Integration of goal-driven, top-down attention and image-driven, bottom-up attention is crucial for visual search. Yet, previous research has mostly focused on models that are purely top-down or bottom-up. Here, we pr...
详细信息
Integration of goal-driven, top-down attention and image-driven, bottom-up attention is crucial for visual search. Yet, previous research has mostly focused on models that are purely top-down or bottom-up. Here, we propose a new model that combines both. The bottom-up component computes the visual salience of scene locations in different feature maps extracted at multiple spatial scales. The topdown component uses accumulated statistical knowledge of the visual features of the desired search target and background clutter, to optimally tune the bottom-up maps such that target detection speed is maximized. Testing on 750 artificial and natural scenes shows that the model’s predictions are consistent with a large body of available literature on human psychophysics of visual search. These results suggest that our model may provide good approximation of how humans combine bottom-up and top-down cues such as to optimize target detection speed.
Games have fascinated people in activities related to entertainment, education, health care, etc. The augmented reality technology, using computational support, brings the game from the computer to the user space, mak...
详细信息
Games have fascinated people in activities related to entertainment, education, health care, etc. The augmented reality technology, using computational support, brings the game from the computer to the user space, making the interaction friendlier. This paper introduces augmented reality and makes considerations on the ARToolKit software, pointing out its interactive processes. The use of augmented reality in the development of games is illustrated by five case studies of games implemented with ARToolKit. The main characteristics of each game and the exploration of the augmented reality resources are discussed.
This paper presents a distributed data mining technique based on a multiagent environment, called SMAMDD (multiagent system for distributed data mining), which uses model integration. Model integration consists in the...
详细信息
This paper presents a distributed data mining technique based on a multiagent environment, called SMAMDD (multiagent system for distributed data mining), which uses model integration. Model integration consists in the amalgamation of local models into a global, consistent one. In each subset, agents perform mining tasks locally and, afterwards, results are merged into a global model. In order to achieve that, agents cooperate by exchanging messages, aiming to improve the process of knowledge discover generating accurate results. The multiagent system for distributed data mining proposed in this paper has been compared with classical machine learning algorithms which are based on model integration as well, simulating a distributed environment. The results obtained show that SMAMDD can produce highly accurate data models
We present an evolutionary algorithm for the inference of context-free grammars from positive and negative examples. The algorithm is based on genetic programming and uses a local optimization operator that is capable...
详细信息
We present an evolutionary algorithm for the inference of context-free grammars from positive and negative examples. The algorithm is based on genetic programming and uses a local optimization operator that is capable of improving the learning task. Ordinary genetic operators are modified so as to bias the search. The system was evaluated using Tomitas language examples and results were compared with another similar approach. Results show that the proposed approach is promising and more robust than the other one.
Established statistical representations of data clusters employ up to second order statistics including mean, variance, and covariance. Strategies for merging clusters have been largely based on intra-and inter-cluste...
详细信息
Established statistical representations of data clusters employ up to second order statistics including mean, variance, and covariance. Strategies for merging clusters have been largely based on intra-and inter-cluster distance measures. The distance concept allows an intuitive interpretation, but it is not designed to merge from the viewpoint of probability distributions. We suggest an alternative strategy to compare clusters based on higher order statistics to capture the underlying probability distributions. Higher order statistics, such as multivariate skewness and kurtosis, enable a more accurate description of the shape of a cluster. Although the original definitions of kurtosis and skewness do require simultaneous involvement of all data points, our finding shows that their estimation can be decomposed into combinations of the cross moments of subsets of data. This decomposable property makes it possible to apply skewness and kurtosis to data stream clustering, where historical data are not accessible. We utilize tests for normality based on skewness and kurtosis to discover cluster pairs that can be merged to produce a less complex normal cluster even if they have different means or covariance structures.
OBF (Orthonormal Basis Function) Fuzzy models have shown to be a promising approach to the areas of nonlinear system identification and control since they exhibit several advantages over those dynamic model topologies...
详细信息
ISBN:
(纸本)9780769531960
OBF (Orthonormal Basis Function) Fuzzy models have shown to be a promising approach to the areas of nonlinear system identification and control since they exhibit several advantages over those dynamic model topologies usually adopted in the literature. Although encouraging application results have been obtained, no automatic procedure had yet been developed to optimize the design parameters of these models. This paper elaborates on the use of a genetic algorithm (GA) especially designed for this task, in which a fitness function based on the Akaike information criterion plays a key role by considering both model accuracy and parsimony aspects. The use of linear (actually affine) and nonlinear local models is also investigated. The proposed methodology is evaluated in the modeling of a real nonlinear magnetic levitation system.
In this paper we address the issue of detecting defects in wood using features extracted from grayscale images. The feature set proposed here is based on the concept of texture and it is computed from the co-occurrenc...
详细信息
In this paper we address the issue of detecting defects in wood using features extracted from grayscale images. The feature set proposed here is based on the concept of texture and it is computed from the co-occurrence matrices. The features provide measures of properties such as smoothness, coarseness, and regularity. Comparative experiments using a color image based feature set extracted from percentile histograms are carried to demonstrate the efficiency of the proposed feature set. Two different learning paradigms, neural networks and support vector machines, and a feature selection algorithm based on multi-objective genetic algorithms were considered in our experiments. The experimental results show that after feature selection, the grayscale image based feature set achieves very competitive performance for the problem of wood defect detection relative to the color image based features
In this paper we address the issue of detecting defects in wood using features extracted from grayscale images. The feature set proposed here is based on the concept of texture and it is computed from the co-occurrenc...
详细信息
This paper analyzes the effect of custom error control schemes on the energy efficiency in Bluetooth sensor networks. The energy efficiency metric considers in just one parameter the energy and reliability constraints...
详细信息
This paper analyzes the effect of custom error control schemes on the energy efficiency in Bluetooth sensor networks. The energy efficiency metric considers in just one parameter the energy and reliability constraints of the wireless sensor networks. New packet types are introduced using some error control strategies in the AUX1 packet, such as Hamming and BCH codes, with and without CRC for error detection. Two adaptive techniques are proposed that change the error control strategy based on the number of hops traversed by a packet through the network. The performance results are obtained through simulations in a channel with Rayleigh fading for networks with different number of hops, showing that error control can improve the energy efficiency of a Bluetooth-based sensor network.
暂无评论