Aleksandra Pizurica

Fusion of pixel and object-based features for weed mapping using unmanned aerial vehicle imagery

Abstract The developments in the use of unmanned aerial vehicles (UAVs) and advanced imaging sensors provide new opportunities for ultra-high resolution (e.g., less than a 10u202fcm ground sampling distance (GSD)) crop field monitoring and mapping in precision agriculture applications. In this study, we developed a strategy for inter- and intra-row weed detection in early season maize fields from aerial visual imagery. More specifically, the Hough transform algorithm (HT) was applied to the orthomosaicked images for inter-row weed detection. A semi-automatic Object-Based Image Analysis (OBIA) procedure was developed with Random Forests (RF) combined with feature selection techniques to classify soil, weeds and maize. Furthermore, the two binary weed masks generated from HT and OBIA were fused for accurate binary weed image. The developed RF classifier was evaluated by 5-fold cross validation, and it obtained an overall accuracy of 0.945, and Kappa value of 0.912. Finally, the relationship of detected weeds and their ground truth densities was quantified by a fitted linear model with a coefficient of determination of 0.895 and a root mean square error of 0.026. Besides, the importance of input features was evaluated, and it was found that the ratio of vegetation length and width was the most significant feature for the classification model. Overall, our approach can yield a satisfactory weed map, and we expect that the obtained accurate and timely weed map from UAV imagery will be applicable to realize site-specific weed management (SSWM) in early season crop fields for reducing spraying non-selective herbicides and costs.

A Robust Sparse Representation Model for Hyperspectral Image Classification

Sparse representation has been extensively investigated for hyperspectral image (HSI) classification and led to substantial improvements in the performance over the traditional methods, such as support vector machine (SVM). However, the existing sparsity-based classification methods typically assume Gaussian noise, neglecting the fact that HSIs are often corrupted by different types of noise in practice. In this paper, we develop a robust classification model that admits realistic mixed noise, which includes Gaussian noise and sparse noise. We combine a model for mixed noise with a prior on the representation coefficients of input data within a unified framework, which produces three kinds of robust classification methods based on sparse representation classification (SRC), joint SRC and joint SRC on a super-pixels level. Experimental results on simulated and real data demonstrate the effectiveness of the proposed method and clear benefits from the introduced mixed-noise model.

Hyperspectral Unmixing Using Double Reweighted Sparse Regression and Total Variation

Spectral unmixing is an important technique in hyperspectral image applications. Recently, sparse regression has been widely used in hyperspectral unmixing, but its performance is limited by the high mutual coherence of spectral libraries. To address this issue, a new sparse unmixing algorithm, called double reweighted sparse unmixing and total variation (TV), is proposed in this letter. Specifically, the proposed algorithm enhances the sparsity of fractional abundances in both spectral and spatial domains through the use of double weights, where one is used to enhance the sparsity of endmembers in spectral library, and the other is introduced to improve the sparsity of fractional abundances. Moreover, a TV-based regularization is further adopted to explore the spatial–contextual information. As such, the simultaneous utilization of both double reweighted

Compressed sensing in MRI with a Markov random field prior for spatial clustering of subband coefficients

l_{1}

Image denoising algorithms : from wavelet shrinkage to nonlocal collaborative filtering

minimization and TV regularizer can significantly improve the sparse unmixing performance. Experimental results on both synthetic and real hyperspectral data sets demonstrate the effectiveness of the proposed algorithm both visually and quantitatively.

Compressed Sensing using sparse binary measurements: A rateless coding perspective

Recent work in compressed sensing of magnetic resonance images (CS-MRI) concentrates on encoding structured sparsity in acquisition or in the reconstruction stages. Subband coefficients of typical images obey a certain structure, which can be viewed in terms of fixed groups (like wavelet trees) or statistically (certain configurations are more likely than others). Approaches using wavelet tree-sparsity have already demonstrated excellent performance in MRI. However, the use of statistical models for spatial clustering of the subband coefficients has not been studied well in CS-MRI yet, although the potentials of such an approach have been indicated. In this paper, we design a practical reconstruction algorithm as a variant of the proximal splitting methods, making use of a Markov Random Field prior model for spatial clustering of subband coefficients. The results for different undersampling patterns demonstrate an improved reconstruction performance compared to both standard CS-MRI methods and methods based on wavelet tree sparsity.

Compressed sensing using sparse adaptive measurements

This paper presents an overview of image denoising algorithms ranging from wavelet shrinkage to patchbased non-local processing. The focus is on the suppression of additive Gaussian noise (white and coloured). A great attention is devoted to explaining the main underlying ideas and concepts of representative approaches with illustrative examples, accessible also to non-experts in the field. A Bayesian perspective of wavelet shrinkage is given, with different instances of spatial context modelling (including local spatial activity indicators, Markov Random Fields, Hidden Markov Tree models and Gaussian Scale Mixture models). Extensions to other transform domains (curvelets and other generalizations of wavelets) are addressed too, showing the benefits in terms of image quality. Patch-based image denoising is illustrated with principles of non-local means filtering and collaborative filtering, explaining also the connections with dictionary learning. Some general notes on the performance comparison are given, by summarizing the benefits and limitations of various approaches against each other, and pointing to some of the current trends in the field.

Multiscale Superpixel-Level Subspace-Based Support Vector Machines for Hyperspectral Image Classification

Compressed Sensing (CS) methods using sparse binary measurement matrices and iterative message-passing recovery procedures have been recently investigated due to their low computational complexity and excellent performance. Drawing much of inspiration from sparse-graph codes such as Low-Density Parity-Check (LDPC) codes, these studies use analytical tools from modern coding theory to analyze CS solutions. In this paper, we consider and systematically analyze the CS setup inspired by a class of efficient, popular and flexible sparse-graph codes called rateless codes. The proposed rateless CS setup is asymptotically analyzed using tools such as Density Evolution and EXIT charts and fine-tuned using degree distribution optimization techniques.