Recognition of open vocabulary, online handwritten pages in Tamil script

In this work, we describe a system, which recognises open vocabulary, isolated, online handwritten Tamil words and extend it to recognize a paragraph of writing. We explain in detail each step involved in the process: segmentation, preprocessing, feature extraction, classification and bigram-based post-processing. On our database of 45,000 handwritten words obtained through tablet PC, we have obtained symbol level accuracy of 78.5% and 85.3% without and with the usage of post-processing using symbol level language models, respectively. Word level accuracies for the same are 40.1% and 59.6%. A line and word level segmentation strategy is proposed, which gives promising results of 100% line segmentation and 98.1% word segmentation accuracies on our initial trials of 40 handwritten paragraphs. The two modules have been combined to obtain a full-fledged page recognition system for online handwritten Tamil data. To the knowledge of the authors, this is the first ever attempt on recognition of open vocabulary, online handwritten paragraphs in any Indian language.

Propuestas para una segmentación estratégica del mercado ecológico

[ES] La creciente concienciación medioambiental y la necesidad de atender a las nuevas demandas ecológicas del mercado, obligan a las empresas a desarrollar instrumentos de análisis para ahondar en su conocimiento. La segmentación de mercados es un instrumento analítico válido para inferir características diferenciales de los consumidores ecológicos y adecuar la estrategia de marketing a las preferencias de los segmentos detectados. En este trabajo analizamos las limitaciones de los criterios de segmentación tradicionales y las ventajas de las segmentaciones multicriterio para recoger la complejidad inherente al consumidor ecológico y definir la estrategia de marketing ecológico. Así mismo, presentamos algunas de las segmentaciones del mercado ecológico más relevantes.

Travel time tomographic studies of seismic structures around subducted lithospheric slabs

The nature of the subducted lithospheric slab is investigated seismologically by tomographic inversions of ISC residual travel times. The slab, in which nearly all deep earthquakes occur, is fast in the seismic images because it is much cooler than the ambient mantle. High resolution three-dimensional P and S wave models in the NW Pacific are obtained using regional data, while inversion for the SW Pacific slabs includes teleseismic arrivals. Resolution and noise estimations show the models are generally well-resolved.

The slab anomalies in these models, as inferred from the seismicity, are generally coherent in the upper mantle and become contorted and decrease in amplitude with depth. Fast slabs are surrounded by slow regions shallower than 350 km depth. Slab fingering, including segmentation and spreading, is indicated near the bottom of the upper mantle. The fast anomalies associated with the Japan, Izu-Bonin, Mariana and Kermadec subduction zones tend to flatten to sub-horizontal at depth, while downward spreading may occur under parts of the Mariana and Kuril arcs. The Tonga slab appears to end around 550 km depth, but is underlain by a fast band at 750-1000 km depths.

The NW Pacific model combined with the Clayton-Comer mantle model predicts many observed residual sphere patterns. The predictions indicate that the near-source anomalies affect the residual spheres less than the teleseismic contributions. The teleseismic contributions may be removed either by using a mantle model, or using teleseismic station averages of residuals from only regional events. The slab-like fast bands in the corrected residual spheres are are consistent with seismicity trends under the Mariana Tzu-Bonin and Japan trenches, but are inconsistent for the Kuril events.

The comparison of the tomographic models with earthquake focal mechanisms shows that deep compression axes and fast velocity slab anomalies are in consistent alignment, even when the slab is contorted or flattened. Abnormal stress patterns are seen at major junctions of the arcs. The depth boundary between tension and compression in the central parts of these arcs appears to depend on the dip and topology of the slab.

Segmentação e classificação de imagens destinadas ao mapeamento digital do Parque Nacional da Serra dos Órgãos (PARNASO)

Na perspectiva ambiental, o Parque Nacional da Serra dos Órgãos (PARNASO) é o Parque Nacional mais pesquisado no Brasil e configura-se como uma importante Unidade de Conservação inserida no estado do Rio de Janeiro, devido à sua importância ambiental para o estado. Localizado em quatro municípios da região serrana: Teresópolis, Petrópolis, Magé e Guapimirim foi constatado que essa área tem passado por alguns problemas, relativamente recentes, de ocupação desordenada devido à expansão urbana em sua vizinhança caracterizada por pressão antrópica. Através do processamento de imagens digitais, mais especificamente as etapas de segmentação e classificação, foi possível ilustrar o processo de ocupação humana por meio de documentos cartográficos. Além de estes processos possibilitarem a geração de mapas de uso da Terra e cobertura vegetal, com o intuito de auxiliar e dar fomento à execução de atividades, o mapeamento digital configura-se numa importante ferramenta para a análise ambiental, contribuindo para o posterior zoneamento da área de estudo. Adotaram-se classes temáticas de uso e ocupação da Terra com o propósito de permitir a classificação das imagens digitais trabalhadas. São elas: afloramento rochoso, área urbana, agricultura e vegetação. Estudos foram feitos no sentido de indicar e explorar as funcionalidades das ferramentas SPRING e DEFINIENS e resultados foram comparados a partir do uso de imagens LANDSAT, CBERS, SPOT e IKONOS chegando-se a resultados de que no sistema SPRING, os melhores parâmetros a serem escolhidos foram similaridade 10 e área 400. Já para o sistema DEFINIENS, constatou-se que o processo de segmentação multinível permitiu o alcance de resultados mais rápidos, do ponto de vista computacional, do que o processo de segmentação único utilizado normalmente entre os sistemas de processamento de imagens digitais como o SPRING. Já sob a ótica do processo de classificação de imagens, a pesquisa constituiu em avaliar este mecanismo por meio de dois indicadores: o de exatidão/acurácia e o índice Kappa. Neste sentido, observaram-se tendências de melhores resultados no sistema SPRING.

Processamento digital de imagens destinado ao mapeamento da ocupação humana do alto Bonfim e análise histórica dos remanescentes florestais entre os anos de 1965 a 2006.

Através do processamento de imagens digitais, mais especificamente as etapas de segmentação e classificação, foi possível analisar o processo de ocupação humana da bacia hidrográfica do rio Bonfim, localizada no município de Petrópolis, no estado do Rio de Janeiro. Este processo possibilitou a geração de mapas de uso da terra e cobertura vegetal e configurou-se numa importante etapa para avaliação ambiental capaz de auxiliar e dar fomento à execução de atividades de gestão e monitoramento do meio ambiente e de análise histórica dos remanescentes florestais ao longo dos últimos anos. Nesta pesquisa foram adotadas classes temáticas com o propósito de permitir a classificação das imagens digitais na escala 1/40.000. As classes adotadas foram: afloramento rochoso e vegetação rupestre; obras e edificações; áreas agrícolas e vegetação. Estudos foram feitos no sentido de indicar o melhor método de classificação. Primeiramente, efetuou-se a classificação no sistema SPRING, testando-se os melhores parâmetros de similaridade e área na detecção de fragmentos, somente da classe vegetação. Houve tentativa de classificar as demais classes de uso diretamente pelo sistema SPRING, mas esta classificação não foi viável por apresentar conflitos em relação às classes, desta forma, neste sistema foi feita somente a classificação e quantificação da classe vegetação. Visando dar continuidade a pesquisa, optou-se por executar uma interpretação visual, através do sistema ArcGis, para todas as classes de uso do solo, possibilitando o mapeamento da dinâmica de evolução humana, diante da floresta de mata atlântica na área de estudos e análise histórica de seus remanescentes entre os anos dos anos 1965, 1975, 1994 e 2006.

Improving reverberant VTS for hands-free robust speech recognition

Model-based approaches to handling additive background noise and channel distortion, such as Vector Taylor Series (VTS), have been intensively studied and extended in a number of ways. In previous work, VTS has been extended to handle both reverberant and background noise, yielding the Reverberant VTS (RVTS) scheme. In this work, rather than assuming the observation vector is generated by the reverberation of a sequence of background noise corrupted speech vectors, as in RVTS, the observation vector is modelled as a superposition of the background noise and the reverberation of clean speech. This yields a new compensation scheme RVTS Joint (RVTSJ), which allows an easy formulation for joint estimation of both additive and reverberation noise parameters. These two compensation schemes were evaluated and compared on a simulated reverberant noise corrupted AURORA4 task. Both yielded large gains over VTS baseline system, with RVTSJ outperforming the previous RVTS scheme. © 2011 IEEE.

Stratigraphic Model Predictions of Geoacoustic Properties

Geoacoustic properties of the seabed have a controlling role in the propagation and reverberation of sound in shallow-water environments. Several techniques are available to quantify the important properties but are usually unable to adequately sample the region of interest. In this paper, we explore the potential for obtaining geotechnical properties from a process-based stratigraphic model. Grain-size predictions from the stratigraphic model are combined with two acoustic models to estimate sound speed with distance across the New Jersey continental shelf and with depth below the seabed. Model predictions are compared to two independent sets of data: 1) Surficial sound speeds obtained through direct measurement using in situ compressional wave probes, and 2) sound speed as a function of depth obtained through inversion of seabed reflection measurements. In water depths less than 100 m, the model predictions produce a trend of decreasing grain-size and sound speed with increasing water depth as similarly observed in the measured surficial data. In water depths between 100 and 130 m, the model predictions exhibit an increase in sound speed that was not observed in the measured surficial data. A closer comparison indicates that the grain-sizes predicted for the surficial sediments are generally too small producing sound speeds that are too slow. The predicted sound speeds also tend to be too slow for sediments 0.5-20 m below the seabed in water depths greater than 100 m. However, in water depths less than 100 m, the sound speeds between 0.5-20-m subbottom depth are generally too fast. There are several reasons for the discrepancies including the stratigraphic model was limited to two dimensions, the model was unable to simulate biologic processes responsible for the high sound-speed shell material common in the model area, and incomplete geological records necessary to accurately predict grain-size

Boundary Characterization Experiment Series Overview

Ocean acoustic propagation and reverberation in continental shelf regions is often controlled by the seabed and sea surface boundaries. A series of three multi-national and multi-disciplinary experiments was conducted between 2000-2002 to identify and measure key ocean boundary characteristics. The frequency range of interest was nominally 500-5000 Hz with the main focus on the seabed, which is generally considered as the boundary of greatest importance and least understood. Two of the experiments were conducted in the Mediterranean in the Strait of Sicily and one experiment in the North Atlantic with sites on the outer New Jersey Shelf (STRATAFORM area) and on the Scotian Shelf. Measurements included seabed reflection, seabed, surface, and biologic scattering, propagation, reverberation, and ambient noise along with supporting oceanographic, geologic, and geophysical data. This paper is primarily intended to provide an overview of the experiments and the strategies that linked the various measurements together, with detailed experiment results contained in various papers in this volume and other sources

曲线演化图像分割和森林烟火预警系统

随着计算机技术，图像采集技术和数据存储技术等的进步，图像处理的应用领域越来越广泛。很多的应用系统是综合利用了电子，通讯和图像处理等技术而开发出来的，图像处理往往是系统的核心部分。图像分割是图像处理的核心技术，也是图像处理技术中的难点。所以研究图像分割技术具有非常重要的意义。 传统的图像分割方法有：使用模板对图像进行边缘检测等；利用滤波处理，频谱分析等数字信号处理处理技术进行分割。80年代末以来，偏微分方程方法越来越多地应用到图像分割领域中，已成为图像分割的有力工具。本文对基于偏微分方程的图像分割方法进行研究，介绍单开曲线演化分割算法，并基于Mumford-Shah模型提出一种带状目标分割方法。这种方法能将图像中的带状区域从图像中分割出来－这里假定带状区域的边界可用单值函数表示。与其它方法，如边缘检测分割，C-V模型分割和单开曲线分割相比，本文提出的方法得到的分割结果有与目标的边界吻合的更好，抗噪能力强等优点。 本文介绍了通过对可见光摄像机所拍摄图像进行分析来检测火的森林烟火预警系统。该系统是通过检测烟的存在来判断是否有火情。图像处理软件是森林烟火预警系统的核心组成部分。评价火灾预警系统性能有两个标准。一个是一旦发生火灾，预警系统能否快速地发出火警信号；另一个是在没有火情时，预警系统是否不报警，即误警率是否低。图像分割在设计图像处理算法时，主要在两个地方得到应用。在图像预处理阶段，利用单开曲线演化分割算法或带状区域的分割算法将森林区域分割出来。这样是为了在对图像进行处理时消除非森林区域中的目标对识别结果的影响，降低误警率。在图像处理阶段，利用图像分割算法将烟从图像中分割出来，准确及时报警。

基于DM642的车牌识别系统设计

本文设计与实现了一种基于TMS320DM642的车牌识别系统,详细阐述了该系统的硬件构成、软件流程、检测算法以及针对DSP处理器进行的系统优化。系统通过摄像头获取汽车牌照图像,以TMS320DM642处理器为核心建立硬件平台,完成车牌定位,倾斜角校正,字符分割,字符识别等一系列算法。实验结果表明基于TMS320DM642的车牌识别系统准确、有效,应用前景广泛。

Modelling uncertainty in agricultural image analysis

C.M. Onyango, J.A. Marchant and R. Zwiggelaar, 'Modelling uncertainty in agricultural image analysis', Computers and Electronics in Agriculture 17 (3), 295-305 (1997)

Integrated Chest Image Analysis System "BU-MIA"

We introduce "BU-MIA," a Medical Image Analysis system that integrates various advanced chest image analysis methods for detection, estimation, segmentation, and registration. BU-MIA evaluates repeated computed tomography (CT) scans of the same patient to facilitate identification and evaluation of pulmonary nodules for interval growth. It provides a user-friendly graphical user interface with a number of interaction tools for development, evaluation, and validation of chest image analysis methods. The structures that BU-MIA processes include the thorax, lungs, and trachea, pulmonary structures, such as lobes, fissures, nodules, and vessels, and bones, such as sternum, vertebrae, and ribs.

MosaicShape: Stochastic Region Grouping with Shape Prior

A novel method that combines shape-based object recognition and image segmentation is proposed for shape retrieval from images. Given a shape prior represented in a multi-scale curvature form, the proposed method identifies the target objects in images by grouping oversegmented image regions. The problem is formulated in a unified probabilistic framework and solved by a stochastic Markov Chain Monte Carlo (MCMC) mechanism. By this means, object segmentation and recognition are accomplished simultaneously. Within each sampling move during the simulation process,probabilistic region grouping operations are influenced by both the image information and the shape similarity constraint. The latter constraint is measured by a partial shape matching process. A generalized parallel algorithm by Barbu and Zhu,combined with a large sampling jump and other implementation improvements, greatly speeds up the overall stochastic process. The proposed method supports the segmentation and recognition of multiple occluded objects in images. Experimental results are provided for both synthetic and real images.

An Active Pattern Recognition Architecture for Mobile Robots

An active, attentionally-modulated recognition architecture is proposed for object recognition and scene analysis. The proposed architecture forms part of navigation and trajectory planning modules for mobile robots. Key characteristics of the system include movement planning and execution based on environmental factors and internal goal definitions. Real-time implementation of the system is based on space-variant representation of the visual field, as well as an optimal visual processing scheme utilizing separate and parallel channels for the extraction of boundaries and stimulus qualities. A spatial and temporal grouping module (VWM) allows for scene scanning, multi-object segmentation, and featural/object priming. VWM is used to modulate a tn~ectory formation module capable of redirecting the focus of spatial attention. Finally, an object recognition module based on adaptive resonance theory is interfaced through VWM to the visual processing module. The system is capable of using information from different modalities to disambiguate sensory input.

Synchronized Oscillations During Cooperative Feature Lining in a Cortical Model of Visual Perception

A neural network model of synchronized oscillations in visual cortex is presented to account for recent neurophysiological findings that such synchronization may reflect global properties of the stimulus. In these experiments, synchronization of oscillatory firing responses to moving bar stimuli occurred not only for nearby neurons, but also occurred between neurons separated by several cortical columns (several mm of cortex) when these neurons shared some receptive field preferences specific to the stimuli. These results were obtained for single bar stimuli and also across two disconnected, but colinear, bars moving in the same direction. Our model and computer simulations obtain these synchrony results across both single and double bar stimuli using different, but formally related, models of preattentive visual boundary segmentation and attentive visual object recognition, as well as nearest-neighbor and randomly coupled models.