No description available.

O monitoramento do meio ambiente e suas mudanças requer a análise de uma grade quantidade de imagens muitas vezes coletadas por satélites. No entanto, variações nos sinais devido a mudanças nas condições atmosféricas frequentemente resultam num deslocamento da distribuição dos dados para diferentes locais e datas. Isso torna difícil a distinção dentre as várias classes de uma base de dados construída a partir de várias imagens. Neste trabalho introduzimos uma nova abordagem de classificação supervisionada, chamada Classifica-Normaliza-Classifica (CNC), para amenizar o problema de deslocamento dos dados. A proposta é implementada usando dois classificadores. O primeiro é treinado em imagens não normalizadas de refletância de topo de atmosfera para distinguir dentre pixels de uma classe de interesse (CDI) e pixels de outras categorias (e.g. floresta versus não-floresta). Dada uma nova imagem de teste, o primeiro classificador gera uma segmentação das regiões da CDI e então um vetor mediano é calculado para os valores espectrais dessas áreas. Então, esse vetor é subtraído de cada pixel da imagem e portanto fixa a distribuição de dados de diferentes imagens num mesmo referencial. Finalmente, o segundo classificador, que é treinado para minimizar o erro de classificação em imagens já centralizadas pela mediana, é aplicado na imagem de teste normalizada no segundo passo para produzir a segmentação binária final. A metodologia proposta foi testada para detectar desflorestamento em pares de imagens co-registradas da Landsat 8 OLI sobre a floresta Amazônica. Experimentos usando imagens multiespectrais de refletância de topo de atmosfera mostraram que a CNC obteve maior acurácia na detecção de desflorestamento do que classificadores aplicados em imagens de refletância de superfície fornecidas pelo United States Geological Survey. As acurácias do método proposto também se mostraram superiores às obtidas pelas máscaras de desflorestamento do programa PRODES. / Monitoring natural environments and their changes over time requires the analysis of a large amount of image data, often collected by orbital remote sensing platforms. However, variations in the observed signals due to changing atmospheric conditions often result in a data distribution shift for different dates and locations making it difficult to discriminate between various classes in a dataset built from several images. This work introduces a novel supervised classification framework, called Classify-Normalize-Classify (CNC), to alleviate this data shift issue. The proposed scheme uses a two classifier approach. The first classifier is trained on non-normalized top-of-the-atmosphere reflectance samples to discriminate between pixels belonging to a class of interest (COI) and pixels from other categories (e.g. forest vs. non-forest). At test time, the estimated COI’s multivariate median signal, derived from the first classifier segmentation, is subtracted from the image and thus anchoring the data distribution from different images to the same reference. Then, a second classifier, pre-trained to minimize the classification error on COI median centered samples, is applied to the median-normalized test image to produce the final binary segmentation. The proposed methodology was tested to detect deforestation using bitemporal Landsat 8 OLI images over the Amazon rainforest. Experiments using top-of-the-atmosphere multispectral reflectance images showed that the deforestation was mapped by the CNC framework more accurately as compared to running a single classifier on surface reflectance images provided by the United States Geological Survey (USGS). Accuracies from the proposed framework also compared favorably with the benchmark masks of the PRODES program.

In this dissertation, the development and performance evaluation of new techniques for image registration and image geometric normalization, which are based on feature points extracted from images are investigated. A feature point extraction method based on scale-interaction of Mexican-hat wavelets is proposed. This feature point extractor can handle images of different scales by using a range of scaling factors for the Mexican-hat wavelet leading to feature points for different scaling factors. Experimental results show that the extracted feature points are invariant to image rotation and translation, and are robust to image degradations such as blurring, noise contamination, brightness change, etc. Further, the proposed feature extractor can handle images with scale change efficiently. A new algorithm is proposed for registration of geometrically distorted images, which may have partial overlap and may have undergone additional degradations. The global 2D affine transformations are considered in the registration process. Three main steps constitute the algorithm: extracting feature point using a feature point extractor based on scale-interaction of Mexican-hat wavelets, obtaining the correspondence between the feature points of the reference and the target images using Zernike moments of neighborhoods centered on the feature points, and estimating the transformation parameters between the first and the second images using an iterative weighted least squares algorithm. Experimental results show that the proposed algorithm leads to excellent registration accuracy using several types of images, even in cases with partial overlap between images. Further, it is robust against many image degradations and it can handle images of different scales effectively. A new technique for image geometric normalization is proposed. The locations of a set of feature points, extracted from the image, are used to obtain the normalization parameters needed to normalize the image. The geometric distortions considered in the proposed normalization technique include translation, rotation, and scaling. Experimental results show that the proposed technique yields good normalization accuracy and it is robust to many image degradations such as image compression, brightness change, noise contamination and image cropping. A blind watermarking technique for images is proposed, as an example of the possible applications of the presented geometric normalization technique. In order to enhance robustness of the watermarking technique to geometric distortions, the normalization technique is used to normalize the image, to be watermarked, during the embedding process. In the watermark detection stage, the normalization parameters for the possibly distorted watermarked image are obtained and used to transform the watermark into its normalized form. The transformed watermark is, then, correlated with the image to indicate whether the watermark is present in the image or not. Experimental results show that the proposed watermarking technique achieves good robustness to geometric distortions that include image translation, rotation, and scaling.

O monitoramento do meio ambiente e suas mudanças requer a análise de uma grade quantidade de imagens muitas vezes coletadas por satélites. No entanto, variações nos sinais devido a mudanças nas condições atmosféricas frequentemente resultam num deslocamento da distribuição dos dados para diferentes locais e datas. Isso torna difícil a distinção dentre as várias classes de uma base de dados construída a partir de várias imagens. Neste trabalho introduzimos uma nova abordagem de classificação supervisionada, chamada Classifica-Normaliza-Classifica (CNC), para amenizar o problema de deslocamento dos dados. A proposta é implementada usando dois classificadores. O primeiro é treinado em imagens não normalizadas de refletância de topo de atmosfera para distinguir dentre pixels de uma classe de interesse (CDI) e pixels de outras categorias (e.g. floresta versus não-floresta). Dada uma nova imagem de teste, o primeiro classificador gera uma segmentação das regiões da CDI e então um vetor mediano é calculado para os valores espectrais dessas áreas. Então, esse vetor é subtraído de cada pixel da imagem e portanto fixa a distribuição de dados de diferentes imagens num mesmo referencial. Finalmente, o segundo classificador, que é treinado para minimizar o erro de classificação em imagens já centralizadas pela mediana, é aplicado na imagem de teste normalizada no segundo passo para produzir a segmentação binária final. A metodologia proposta foi testada para detectar desflorestamento em pares de imagens co-registradas da Landsat 8 OLI sobre a floresta Amazônica. Experimentos usando imagens multiespectrais de refletância de topo de atmosfera mostraram que a CNC obteve maior acurácia na detecção de desflorestamento do que classificadores aplicados em imagens de refletância de superfície fornecidas pelo United States Geological Survey. As acurácias do método proposto também se mostraram superiores às obtidas pelas máscaras de desflorestamento do programa PRODES. / Monitoring natural environments and their changes over time requires the analysis of a large amount of image data, often collected by orbital remote sensing platforms. However, variations in the observed signals due to changing atmospheric conditions often result in a data distribution shift for different dates and locations making it difficult to discriminate between various classes in a dataset built from several images. This work introduces a novel supervised classification framework, called Classify-Normalize-Classify (CNC), to alleviate this data shift issue. The proposed scheme uses a two classifier approach. The first classifier is trained on non-normalized top-of-the-atmosphere reflectance samples to discriminate between pixels belonging to a class of interest (COI) and pixels from other categories (e.g. forest vs. non-forest). At test time, the estimated COI’s multivariate median signal, derived from the first classifier segmentation, is subtracted from the image and thus anchoring the data distribution from different images to the same reference. Then, a second classifier, pre-trained to minimize the classification error on COI median centered samples, is applied to the median-normalized test image to produce the final binary segmentation. The proposed methodology was tested to detect deforestation using bitemporal Landsat 8 OLI images over the Amazon rainforest. Experiments using top-of-the-atmosphere multispectral reflectance images showed that the deforestation was mapped by the CNC framework more accurately as compared to running a single classifier on surface reflectance images provided by the United States Geological Survey (USGS). Accuracies from the proposed framework also compared favorably with the benchmark masks of the PRODES program.

O monitoramento do meio ambiente e suas mudanças requer a análise de uma grade quantidade de imagens muitas vezes coletadas por satélites. No entanto, variações nos sinais devido a mudanças nas condições atmosféricas frequentemente resultam num deslocamento da distribuição dos dados para diferentes locais e datas. Isso torna difícil a distinção dentre as várias classes de uma base de dados construída a partir de várias imagens. Neste trabalho introduzimos uma nova abordagem de classificação supervisionada, chamada Classifica-Normaliza-Classifica (CNC), para amenizar o problema de deslocamento dos dados. A proposta é implementada usando dois classificadores. O primeiro é treinado em imagens não normalizadas de refletância de topo de atmosfera para distinguir dentre pixels de uma classe de interesse (CDI) e pixels de outras categorias (e.g. floresta versus não-floresta). Dada uma nova imagem de teste, o primeiro classificador gera uma segmentação das regiões da CDI e então um vetor mediano é calculado para os valores espectrais dessas áreas. Então, esse vetor é subtraído de cada pixel da imagem e portanto fixa a distribuição de dados de diferentes imagens num mesmo referencial. Finalmente, o segundo classificador, que é treinado para minimizar o erro de classificação em imagens já centralizadas pela mediana, é aplicado na imagem de teste normalizada no segundo passo para produzir a segmentação binária final. A metodologia proposta foi testada para detectar desflorestamento em pares de imagens co-registradas da Landsat 8 OLI sobre a floresta Amazônica. Experimentos usando imagens multiespectrais de refletância de topo de atmosfera mostraram que a CNC obteve maior acurácia na detecção de desflorestamento do que classificadores aplicados em imagens de refletância de superfície fornecidas pelo United States Geological Survey. As acurácias do método proposto também se mostraram superiores às obtidas pelas máscaras de desflorestamento do programa PRODES. / Monitoring natural environments and their changes over time requires the analysis of a large amount of image data, often collected by orbital remote sensing platforms. However, variations in the observed signals due to changing atmospheric conditions often result in a data distribution shift for different dates and locations making it difficult to discriminate between various classes in a dataset built from several images. This work introduces a novel supervised classification framework, called Classify-Normalize-Classify (CNC), to alleviate this data shift issue. The proposed scheme uses a two classifier approach. The first classifier is trained on non-normalized top-of-the-atmosphere reflectance samples to discriminate between pixels belonging to a class of interest (COI) and pixels from other categories (e.g. forest vs. non-forest). At test time, the estimated COI’s multivariate median signal, derived from the first classifier segmentation, is subtracted from the image and thus anchoring the data distribution from different images to the same reference. Then, a second classifier, pre-trained to minimize the classification error on COI median centered samples, is applied to the median-normalized test image to produce the final binary segmentation. The proposed methodology was tested to detect deforestation using bitemporal Landsat 8 OLI images over the Amazon rainforest. Experiments using top-of-the-atmosphere multispectral reflectance images showed that the deforestation was mapped by the CNC framework more accurately as compared to running a single classifier on surface reflectance images provided by the United States Geological Survey (USGS). Accuracies from the proposed framework also compared favorably with the benchmark masks of the PRODES program.

Prior to 1960's, the fingerprint analysis was carried out manually by human experts and for forensic purposes only. Automated fingerprint identification systems (AFIS) have been developed during the last 50 years. The success of AFIS resulted in that its use expanded beyond forensic applications and became common also in civilian applications. Mobile phones and computers equipped with fingerprint sensing devices for fingerprint-based user identification are common today. Despite the intense development efforts, a major problem in automatic fingerprint identification is to acquire reliable matching features from fingerprint images with poor quality. Images where the fingerprint pattern is heavily degraded usually inhibit the performance of an AFIS system. The performance of AFIS systems is also reduced when matching fingerprints of individuals with large age variations. This doctoral thesis presents contributions within the field of fingerprint image enhancement, segmentation and minutiae detection. The reliability of the extracted fingerprint features is highly dependent on the quality of the obtained fingerprints. Unfortunately, it is not always possible to have access to high quality fingerprints. Therefore, prior to the feature extraction, an enhancement of the quality of fingerprints and a segmentation are performed. The segmentation separates the fingerprint pattern from the background and thus limits possible sources of error due to, for instance, feature outliers. Most enhancement and segmentation techniques are data-driven and therefore based on certain features extracted from the low quality fingerprints at hand. Hence, different types of processing, such as directional filtering, are employed for the enhancement. This thesis contributes by proposing new research both for improving fingerprint matching and for the required pre-processing that improves the extraction of features to be used in fingerprint matching systems. In particular, the majority of enhancement and segmentation methods proposed herein are adaptive to the characteristics of each fingerprint image. Thus, the methods are insensitive towards sensor and fingerprint variability. Furthermore, introduction of the higher order statistics (kurtosis) for fingerprint segmentation is presented. Segmentation of the fingerprint image reduces the computational load by excluding background regions of the fingerprint image from being further processed. Also using a neural network to obtain a more robust minutiae detector with a patch rejection mechanism for speeding up the minutiae detection is presented in this thesis.

Orientador: Rubens Augusto Camargo Lamparelli / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Agricola / Made available in DSpace on 2018-08-09T05:37:08Z (GMT). No. of bitstreams: 1 Mercante_Erivelto_D.pdf: 2989061 bytes, checksum: 75872d2f0b562b1a4ee7798a720abb60 (MD5) Previous issue date: 2007 / Resumo: Estudos e pesquisas referentes ao acompanhamento da produção agropecuária têm um peso determinante e estratégico na economia do país. Nos últimos anos, essas pesquisas vêm sofrendo grandes transformações para se tornarem menos subjetivas. A associação das técnicas de sensoriamento remoto e métodos estatísticos podem proporcionar uma visão sinóptica das áreas semeadas, gerando, assim informações sobre a área plantada das culturas e a variabilidade existente nelas. Entretanto, a utilização dos dados provenientes de imagens de satélites está condicionada, principalmente, às propriedades de refletância e absorção dos componentes da superfície e pelo comportamento da atmosfera. Dentre as culturas de grande valor econômico, a soja (Glycine max (L) Merrill.) se destaca como um dos principais produtos da agricultura brasileira, assumindo grande importância econômica nas exportações. O estado do Paraná se destaca como um dos maiores produtores agrícolas do país, e a sua economia é baseada principalmente na agricultura voltada para a produção de grãos. Neste contexto, o objetivo da pesquisa foi estudar a relação entre o comportamento espectral da cultura de soja com a ao longo de seu ciclo de desenvolvimento, gerando informações e metodologias para auxiliar no acompanhamento da produção e estimativa de área da cultura na região Oeste do Paraná. As áreas monitoradas abrangem 36 municípios e duas áreas agrícolas comerciais localizadas próximas ao município de Cascavel/PR. Foram utilizadas imagens do satélite Landsat 5/TM (cena órbita/ponto 223/77) e imagens do satélite Terra sensor MODIS (produto MOD13Q1), caracterizando, assim, a passagem de escala entre as imagens dos dois sensores com resoluções espaciais diferentes. Dados de produtividade da cultura foram coletados nas escalas local (para as duas áreas monitoradas) e regional junto à Secretária de Agricultura e Abastecimento do Paraná ¿ SEAB (para os 36 municípios). A cultura da soja foi monitorada nas safras 2003/2004 e 2004/2005 utilizando imagens dos satélites. Para a utilização das imagens do satélite Landsat 5/TM de forma multitemporal foram realizados ainda os procedimentos de correção atmosférica e normalização de imagens. No intuito de caracterizar a resposta espectral da biomassa da cultura da soja e a sua relação com a produtividade final, geraram-se imagens referentes aos índices de vegetação NDVI e GVI. Foram realizadas análises de correlação e regressão entre dados de produtividade (variável predita) e dados espectrais (variável preditora) oriundos dos índices de vegetação (NDVI e GVI) em nível municipal (36 municípios) e local (duas áreas). Os resultados obtidos com a correção atmosférica e a normalização de imagens são coerentes quanto ao comportamento espectral dos alvos, vegetação e solo. Após o mapeamento das áreas com a cultura da soja (¿máscaras de soja¿), por meio das imagens temporais do Landsat 5/TM, foi possível realizar a passagem de escala para o sensor MODIS. O comportamento espectral da cultura se mostrou diferente para as imagens Landsat 5/TM com os tratamentos de refletância aparente, de superfície e de normalização. Por meio dos gráficos dos perfis espectrais traduzidos pelos índices NDVI e GVI foi possível acompanhar o ciclo de desenvolvimento da cultura da soja. As melhores correlações e regressões lineares entre os parâmetros espectrais e a produtividade final ocorreram quando considerado todo o ciclo de desenvolvimento da cultura. Quanto aos índices de vegetação utilizados NDVI e GVI, observou-se que o GVI teve comportamento com menor variação quando analisado os resultados das regressões. Em síntese, os resultados demonstraram que utilizando somente técnicas de modelagem estatística com dados espectrais foi possível estimar em até 85% a variabilidade encontrada na produtividade da soja / Abstract: Studies and researches referring to attendance of the agropecuary production have a determinant strategic importance in the economy of the country. In the last years, those researches have been suffering great transformations to become less subjectives. The association of the remote sensing technique and statistic methods can provide a synoptic view of the seeded areas, in this manner producing information about the planted area of the crops and the variability on them. However, the use of data from the images of satellite is stipulated mainly to the reflectivity properties and absorption of the surface components and also for the atmosphere behaviour. Amongst the crops of great economic value, the soybean (Glycine max (L) Merrill.) emphasizes as one of the main product of the Brazilian agriculture, assuming big economic importance in the exportations. The Paraná state stands out as one of the biggest agricultural producers of the country, and its economy is based mainly towards to the production of grains. In this context, the aim of the research was study the relation between the spectral behaviour of the soybean crop with the end yield along its development cycle, producing information and methodologies to assist the attendance of the production and area estimation of the western Paraná region crop. The areas monitored reach 36 municipal districts and two agricultural commercial areas located near of Cascavel/PR. It was used images from satellite Landsat 5/TM (orbit/point 223/77) and images of the satellite Terra sensor MODIS (product MOD13Q1), then characterizing the crossing scale between the images from both sensors with different spatial resolution. Crop yield data were collected, in the local scale (to both monitored areas) and in the regional scale together with the Secretary of Agriculture (SEAB) (to the 36 municipal districts). Through the satellites images, the soybean crop was monitored in the 2003/2004 and 2004/2005 harvests. For the use of images from the satellite Landsat 5/TM of multitemporal form was realized the proceeding of atmospheric correction and image normalization. In the aim of distinguishing the biomass spectral answer of the soybean crop and its relation to the end yield, it was created images referring to vegetation indexes NDVI and GVI. It was accomplished analyses of correlation and regression between the yield data (variable predict) and spectral data (variable predictor) derived from vegetation indexes (NDVI and GVI) in municipal level (36 municipal districts) and local (two areas). The results obtained with the atmospheric correction and image normalization presented coherent as to the spectral behaviour of the vegetation and soil target. After the area mapping with the soybean crop (soybean mask) by the temporary images of Landsat 5/TM, it was possible to realize the cross-scale to the MODIS sensor. The spectral behaviour of the crop was showed different in the Landsat 5/TM images to the apparent reflectance usage, surface and normalization. By mean of the graphics of spectral profile translated from the NDVI and GVI indexes was possible follow the development cycle of the soybean crop. The best correlations and linear regressions between the spectral parameters and the end yield will occur when it is considered all the cycle of the crop development. For the matters of vegetation indexes NDVI and GVI used, it was observed that the GVI had a less variable behaviour when analysed the results of regressions. In the sum up, the results showed that using only statistical modelling techniques with spectral data was possible estimate 85% of variability found in the soybean yield / Doutorado / Planejamento e Desenvolvimento Rural Sustentável / Doutor em Engenharia Agrícola

This thesis deals with digital image stabilization. It contains a brief overview of the problem and available methods for digital image stabilization. The aim was to design and implement image stabilization system in JAVA, which is designed for RapidMiner. Two new stabilization methods have been proposed. The first is based on the motion estimation and motion compensation using Full-search and Three-step search algorithms. The basis of the second method is the detection of object boundaries. The functionality of the proposed method was tested on video sequences with contain visible shake of the scene, which has beed created for this purpose. Testing results show that with the proper set of input parameters for the object border detection method, successful stabilization of the scene is achieved. The rate of error reduction between images is approximately about 65 to 85%. The output of the method is stabilized image sequence and a set of metadata collected during stabilization, which can be further processed in an environment of RapidMiner.