Search

Person re-identification is still a useful and challenging task in pattern recognition and computer vision. Based on a probe image, the task focuses on identifying a set of matching images from a gallery set. In this work, a novel multi-feature fusion-based person re-identification framework is proposed. The key technique is to intelligently fuze hand-crafted feature and deeply-learned feature. More specifically, hand-crafted features from both local and global region are extracted from each image. Afterward, a novel Convolutional Neural Networks (CNN) is trained based on combining the three datasets, wherein robust deep features can be obtained. Finally, the above hand-crafted feature as well as the deep features are optimally fuzed for calculating a robust re-identification result. Extensive experimental results on three state-of-the-art data sets have demonstrated the effectiveness of our method.

Facial expression recognition (FER) plays an important role in the applications of human computer interaction. Given the wide use of convolutional neural networks (CNNs) in automatic video and image classification systems, higher-level features can be automatically learned from hierarchical neural networks with big data. However, learning CNNs require large amount of training data for adequate generalization, while the Scale-invariant feature transform (SIFT) does not need large training samples to generate useful feature. In this paper, we propose a new hybrid feature representation for the recognition of facial expressions from a single image frame that uses a combination of SIFT and deep-learning feature of different level extracted from the CNN model, then adopt the combined features and classify the expression by support vector machines (SVM). The performance of the proposed method has been validated on public CK+ databases. To evaluate the generalization ability of our method, we also performed an experiment on a cross-database environment. Experimental results show that the proposed approach can achieve better classification rates compared with state-of-art CNN methods, which indicate the considerable potential of combining shallow feature with deep-learning feature.

Aiming at the problem that the manual monitoring cannot guarantee the safe and stable operation of the power operation at all times, according to the principle and development status of the feature extraction and the target detection technology, the rational application of this technology in the power operation process is analyzed to ensure the reliability of the power supply. At present, the patrol work of the overhead power lines basically adopts the manual patrol mode. This mode of operation is of great labor intensity, great danger and low efficiency. Therefore, it is very necessary to extract the specific characteristics of the electric power operation according to the specific conditions of the specific electric power operation process, and systematically develop the target detection technology and explore the reasonable application strategies, which is extremely important.

Distributed energy as an important part of the national distribution power grid aspect, which in addition to its grid technologies, power quality, harmonic control, and network protection, and many other requirements, but it also the need to achieve the accurate measurement of the distributed power generation. Because of distributed power generation has rapid flow switching, dynamic load changes and other reasons; it has brought the new challenge to the current metering technology on the distributed power generation capacity. This article starts from the impact of distributed power generation on measurement technology and analyzes the current situation of the photovoltaic power generation metering and its problems and discusses the new technology of the distributed power energy measurement from the impact of distributed generation on measurement technology.

Use of path expressions is a common feature in most XML query languages, and many evaluation methods for path expression queries have been proposed recently. However, there are few researches on the issue of optimizing regular path expression queries. In this paper, two kinds of path expression optimization principles are proposed, named path shortening and path complementing, respectively. The path shortening principle reduces the querying cost by shortening the path expressions with the knowledge of XML schema. While the path complementing principle substitutes the user queries with the equivalent lower-cost path expressions. The experimental results show that these two techniques can largely improve the performance of path expression query processing.

Most query languages in XML database systems use Regular Path Expressions (RPE) to query or extract data from databases and some query processing and optimization techniques have been proposed for RPEs. Conceptually XML documents are collections of path instances. Each path instance should conform to an XML element tag sequence, called path schema. A RPE query can be written as an automaton that can represent a language, while path schemas can be seen as sentences. In this paper, a novel RPE computing approach, automaton match (AM), is proposed. AM queries the RPEs by matching the automatons with path schemas. The experimental results show AM is quite efficient for computing RPE queries.

Path query is one of the most frequently used components by the various XML query languages. Most of the proposed methods compute path queries in instance space, i.e. directly facing the XML instances, such as XML tree traversal and containment join ways. As a query method based on automata technique, automata match (AM) can evaluate path expression queries in schema space so that it allows efficient computation of complex queries on vast amount of data. This paper introduces how to construct query automata in order to compute all regular expression queries including those with wildcards. Furthermore, a data structure named schema automata is proposed to evaluate containment queries that are very difficult from the conventional automata point of view. To improve the efficiency of schema automata, methods to reduce and persistent them are proposed. Finally, performance study of the proposed methods are given.