transforms that are highly discriminative. Using ECO features provides several advantages over other object detection algorithms including: no need for a human expert to build feature sets or tune their parameters, ability to generate specialized feature sets for different objects, and no limitations to certain types of image sources. We show in our experiments that ECO features perform better or comparable with hand-crafted state-of-the-art object recognition algorithms. An analysis is given of ECO features which includes a visualization of ECO features and improvements made to the algorithm.

A Confidence Ranked Co-Occurrence Approach for Accurate Object Recognition in Highly Complex Scenes

Real-time and accurate classification of objects in highly complex scenes is an important problem for the Computer Vision community due to its many application areas. While boosting methods with the sliding window approach provide fast processing and accurate results for particular object categories, they cannot achieve the desired performance for more involved categories of objects. Recent research in Computer Vision has shown that exploiting object context through relational dependencies between object categories leads to improved accuracy in object recognition. While efforts in collective classification in images have resulted in complex algorithms suitable for offline processing, the real-time nature of the problem requires the use of simpler algorithms. In this paper, we propose a simple iterative algorithm for collective classification of all objects in an image, exploiting the global co-occurrence frequencies of object categories. The proposed algorithm uses multiple detectors trained using Gentle Boosting, where the category of the most confident estimate is propagated through the co-occurrence relations to determine the categories of the remaining unclassified objects. Experiments on a real-world dataset demonstrate the superiority of our approach over using Gentle Boosting alone as well as classic collective classification approaches modeling the full joint distribution for each object in the scene.

An Improved Edge Structure Weighted Hausdorff Distance for Template Matching

Template matching based on a Hausdorff distance (HD) approach become popular for object recognition. In this paper, we present a newly improved edge structure weighted HD (ESW-HD) algorithm for object recognition. We use edge points as the feature of the model, and construct the structure tensor by edge intensity and edge gradient. Then, the HD is weighted by the structure tensors. This work illustrates the ESW-HD algorithm by template edge matching which uses edge points and its edge adjacent structure information to perform the image matching. The experimental results show that the improved HD matching method can achieve a good performance level in terms of matching accuracy, even in a noisy environment when compared with the conventional approaches for object recognition.

Object recognition using Gabor co-occurrence similarity

We present an object recognition approach using co-occurrence similarities of Gabor magnitude textures in this paper. A novel image descriptor, multichannel Gabor magnitude co-occurrence matrices (MGMCMs), is designed to characterize Gabor textures for object representation and similarity matching. The descriptor is a generalization of multichannel color co-occurrence matrices (MCMs), which focus on using robust and discriminative magnitude textures in filtered images. Our approach starts from Gabor wavelet transformation of each object image. An exploratory learning algorithm is proposed for learning channel-adaptive magnitude truncation parameters and level parameters. This allows us to design the magnitude quantization that can reduce overall biased and peaked levels of resulting feature distributions in each channel, to avoid over-sparse co-occurrence distributions on average. The direction-based grouping is adopted for computational complexity reduction of MGMCMs extraction under a specific neighborhood mode on the grouped rescaled magnitude images of per object image. When each MGMCM is treated as a probability distribution lying on a multinomial manifold, we represent per object image as a point on a product multinomial manifold. Using multinomial geometry and metric extension technique, we construct the p-order Minkowski co-occurrence information distance for similarity matching between the albums of Gabor magnitude textures. The feasibility and effectiveness of the approach is validated by the experimental results on the Yale and FERET face databases, PolyU palmprint database, COIL-20 object database and Zurich buildings database.

Rigid 3D geometry matching for grasping of known objects in cluttered scenes

In this paper, we present an efficient 3D object recognition and pose estimation approach for grasping procedures in cluttered and occluded environments. In contrast to common appearance-based approaches, we rely solely on 3D geometry information. Our method is based on a robust geometric descriptor, a hashing technique and an efficient, localized RANSAC-like sampling strategy. We assume that each object is represented by a model consisting of a set of points with corresponding surface normals. Our method simultaneously recognizes multiple model instances and estimates their pose in the scene. A variety of tests shows that the proposed method performs well on noisy, cluttered and unsegmented range scans in which only small parts of the objects are visible. The main procedure of the algorithm has a linear time complexity resulting in a high recognition speed which allows a direct integration of the method into a continuous manipulation task. The experimental validation with a seven-degree-of-freedom Cartesian impedance controlled robot shows how the method can be used for grasping objects from a complex random stack. This application demonstrates how the integration of computer vision and soft-robotics leads to a robotic system capable of acting in unstructured and occluded environments.

Color-based underwater object recognition using water light attenuation

In this article we present a new approach for object recognition in a robotic underwater context. Color is an attractive feature because of its simplicity and its robustness to scale changes, object positions and partial occlusions. Unfortunately, in the underwater medium, the colors are modified by attenuation and are not constant with the distance. To perform a color-based recognition of an object, we develop an algorithm robust with respect to the attenuation which takes into account the light modification during its path between the light source and the camera. Therefore, a given underwater object can be identified in an image by detecting all the colors compatible with its prior known color. Our method is fast, robust and needs a very few computers resources. We successfully used it when experimenting in the sea using a system we built. It is suitable for robotic applications where computers resources are limited and shared between various embedded devices. This novel concept enables the use of the color in many applications such as target interception, object tracking or obstacle detection.

Learning the shape manifold to improve object recognition

This paper presents an approach for object recognition and shape retrieval in binary images. Object description is accomplished by contour and region-based solutions. Fusion of two categories of shape descriptors causes a considerable performance improvement in retrieval. In order to improve the recognition rate, it is important to extract the intrinsic object feature vectors in such a way that the geometrical distance between two such vectors matches the semantic relation between the two objects from which they were extracted. The main contribution in this paper is learning how to map the samples in high-dimensional observation space into the new manifold space so that the geometrically closer vectors belong to near semantics. Feature extraction is done by Isomap that belongs to a non-linear feature extraction algorithm. In the proposed method, the shortest path in the dissimilarity graph leads to less value between the samples in the same category by denoising the edge values. Experiments show that the geometrical distance between the samples on the manifold space are more compatible to the semantic distance of them. The proposed method has been compared to some well-known approaches by a variety of shape databases which includes MPEG-7 Part B, Kimia silhouettes, and Fish. The results show the effectiveness and validity of the method.

Bounding Box Object Localization Based On Image Superpixelization

One of successful approaches for object localization and recognition is sliding window approach where different candidate windows are found and evaluated and the best window is selected to represent the object. To avoid exhaustive search over all windows locations, Efficient Sub-window Search (ESS) [1] algorithm were proposed to efficiently find the best window among all sub-windows using branch and bound technique. In case of multi-class multiple object detection problems, such methods are time consuming. To handle this issue, efficient multi-object detection approach is proposed based on image superpixelization We utilize image superpixels in 2 points. (a) Such preprocessing stage is done once for an image, hence multiple detections could be fast. (b) We observe that image superpixels could help in identifying the promising candidate sub-windows, hence evaluating all sub-windows could be avoided. An efficient brute force sub-window search algorithm is proposed based on these observations. Moreover, for improving its performance, the algorithm is integrated with ESS algorithm. The proposed algorithms are assessed on the PASCAL 2006 and PASCAL 2007 test-set and show that they are faster than the state-of-the art sub-window search algorithms, while achieving a comparable performance comparing with them.

Light-Ontology Classification for Efficient Object Detection using a Hierarchical Tree Structure

Abstract This paper proposes a ontology of tree structure approach for adaptive object recognition in a situation-variant environment. In this paper, we introduce a new concept, ontology of tree structure ontology, for context sensitivity, as we found that many developed systems work in a context-invariant environment. Due to the effects of illumination on a supreme obstinate designing context-sensitive recognition system, we have focused on designing such a context-variant system using ontology of tree structure. Ontology can be defined as an explicit specification of conceptualization of a domain typically captured in an abstract model of how people think about things in the domain. People produce ontologies to understand and explain underlying principles and environmental factors. In this research, we have proposed context ontology, context modeling, context adaptation, and context categorization to design ontology of tree structure based on illumination criteria. After selecting the proper light-ontology domain, we benefit from selecting a set of actions that produces better performance on that domain. We have carried out extensive experiments on these concepts in the area of object recognition in a dynamic changing environment, and we have achieved enormous success, which will enable us to proceed on our basic concepts.

Machine Learning Approach for Object Recognition

Object recognition is basically invariant to the dramatic changes caused in objects' appearance such as location, size, viewpoint, illumination, occlusion and more by the variability in viewing conditions. In this paper, we employ an efficient approach for object recognition using invariant features and machine learning technique. The invariant features namely color, shape and texture invariant features of the objects are extracted separately with the aid of suitable feature extraction techniques. In the proposed approach, we integrate the color, shape and texture invariant features of the objects at the feature level, so as to improve the recognition performance. The fused feature set serves as the pattern for the forthcoming processes involved in the proposed approach. We employed the pattern recognition algorithms, like Discriminative Canonical Correlation (DCC) and attain distinct or identical results concerned with false positives. Our proposed approach is evaluated on the ALOI collection, a large collection of object images consists of 1000 objects recorded under various imaging circumstances. The experiments clearly demonstrate that our proposed approach significantly outperforms the state-of-the- art methods for combining color, shape and texture features. The proposed method is shown to be effective under a wide variety of imaging conditions.

Improved integral imaging approach for 3D object recognition

Improved integral imaging approach for 3D object recognition

객체인식과 분석을 위한 3D 체인코드의 적용

객체의 형상을 결정할 수 있는 요소는 크기, 경사도와 경사방향, 곡률, 기복변화, 요소간의 특성관계, 선 또는 면이 이루는 각, 특징점의 분포형태 등 매우 다양하다. 객체가 존재하는 공간은 3차원이지만 대부분의 경우 2차원 공간상에서 객체를 표현하고 이를 기반으로 객체인식을 수행하고 있다. 본 연구에서는 3차원 공간에서 객체의 형태를 판단하기 위한 방법을 제안하기 위하여 기존의 체인코드를 3차원으로 확장시켜 큐브형태의 연산자의 적용을 수행하였다. 일련의 코드로 생성되는 3D 체인코드를 분석하여 객체를 정형화하여 하향식 방법에 의한 객체 모델링의 기반이 될 수 있음을 보여 주었다. 또한 점 데이터 분할에 적용할 수 있으며 이는 라이다 데이터에 의한 모델링에 사용될 수 있다. There are various factors for determining object shape, such as size, slope and its direction, curvature, length, surface, angles between lines or planes, distribution of the model key points, and so on. Most of the object description and recognition methods are for the 2D space not for the 3D object space where the objects actually exist. In this study, 3D chain code operator, which is basically extension of 2D chain code, was proposed for object description and analysis in 3D space. Results show that the sequence of the 3D chain codes could be basis of a top-down approach for object recognition and modeling. In addition, the proposed method could be applicable to segment point cloud data such as LiDAR data.

Modulating Shape Features by Color Attention for Object Recognition

Bag-of-words based image representation is a successful approach for object recognition. Generally, the subsequent stages of the process: feature detection, feature description, vocabulary construction and image representation are performed independent of the intentioned object classes to be detected. In such a framework, it was found that the combination of different image cues, such as shape and color, often obtains below expected results. This paper presents a novel method for recognizing object categories when using multiple cues by separately processing the shape and color cues and combining them by modulating the shape features by category-specific color attention. Color is used to compute bottom-up and top-down attention maps. Subsequently, these color attention maps are used to modulate the weights of the shape features. In regions with higher attention shape features are given more weight than in regions with low attention. We compare our approach with existing methods that combine color and shape cues on five data sets containing varied importance of both cues, namely, Soccer (color predominance), Flower (color and shape parity), PASCAL VOC 2007 and 2009 (shape predominance) and Caltech-101 (color co-interference). The experiments clearly demonstrate that in all five data sets our proposed framework significantly outperforms existing methods for combining color and shape information.

A Scalable Tree-Based Approach for Joint Object and Pose Recognition

Recognizing possibly thousands of objects is a crucial capability for an autonomous agent to understand and interact with everyday environments. Practical object recognition comes in multiple forms: Is this a coffee mug (category recognition). Is this Alice's coffee mug? (instance recognition). Is the mug with the handle facing left or right? (pose recognition). We present a scalable framework, Object-Pose Tree, which efficiently organizes data into a semantically structured tree. The tree structure enables both scalable training and testing, allowing us to solve recognition over thousands of object poses in near real-time. Moreover, by simultaneously optimizing all three tasks, our approach outperforms standard nearest neighbor and 1-vs-all classifications, with large improvements on pose recognition. We evaluate the proposed technique on a dataset of 300 household objects collected using a Kinect-style 3D camera. Experiments demonstrate that our system achieves robust and efficient object category, instance, and pose recognition on challenging everyday objects.

Artificial Neural Network (ANN) based Object Recognition Using Multiple Feature Sets

In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of various objects is explored using multiple features. The objective is to configure and train an ANN to be capable of recognizing an object using a feature set formed by Principal Component Analysis (PCA), Frequency Domain and Discrete Cosine Transform (DCT) components. The idea is to use these varied components to form a unique hybrid feature set so as to capture relevant details of objects for recognition using a ANN which for the work is a Multi Layer Perceptron (MLP) trained with (error) Back Propagation learning.

Accurate Object Recognition with Shape Masks

In this paper we propose an object recognition approach that is based on shape masks--generalizations of segmentation masks. As shape masks carry information about the extent (outline) of objects, they provide a convenient tool to exploit the geometry of objects. We apply our ideas to two common object class recognition tasks--classification and localization. For classification, we extend the orderless bag-of-features image representation. In the proposed setup shape masks can be seen as weak geometrical constraints over bag-of-features. Those constraints can be used to reduce background clutter and help recognition. For localization, we propose a new recognition scheme based on high-dimensional hypothesis clustering. Shape masks allow to go beyond bounding boxes and determine the outline (approximate segmentation) of the object during localization. Furthermore, the method easily learns and detects possible object viewpoints and articulations, which are often well characterized by the object outline. Our experiments reveal that shape masks can improve recognition accuracy of state-of-the-art methods while returning richer recognition answers at the same time. We evaluate the proposed approach on the challenging natural-scene Graz-02 object classes dataset.

Nonparametric Scene Parsing via Label Transfer

While there has been a lot of recent work on object recognition and image understanding, the focus has been on carefully establishing mathematical models for images, scenes, and objects. In this paper, we propose a novel, nonparametric approach for object recognition and scene parsing using a new technology we name label transfer. For an input image, our system first retrieves its nearest neighbors from a large database containing fully annotated images. Then, the system establishes dense correspondences between the input image and each of the nearest neighbors using the dense SIFT flow algorithm [28], which aligns two images based on local image structures. Finally, based on the dense scene correspondences obtained from SIFT flow, our system warps the existing annotations and integrates multiple cues in a Markov random field framework to segment and recognize the query image. Promising experimental results have been achieved by our nonparametric scene parsing system on challenging databases. Compared to existing object recognition approaches that require training classifiers or appearance models for each object category, our system is easy to implement, has few parameters, and embeds contextual information naturally in the retrieval/alignment procedure.

Dense statistic versus sparse feature-based approach for 3D object recognition

In this article we introduce and compare two approaches towards automatic classification of 3D objects in 2D images. The first one is based on statistical modeling of wavelet features. It estimates probability density functions for all possible object classes considered in a particular recognition task. The second one uses sparse local features. For training, SURF features are extracted from the training images. During the recognition phase, features from the image are matched geometrically, providing the best fitting object for the query image. Experiments were performed for different training sets using more than 40 000 images with different backgrounds. Results show very good classification rates for both systems and point out special characteristics for each approach, which make them more suitable for different applications.