In this paper, the advantages of ensemble methods are adapted to image categorization. A novel method is introduced for image categorization by constructing vocabulary ensembles using different clustering algorithms i...
详细信息
In this paper, the advantages of ensemble methods are adapted to image categorization. A novel method is introduced for image categorization by constructing vocabulary ensembles using different clustering algorithms in the popular vocabulary approach. The vocabulary approach describes an image as a bag of discrete visual words, where the frequency distributions of these words are used for image categorization. Based on vocabularies formed by various clustering algorithms, a classifier ensemble is learned, which can jointly exploit different data structure of high dimensional descriptors. High classification accuracies of the proposed algorithm are demonstrated on three different datasets.
A particular kind of invariants known as moment invariants have been used for object recognition and identification due to its invariabilities under translation, rotation and similitude transformation and approximate ...
详细信息
A particular kind of invariants known as moment invariants have been used for object recognition and identification due to its invariabilities under translation, rotation and similitude transformation and approximate invariability even under small-scale perspective transformation, while contours representing shapes and structures play key role in people's tasks of reorganization and identification. A combining approach incorporating contours into moment invariant is proposed in this paper such that advantages of both are employed. Redefined on contours of both the whole objects and their inners, not just are moment invariants related to geometric information but they are less sensitive to non-essential changes and less complex for computations. Vectors formed by invariants act as features of objects and are classified by Nearest Neighborhood. If common features of kinds of objects are defined in advance, this algorithm can be used for objects belonging to those kinds. Time and space complex are rather low and experimental results show desirable accuracy.
It might be a very effective way for computational vision model to utilize the productions of neurophysiology and biology. Retina does astonishing work in the early vision. In this paper, according to anatomic structu...
详细信息
It might be a very effective way for computational vision model to utilize the productions of neurophysiology and biology. Retina does astonishing work in the early vision. In this paper, according to anatomic structure, a multi-layer digital retina model is presented to simulate biological retina and it is placed in a physical visual field of a reduced eye to analyze why the retina can be capable of fulfilling all tasks, and further more, to analyze the characteristics of every layer cell in retina in information processing. The model is designed to achieve a kind of balance among hardware complexity, computing load and performance on the condition of sufficient information collection. This research also contributes to the design and implementation of artificial retina chips to improve perception of visually impaired patients.
It has been proved that acquired training is important to the development of stereopsis experience. Month-old babies already have the initial experience of invariance recognition of 3D objects. There is a slight lack ...
详细信息
It has been proved that acquired training is important to the development of stereopsis experience. Month-old babies already have the initial experience of invariance recognition of 3D objects. There is a slight lack of precision in the interpretation of biological vision. However, the small cost and the fast speed in calculation meet the requirements of invariance recognition, the rich visual experience in which play an important role. But what is the experience, how to acquire and how to use, these problems have never been satisfactorily resolved. In this paper we simulate the learning of visual experience in children, and solve a view angle estimated problem by using self-organizing network, which make the hidden experience clarified. Compared to the Classic camera calibration, which a large number of parameters need to be estimated, this method needs only one image and does not aim to 3D reconstruction. By avoiding the complex calibration and registration process, an amount of computation has been reduced. Visual experiences are all obtained from the most ordinary examples, and the characterization based on the geometric feature. Therefore, this method has strong expansibility and good generalization ability.
Knowledge-based problem solving requires a conceptual system that is comparatively rich and complete, especially when the problem is domain *** methods about knowledge acquisition, representation and usage in classica...
详细信息
Knowledge-based problem solving requires a conceptual system that is comparatively rich and complete, especially when the problem is domain *** methods about knowledge acquisition, representation and usage in classical Knowledge Engineering can only adapt themselves to domain restricted *** is because it doesn't take a developmental view to construct conceptual system and consequently it is confronted with Framework *** cognitive Psychology, the study of conceptual system has an in-depth cognitive investigate on issues of development and ***, there lack investigations on the details of system construction and *** on the theory of Developmental Psychology, this paper proposes an object-based representation method for conceptual system, focusing on the representation and development of concepts on four levels: Implicit (I), Explicit 1 (E1), Explicit 2 (E2) and Explicit 3 (E3) *** will contribute well to the adaptability and flexibility in the reasoning and problem solving of knowledge-based systems.
Machine vision is an active branch of Artificial Intelligence. An important problem in this area is the balance among efficiency, accuracy and huge computing. The visual system of human can keep watchfulness to the pe...
详细信息
Machine vision is an active branch of Artificial Intelligence. An important problem in this area is the balance among efficiency, accuracy and huge computing. The visual system of human can keep watchfulness to the perimeter of visual field while at same time their central attention is focused to the center of visual field for fine information processing. This mechanism of computing resource assignment could ease the demand for huge and complex hardware structure. Therefore designing computer model based on biological visual
One of the interferences between inheritance and concurrency is inheritance anomaly. From the view of cognitive computational neuroscience, a direct information representation method is presented based on neural syste...
详细信息
One of the interferences between inheritance and concurrency is inheritance anomaly. From the view of cognitive computational neuroscience, a direct information representation method is presented based on neural system dynamics and graphic theory. A group of neurons and their connections representing perceptual information directly and the dynamical behaviors of neurons are defined firstly, and then a two-layer neural network is designed to record characteristics of stimulus and connect a specialized neural circuit that responding to the perception of that stimulus respectively. This could be achieved by the structure learning algorithm. The circuit constituted by neurons in two layers is also served as an associative memory of stimulus whose credibility is decided by the degree of connection of the circuit. The direct representation method is of very significance to the research of semantic representation and inference driven by semantics in artificial intelligence.
暂无评论