Detecção de região saliente em imagens usando dissimilaridade de cor e amostragem por pixeis aleatórios

The purpose of this work is to demonstrate and to assess a simple algorithm for automatic estimation of the most salient region in an image, that have possible application in computer vision. The algorithm uses the connection between color dissimilarities in the image and the image’s most salient region. The algorithm also avoids using image priors. Pixel dissimilarity is an informal function of the distance of a specific pixel’s color to other pixels’ colors in an image. We examine the relation between pixel color dissimilarity and salient region detection on the MSRA1K image dataset. We propose a simple algorithm for salient region detection through random pixel color dissimilarity. We define dissimilarity by accumulating the distance between each pixel and a sample of n other random pixels, in the CIELAB color space. An important result is that random dissimilarity between each pixel and just another pixel (n = 1) is enough to create adequate saliency maps when combined with median filter, with competitive average performance if compared with other related methods in the saliency detection research field. The assessment was performed by means of precision-recall curves. This idea is inspired on the human attention mechanism that is able to choose few specific regions to focus on, a biological system that the computer vision community aims to emulate. We also review some of the history on this topic of selective attention.

Padrões de organização das assembleias de aves florestais em uma paisagem fragmentada de Floresta Atlântica

Brazil is home to one of the richest avifaunas the world, which is subject to high levels of environmental degradation, in particular forest fragmentation. The Atlantic Forest biome depicts this history of devastation and today remains as small isolated fragments on highly degraded landscapes. This project aimed to evaluate the effects of forest fragmentation in an area with Atlantic Forest remnants in northern Paraná (Brazil) on the distribution and the organization of assemblage of forest birds and tested the hypothesis that the structure of the assembly in the fragments is different than expected by chance. We did four qualitative samplings of birds in three sets of forest fragments in the landscape, each with three fragments: large, medium and small. The method applied in the sampling was point counts along transects, traveled randomly for four hours in each fragment. Samples were taken in two periods: from September to November / 2013, and between March and May / 2014. The structure of the meeting was assessed by rates of co-occurring species (Checkerboard and CScore) and α diversity patterns (wealth) and β (turnover of species), while the landscape structure was analyzed from the parameters: area, distance between fragments, fractal dimension, edge density, fragment shape index and nuclear area index. The null hypothesis of no structure in the assembly of birds in the landscape was tested with null models from the co-occurrence indexes. The effects of landscape structure on the assembly of the structure were analyzed by the Mantel test and principal component analysis (PCA). The assembly of the structure in the landscape showed a pattern of spatiotemporal organization significantly different from that expected by chance, revealing a structure most influenced by segregation of the species. The fragments showed significant differences in richness, unlike sets of fragments, indicating relative homogeneity in the landscape structure. The differences between the size and the distance between the fragments significantly influenced the patterns of organization of the meeting of forest birds in the landscape and patterns of α and β diversity, indicating that the higher the fragment and smaller distances between them, more the standard of species cooccurrence is different than expected by chance. Thus, the fragmented landscape of remnants of the northern Paraná Atlantic Forest still has availability of environmental resources and physical characteristics that allow a persistent organizational structure of the assembly of forest birds in space over time.

Comparação do controle do inversor trifásico conectado à rede com filtro LCL considerando o amortecimento passivo e ativo

Distributed generation systems must fulfill standards specifications of current harmonics injected to the grid. In order to satisfy these grid requirements, passive filters are connected between inverter and grid. This work compares the characteristic response of the traditional inductive (L) filter with the inductive-capacitive-inductive (LCL) filter. It is shown that increasing the inductance L leads to a good ripple current suppression around the inverter switching frequency. The LCL filter provides better harmonic attenuation and reduces the filter size. The main drawback is the LCL filter impedance, which is characterized by a typical resonance peak, which must be damped to avoid instability. Passive or active techniques can be used to damp the LCL resonance. To address this issue, this dissertation presents a comparison of current control for PV grid-tied inverters with L filter and LCL filter and also discuss the use of active and passive damping for different regions of resonance frequency. From the mathematical models, a design methodology of the controllers was developed and the dynamic behavior of the system operating in closed loop was investigated. To validate the studies developed during this work, experimental results are presented using a three-phase 5kW experimental platform. The main components and their functions are discussed in this work. Experimental results are given to support the theoretical analysis and to illustrate the performance of grid-connected PV inverter system. It is shown that the resonant frequency of the system, and sampling frequency can be associated in order to calculate a critical frequency, below which is essential to perform the damping of the LCL filter. Also, the experimental results show that the active buffer per virtual resistor, although with a simple development, is effective to damp the resonance of the LCL filter and allow the system to operate stable within predetermined parameters.