For color image fusion, existing traditional algorithms regard each color image pixel as a scalar which may lose the relationships among spectral channels. In this paper, a novel multi-focus image fusion algorithm using geometric algebra based Discrete Fourier Transform (GA-DFT) is proposed. First, a novel Discrete Fourier Transform based on geometric algebra (GA) is provided. Second, the proposed algorithm represents the color image as a GA multivector by mapping the RGB channels into the GA space. Third, the variance is selected as a contrast measure to fuse the source images in GA space. The experimental results demonstrate that the proposed algorithm can achieve the comparable performance with its counterparts in terms of subjective quality evaluation, and outperform most of the traditional algorithms in terms of objective quality metrics, indicating good performance in image fusion applications.
Direction-of-arrival (DOA) estimation plays an important role in array signal processing, and the Estimating Signal Parameter via Rotational Invariance Techniques (ESPRIT) algorithm is one of the typical super resolution algorithms for direction finding in an electromagnetic vector-sensor (EMVS) array; however, existing ESPRIT algorithms treat the output of the EMVS array either as a “long vector”, which will inevitably lead to loss of the orthogonality of the signal components, or a quaternion matrix, which may result in some missing information. In this paper, we propose a novel ESPRIT algorithm based on Geometric Algebra (GA-ESPRIT) to estimate 2D-DOA with double parallel uniform linear arrays. The algorithm combines GA with the principle of ESPRIT, which models the multi-dimensional signals in a holistic way, and then the direction angles can be calculated by different GA matrix operations to keep the correlations among multiple components of the EMVS. Experimental results demonstrate that the proposed GA-ESPRIT algorithm is robust to model errors and achieves less time complexity and smaller memory requirements.
An extended model of multilayer perceptron (MLP) based on reduced geometric algebra (RGA), namely RGA-MLP, is proposed for multi-dimensional signal processing. The RGA-MLP model treats multi-dimensional signals as multivectors in RGA space and all neuronal parameters such as inputs, connection weights, activation function and outputs, and also operators are encoded by RGA. The RGA-based back propagation (BP) algorithm is also provided. Thanks to the commutative property of RGA, multi-dimensional signals can be processed in a holistic manner which avoids losing relationship of multiple dimensions. The experiments demonstrate that the RGA-MLP model outperforms the traditional real-valued MLP model and quaternion based MLP model (QMLP) with faster convergence rate, higher classification accuracy and Lower computational complexity.
The feature extraction for multispectral images plays an important role in many computer vision applications. Recently, geometric algebra (GA) based scale invariant feature transform algorithm (GA-SIFT) and GA based speeded up robust Features algorithm (GA-SURF), have been proposed to extract feature of multispectral image in GA space. However those methods are difficult to be implemented in real-time applications. Now, the challenge is to design a new algorithm to extract the features of multispectral image more efficiently and quickly, so that it can be used in real-time applications. Although the proposed fast retina keypoint (FREAK) algorithm is faster to compute and more robust than SIFT and SURF, it can not be utilized to extract features directly for multispectral images. In this paper, we propose a novel fast retina keypoint extraction algorithm based on GA, named as GA-FREAK, for multispectral images. Firstly, the multispectral images are represented as multivectors in GA space, then the interest points are detected with the procedure of FREAK in GA space. Finally, our experiments demonstrate that the GA-FREAK is faster and more robust than some previous algorithms in multispectral images. It is expected that the proposed GA-FREAK will be a competitive alternative in real-time applications of multispectral images.
Electric load forecasting is a prominent topic in energy research. Support vector regression (SVR) has extensively and successfully achieved good performance in electric load forecasting. Clifford support vector regression (CSVR) realizes multiple outputs by the Clifford geometric algebra which can be used in multistep forecasting of electric load. However, the effect of input is different from the forecasting value. Since the load forecasting value affects the energy reserve and distribution in the energy system, the accuracy is important in electric load forecasting. In this study, a fuzzy support vector machine is proposed based on geometric algebra named Clifford fuzzy support vector machine for regression (CFSVR). Through fuzzy membership, different input points have different contributions to deciding the optimal regression hyperplane. We evaluate the performance of the proposed CFSVR in fitting tasks on numerical simulation, UCI data set and signal data set, and forecasting tasks on electric load data set and NN3 data set. The result of the experiment indicates that Clifford fuzzy support vector machine for regression has better performance than CSVR and SVR of other algorithms which can improve the accuracy of electric load forecasting and achieve multistep forecasting.
3 June 2022 in International Journal of Remote Sensing
Semantic segmentation is a commonly used intelligent interpretation method. However, semantic segmentation via deep learning usually faces several problems: resolution loss due to downsampling, difficulty in obtaining context information, and imprecise object boundaries. These problems lead to a poor performance when using satellite images instead of conventional images and limit the value of intelligent interpretation of satellite images in practical application. In this study, we construct a Qilian Ecological Resource Extraction Dataset (QERED) with multiple scene satellite images of Qilian after preprocessing via correction, fusion, mosaic, and annotation. We then propose a segmentation network, namely, High-resolution Resource Extracting Network (HrreNet), by using high-resolution feature representation, multi-scale context fusion, boundary refinement with relearning, and structural similarity loss. Experiments on QERED show that HrreNet greatly improves performance on small-size objects and slightly improves performance on larger-size objects. HrreNet achieves the best result of 73.36% mIoU on a moderate-resolution satellite dataset CCFD and 80.87% mIoU on a high-resolution satellite dataset Vaihingen.