The problem of limited tagged training samples and unknown the number of classes is challenging for the classification of remote sensing scenes. This paper presents a new GMRF self-supervised algorithm for SAR image. ...
详细信息
The problem of limited tagged training samples and unknown the number of classes is challenging for the classification of remote sensing scenes. This paper presents a new GMRF self-supervised algorithm for SAR image. We add a GOF process in the process of estimating GMM parameters by EM algorithm, which can not only dynamically select the best number of significant classes but also provides an initial feature parameter to calculate the MRF minimum energy. After the iterative region label and region growth cycle, iteration is combined with the Mll context model to obtain the best mark of each region. Since the initial feature parameter selection of the MRF is not random, the operation efficiency is also improved while reducing the number of iteration cycles of the algorithm. The experiment validates that our design not only solves the problem of manual input of the number of classes but also provide the better output result graph in terms of detail maintenance than the expert interpretation of the truth map in the unsupervised image classification process, and we hope that it could support operation and meet the real-time requirements.
A self-supervised algorithm based on deep learning is designed to estimate the depth of the driving scene. The depth estimation network and pose estimation network designed based on convolutional neural network take t...
详细信息
A self-supervised algorithm based on deep learning is designed to estimate the depth of the driving scene. The depth estimation network and pose estimation network designed based on convolutional neural network take the video information obtained by monocular camera as the input, and output the depth map of each frame of input image and the pose changes between two adjacent frames of input images, respectively. The view synthesis, that is, the image reconstruction loss between two adjacent frame images, is used as the supervision signal to train the neural network. The problem of scale inconsistency in monocular depth estimation is solved through the scale consistency loss, and the weight mask obtained from the scale inconsistency loss is used to solve the dynamic problems and the adverse effects of occluded objects in driving environment. The tests results show that the designed self-supervised depth estimation algorithm based on monocular video information shows high accuracy on the KITTI dataset and almost reaches the same level as the supervisedalgorithm.
Spectral clustering, as an algorithm based on graph theory and spectral theory, has shown excellent performance in the classification tasks of hyperspectral images in recent years. Although better results have been ac...
详细信息
Spectral clustering, as an algorithm based on graph theory and spectral theory, has shown excellent performance in the classification tasks of hyperspectral images in recent years. Although better results have been achieved, some challenges still exist. The inclusion of a priori information can increase the performance of spectral clustering algorithms;however, in practice, it is often unable to meet the demand for a priori information;at the same time, spectral clustering for the quality of the similarity matrix is demanding, and some of the current algorithms are to improve the similarity matrix from the aspect of data planning, and ultimately for the label feature matrices are diluted. In response to the above problems, this paper proposes a self-supervised spectral clustering with spectral embedding (SESSC). The algorithm obtains a low-dimensional representation of the data through spectral embedding, which can simplify clustering while preserving feature information;then uses the similarity matrix about the data and the guidance of the sample constraint information to obtain a new similarity matrix in order to further refine the structural graph, and the result can feed back and optimize the label feature matrix and the low-dimensional representation of the data. Additionally, we introduce a fractional theory in the update of the sample variable matrix, which assures the integrity and validity of the information in the update. Experimental results show that the proposed algorithm has better performance in hyperspectral image classification than existing spectral clustering algorithms.
As a leading graph clustering technique, spectral clustering is one of the most widely used clustering methods that captures complex clusters in data. However, some of its deficiencies, such as the high computational ...
详细信息
As a leading graph clustering technique, spectral clustering is one of the most widely used clustering methods that captures complex clusters in data. However, some of its deficiencies, such as the high computational complexity in eigen decomposition and the guidance without supervised information, limit its real applications. To get rid of the deficiencies, we propose a self-supervised spectral clustering algorithm. In this algorithm, we define an exemplar constraint which reflects the relations between objects and exemplars. We provide the related analysis to show that it is more suitable for unsupervised learning. Based on the exemplar constraint, we build an optimization model for self-supervised spectral clustering so that we can simultaneously learn clustering results and exemplar constraints. Furthermore, we propose an iterative method to solve the new optimization problem. Compared to other existing versions of spectral clustering algorithms, the new algorithm can use the low computational costs to discover a high-quality cluster structure of a data set without prior information. Furthermore, we did a number of experiments of algorithm comparison and parameter analysis on benchmark data sets to illustrate that the proposed algorithm is very effective and efficient. (C) 2022 Elsevier Ltd. All rights reserved.
Purpose An automated accurate segmentation for dynamic contrast-enhanced magnetic resonance (DCE-MR) image sequences is essential for quantification of renal function. A self-supervised strategy is proposed for fully ...
详细信息
Purpose An automated accurate segmentation for dynamic contrast-enhanced magnetic resonance (DCE-MR) image sequences is essential for quantification of renal function. A self-supervised strategy is proposed for fully automatic segmentation of the renal DCE-MR images without using manually labeled data. Methods The proposed strategy employed both temporal and spatial information of the DCE-MR image sequences. First, the kidney area, the seed regions of the cortex, the medulla, and the pelvis were automatically detected in the spatial domain. Subsequently, all the pixels in the kidney were automatically labeled as the cortex, the medulla, or the pelvis based on their time-intensity signal and spatial position using a supervised classifier. The feasibility of the proposed strategy was verified on a dataset of renal DCE-MR images of 14 subjects without history of kidney diseases. Furthermore, the self-supervised strategy and the commonly used traditional unsupervised method were quantitatively compared with a reference manual segmentation by an experienced radiologist, using similarity indexes. Results The average Dice coefficient (ADC) for the segmentations of the proposed self-supervised method is 0.92 using a ransom walker model as the classifier or 0.86 using a K-nearest neighbor model as the classifier. The ADC of the Kmeans-based unsupervised methods with three and six clusters were 0.65 and 0.79, respectively. The Dice coefficients of the self-supervised method were remarkably higher than that of the unsupervised method (one-tailed paired-sample t-test, P-values <10(-3)). Conclusions The results indicate that the proposed self-supervised approach yields a satisfactory similarity with the reference manual segmentation. Compared with the traditional unsupervised clustering method, the new strategy does not require manual intervention during the segmentation process and achieves better results for the segmentation of renal DCE-MR images.
暂无评论