Haikel Salem Alhichri

Deep learning approach for active classification of electrocardiogram signals

In this paper, we propose a novel approach based on deep learning for active classification of electrocardiogram (ECG) signals. To this end, we learn a suitable feature representation from the raw ECG data in an unsupervised way using stacked denoising autoencoders (SDAEs) with sparsity constraint. After this feature learning phase, we add a softmax regression layer on the top of the resulting hidden representation layer yielding the so-called deep neural network (DNN). During the interaction phase, we allow the expert at each iteration to label the most relevant and uncertain ECG beats in the test record, which are then used for updating the DNN weights. As ranking criteria, the method relies on the DNN posterior probabilities to associate confidence measures such as entropy and Breaking-Ties (BT) to each test beat in the ECG record under analysis. In the experiments, we validate the method on the well-known MIT-BIH arrhythmia database as well as two other databases called INCART, and SVDB, respectively. Furthermore, we follow the recommendations of the Association for the Advancement of Medical Instrumentation (AAMI) for class labeling and results presentation. The results obtained show that the newly proposed approach provides significant accuracy improvements with less expert interaction and faster online retraining compared to state-of-the-art methods.

Differential Evolution Extreme Learning Machine for the Classification of Hyperspectral Images

Recently, a new machine learning approach that is termed as the extreme learning machine (ELM) has been introduced in the literature. This approach is characterized by a unified formulation for regression, binary, and multiclass classification problems, and the related solution is given in an analytical compact form. In this letter, we propose an efficient classification method for hyperspectral images based on this machine learning approach. To address the model selection issue that is associated with the ELM, we develop an automatic-solution-based differential evolution (DE). This simple yet powerful evolutionary optimization algorithm uses cross-validation accuracy as a performance indicator for determining the optimal ELM parameters. Experimental results obtained from four benchmark hyperspectral data sets confirm the attractive properties of the proposed DE-ELM method in terms of classification accuracy and computation time.

Virtual circles: a new set of features for fast image registration

In this paper, we propose a novel set of image features called virtual circles, and their use in an efficient image registration algorithm to find translation and scale differences. A virtual circle is a circle with maximal radius encompassing a background area that does not contain edge points. Virtual circles have a number of nice properties such as the fact that they can be extracted efficiently with the help of the distance transform from many types of images. They can also have extra information, such as their radii, which can be used for efficient registration. Furthermore, virtual circles are robust against broken or corrupt edges. On the other hand, they are vulnerable to background noise, but through the use of a heuristic called the smoothness criterion, virtual circles that are less likely to be corrupted by background noise can be selected. Another advantage of the smoothness criterion is that it reduces the number of virtual circles needed, which increases the efficiency of the algorithm.

Fusion of Extreme Learning Machine and Graph-Based Optimization Methods for Active Classification of Remote Sensing Images

In this letter, we propose an efficient multiclass active learning (AL) method for remote sensing image classification. We fuse the capabilities of an extreme learning machine (ELM) classifier and graph-based optimization methods to boost the classification accuracy while minimizing the user interaction. First, we use the ELM to generate an initial label estimation of the unlabeled image pixels. Then, we optimize a graph-based functional energy that integrates the ELM outputs as an initial estimation of the image structure. As for the ELM, the solution to this multiclass optimization problem leads to a system of linear equations. Due to the sparse Laplacian matrix built from the lattice graph defined on the image pixels, the optimization problem is solved in a linear time. In the experiments, we report and discuss the results of the proposed AL method on two very high resolution images acquired by IKONOS-2 and GoeEye-1, as well as the well-known Pavia University hyperspectral image.

Efficient Framework for Palm Tree Detection in UAV Images

The latest developments in unmanned aerial vehicles (UAVs) and associated sensing systems make these platforms increasingly attractive to the remote sensing community. The large amount of spatial details contained in these images opens the door for advanced monitoring applications. In this paper, we use this cost-effective and attractive technology for the automatic detection of palm trees. Given a UAV image acquired over a palm farm, first we extract a set of keypoints using the Scale-invariant Feature Transform (SIFT). Then, we analyze these keypoints with an extreme learning machine (ELM) classifier a priori trained on a set of palm and no-palm keypoints. As output, the ELM classifier will mark each detected palm tree by several keypoints. Then, in order to capture the shape of each tree, we propose to merge these keypoints with an active contour method based on level sets (LSs). Finally, we further analyze the texture of the regions obtained by LS with local binary patterns (LBPs) to distinguish palm trees from other vegetations. Experimental results obtained on UAV images with 3.5 cm of spatial resolution and acquired over two different farms confirm the promising capabilities of the proposed framework.

Using convolutional features and a sparse autoencoder for land-use scene classification

ABSTRACT In this article, we propose a novel approach based on convolutional features and sparse autoencoder (AE) for scene-level land-use (LU) classification. This approach starts by generating an initial feature representation of the scenes under analysis from a deep convolutional neural network (CNN) pre-learned on a large amount of labelled data from an auxiliary domain. Then these convolutional features are fed as input to a sparse AE for learning a new suitable representation in an unsupervised manner. After this pre-training phase, we propose two different scenarios for building the classification system. In the first scenario, we add a softmax layer on the top of the AE encoding layer and then fine-tune the resulting network in a supervised manner using the target training images available at hand. Then we classify the test images based on the posterior probabilities provided by the softmax layer. In the second scenario, we view the classification problem from a reconstruction perspective. To this end we train several class-specific AEs (i.e. one AE per class) and then classify the test images based on the reconstruction error. Experimental results conducted on the University of California (UC) Merced and Banja-Luka LU public data sets confirm the superiority of the proposed approach compared to state-of-the-art methods.

Multi-resolution image registration using multi-class Hausdorff fraction

Recently, a new image registration method, based on the Hausdorff fraction and a multi-resolution search of the transformation space, has been developed in the literature. This method has been applied to problems involving translations, translation and scale, and affine transformations. In this paper, we adapt the above method to the set of similarity transformations. We also introduce a new variant of the Hausdorff fraction similarity measure based on a multiclass approach, which we call the multi-class Hausdorff fraction (MCHF). The multi-class approach is more efficient because it matches feature points only if they are from the same class. To validate our approach, we segment edge maps into two classes which are the class of straight lines and the class of curves, and we apply the new multi-class approach to two image registration examples, using synthetic and real images, respectively. Experimental results show that the multiclass approach speeds up the multi-resolution search algorithm.

Deep Learning Approach for Car Detection in UAV Imagery

This paper presents an automatic solution to the problem of detecting and counting cars in unmanned aerial vehicle (UAV) images. This is a challenging task given the very high spatial resolution of UAV images (on the order of a few centimetres) and the extremely high level of detail, which require suitable automatic analysis methods. Our proposed method begins by segmenting the input image into small homogeneous regions, which can be used as candidate locations for car detection. Next, a window is extracted around each region, and deep learning is used to mine highly descriptive features from these windows. We use a deep convolutional neural network (CNN) system that is already pre-trained on huge auxiliary data as a feature extraction tool, combined with a linear support vector machine (SVM) classifier to classify regions into “car” and “no-car” classes. The final step is devoted to a fine-tuning procedure which performs morphological dilation to smooth the detected regions and fill any holes. In addition, small isolated regions are analysed further using a few sliding rectangular windows to locate cars more accurately and remove false positives. To evaluate our method, experiments were conducted on a challenging set of real UAV images acquired over an urban area. The experimental results have proven that the proposed method outperforms the state-of-the-art methods, both in terms of accuracy and computational time.

Robust Estimation of Water Chlorophyll Concentrations With Gaussian Process Regression and IOWA Aggregation Operators

In this paper, we propose a new framework for estimating water chlorophyll concentrations in remote sensing data based on Gaussian process regression (GPR) and induced ordered weighted averaging(IOWA) operators. First, we construct an ensemble of GPR estimators modeled with different covariance functions. Then, in a second step, we aggregate the predictions of these estimators using IOWA operators. To learn the weights associated with these nonlinear operators, we propose three different approaches called IOWAMVO, IOWAMOP, and IOWAPA. The IOWAMVO is based on the minimization of the variance of the weights with a given orness level. In IOWAMOP, we replace the orness level constraint by an objective related to data fitting. Then we solve the modified optimization problem using a multiobjective optimization evolutionary algorithm based on decomposition. Finally, in IOWAPA, we generate the weights directly from the confidence measures (i.e., output variances) provided by the set of GPR estimators using the concept of prioritization aggregation. Experimental results on in situ and satellite data are reported and discussed.

Using OWA Fusion Operators for the Classification of Hyperspectral Images

In this paper, we propose a novel ensemble-based classification system for improving the classification accuracy of hyperspectral images. To generate the ensemble, we run the mean-shift (MS) algorithm several times on different bands randomly selected from the hyperspectral cube and with distinct kernel width parameters. The resulting set of MS maps are then successively labeled via a pair wise labeling procedure with respect to a spectral-based classification map generated by the support vector machine (SVM) classifier. To this end, for each region in the MS maps, the weighted-majority-voting (WMV) rule is applied to the corresponding pixels in the SVM map. The output of this step is a set of spectral-spatial classification maps termed as SVM-MS maps. In order to generate the final classification result, we propose to aggregate this set of SVM-MS maps using the ordered weighted averaging (OWA) operator. The determination of the associated weights is made using the idea of a stress function. The performance of the proposed classification system is assessed on three different hyperspectral datasets acquired by the Reflective Optics System Imaging Spectrometer (ROSIS-03), the Digital Imagery Collection Experiment (HYDICE) and the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) sensors.