Electrical Impedance Tomography (EIT) is an advanced visualization technique characterized by its cost-effectiveness, rapid response, non-ionizing radiation, and non-intrusiveness compared with other tomography modalities. However, the quality of EIT imaging is compromised by the inherent ill-posed solution and the soft-field effect, which hinder the effective reconstruction of target objects. Various methods have been employed to address these issues, but they can make limited progress in high-quality reconstruction. In this paper, efforts are made by applying the clustering technique to the EIT reconstruction. The key features hidden in all EIT measurements are extracted, and whereby each pixel can be represented by a feature vector. After clustering these vectors over all pixels, we proposed a novel EIT reconstruction algorithm. Experimental results show that the novel algorithm can reconstruct the EIT images more accurately and effectively than the existing algorithms. Consequently, the implementation of a clustering algorithm instead of the existing algorithm offers a practical and effective approach to enhance the EIT imaging quality.

As an advanced detection technique, electrical resistive tomography (ERT) has been applied to detect the solid–liquid two-phase flow velocity based on available ERT measurements. The flow velocity computation by ERT must depend on the relative algorithms, including both the cross-correlation (CC) principle and convolutional neural networks (CNNs). However, these two types of algorithms have poor accuracy and generalization under complex measuring conditions and various flow patterns. To address this issue, in this paper, a hybrid network is proposed that combines a CNN with a reproducing kernel-based support vector machine (RKSVM) technique. The features hidden in ERT measurements are extracted using the CNN, and then the flow velocity is computed by the RKSVM in a high-dimensional feature space. According to the ERT measurements in an actual experimental platform, the results show that the hybrid network has higher accuracy and generalization ability for flow velocity computation compared with the existing CC, RKSVM, and CNN methods.

Electric tomography (ET) is an advanced visualization technique with low-cost, rapid-response, nonradiative, and non-intrusive advantages compared with other tomography modalities. The imaging resolution of ET, however, is significantly low providing the required measurements that are far less than the number of pixels in a detection field. Presented here is a reproducing kernel-based best interpolation (RKBI) method that can greatly increase the number of numeric measurements in the ET process. For a group of available measurements, RKBI has the smallest approximation error compared to existing interpolation methods. Furthermore, the error of RKBI can be easily estimated with no additional hardware and the need for actual measurements. The optimality of RKBI is validated using both theoretical and experimental frameworks, demonstrating that RKBI really improves the spatial resolution and steadiness of ET images.

This paper indicated a potential approach for direct rice production estimation with single high frequency SAR image. A modified cloud model had been built, in which the grain-layer was considered as the top layer and the stems and leaves as the second layer. Beneath these two layers, the water surface was considered as mirror reflection, of which value was extracted from nearby water surface. The simulated annealing algorithm and the leave-one-out method had also been applied. The result suggested that the scheme had good potential to estimate the rice grain number per unit, which was the key indice of yield, from a single TerraSAR image.

With the explosive increase of big data in industry and academic fields, it is necessary to apply large-scale data processing systems to analysis Big Data. Arguably, Spark is state of the art in large-scale data computing systems nowadays, due to its good properties including generality, fault tolerance, high performance of in-memory data processing, and scalability. Spark adopts a flexible Resident Distributed Dataset (RDD) programming model with a set of provided transformation and action operators whose operating functions can be customized by users according to their applications. It is originally positioned as a fast and general data processing system. A large body of research efforts have been made to make it more efficient (faster) and general by considering various circumstances since its introduction. In this survey, we aim to have a thorough review of various kinds of optimization techniques on the generality and performance improvement of Spark. We introduce Spark programming model and computing system, discuss the pros and cons of Spark, and have an investigation and classification of various solving techniques in the literature. Moreover, we also introduce various data management and processing systems, machine learning algorithms and applications supported by Spark. Finally, we make a discussion on the open issues and challenges for large-scale in-memory data processing with Spark.

The convolutional neural network (CNN) has been used in almost all research and application fields. With ever-increasing actual requirements and the explosive growth of available big data, the topology structure of CNN has become more complex, particularly in terms of the number of layers and nodes. The complex structures decrease the generalization capability and efficiency of CNN. To solve the problems, we propose a CNN-induced network. By merely replacing the fully connected layer with n nodes using a Choquet integral (CI) block, a typical CNN can be extended to n! CNNs without increasing the number of layers and nodes. As an example, in the solid-liquid two-phase flow of dredging engineering, the proposed network is evaluated by the computation accuracy of solid-phase fraction (SPF). A large number of samples are collected due to complex flow patterns and fast-changing measurement conditions, but a typical CNN has poor generalization and learning capability on complex datasets. Inversely, the new network solves the above problem easily and practicably. Experiments show that the new network not only provides a robust approach for learning from big dataset but also provides an effective and stable method for SPF computation.