Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-04-01 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at that complex topological structure and various feature details of tree point cloud, this paper proposes an algorithm for extract the curve skeleton based on point cloud contraction. First, In order to directly apply the mesh shrinkage algorithm on the surface of the point cloud, performing local cloud principal component analysis and Delaunay triangulation on the point cloud. Secondly, for the problem that the tree point cloud’s complex topology and the details of the last branch, the curvature normal operator is used to shrink the point cloud. In view of the slenderness of the branches of the trees and the gentle curvature, the modified QEM mesh simplification method is used to fold the triangular mesh into a one-dimensional curve skeleton. Finally, connecting and centering the resulting curve skeleton. The algorithm in this paper operates directly on the point cloud and does not require additional information and pre-processing operations. Good robustness to noise and residual fault clouds. Experiments show that compared with other classical algorithms such as L1 and rosa, the tree point cloud skeleton extracted by the algorithm has a good topological structure, which fully expresses the biological structure and characteristics of trees in the natural environment. The skeleton extraction speed of the tree point cloud is increased by more than 3 times, and the branch reconstruction degree is increased by 25%.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： In the age of big data, it is increasingly difficult to make the community structure mining of large-scale complex networks by using the traditional community discovery algorithm and the accuracy rate is low. Therefore, this research come up with L_1-ECDA, a community discovery algorithm for deep sparse self-encoder based on smooth L_1 norm. This algorithm preprocessed the adjacency matrix of the network diagram with the method based on s Jump; then it established the deep sparse self-encoder based on smooth L_1 norm and get the low dimensional characteristic matrix by training the similarity matrix of the network graph; Finally, it get the network community structure by clustering the low-dimensional feature matrix through the K-means algorithm. Experiments on simulated network and real network data set show that the algorithm of L_1-ECDA improves the accuracy of community recognition effectively. Its accuracy rate is 4% higher than the DBCS algorithm on average, and 5.4% higher than Deepwalk algorithm and CoDDA algorithm on average.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-08-13 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the problem that the existing Internet of things big data feature selection algorithm has low computational efficiency and low scalability, this paper proposed a system architecture that selects features by using improved artificial bee colony. The architecture included a four-layer system and it could efficiently aggregate the effective data and eliminate unwanted data. The entire system was based on the Hadoop platform, MapReduce, and improved ABC algorithms. The method used improved ABC algorithm to select features and it also used a parallel algorithm to support MapReduce, which could efficiently process a huge volume of data sets. It used MapReduce tool to implement the system and it used particle filter for removal of noise. Compare the proposed algorithm with similar algorithms and evaluate the efficiency, accuracy and throughput by using ten different data sets. The results show that the proposed algorithm is more scalable and efficient in selecting features.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-24 Cooperative journals: 《计算机应用研究》
Abstract： At present, semi-automatic interpreting terrace information had certain research progress through satellite remote sensing and low-altitude remote sensing imagery. However, limited by data acquisition cost, accuracy, and single interpretation methods, it was currently limited to large-scale extraction of terraced areas. Accurate extraction of terrace field and area statistics at low cost still require further study. In this study, based on the object-oriented approach, the orthophoto and terrain index of UAV with 0.5m resolution, and the fusion of the two kinds of data were used for extraction of terrace field and area statistic. The results show that with the fusion of orthophoto and terrain index it is superior to those based on a single data source.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： Traditional methods of biomedical named entity recognition（NER） require a large amount of labeled data in the target domain, but the cost of tagging data is expensive. In order to reduce the requirement of labeled data in target domain for NER, the problem of NER in biomedical texts is transformed into a hidden Markov model based on transfer learning. The data sets in the target domain for NER do not need a large amount of labeled data to learn a model for the task by transfer learning. With the help of labeled data in source data sets across a different but related domain, and use the method of data gravitation to evaluate the contribution of samples in the auxiliary data sets about learning a model for the target domain. And calculate the weights of the data from the source domain and the data from the target domain. And then construct the hidden Markov model algorithm(BioTrHMM) based on the transfer learning. The experiment results on GENIA corpus show the BioTrHMM algorithm has better performance than the traditional algorithm of hidden Markov model, only uses small amount of labeled data in target domain.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract： To make full use of the context information of the papers, combined with the construction method of graph model and query vector, this paper proposed a fusion query information personalized citation recommendation method. Built a three layer graph model through three kinds of paper information, and set different parameters on different layers to adjust the jump probability of nodes to different levels; the query vector constructed using word2vec technology can effectively use the text context information, so that similar papers are closer to the distance, and then the candidate papers are predicted and recommended. Computational analyzes performed on the Association of Computational Linguistics Anthology Network dataset showed an average increase of about 7% over recall@N and an average increase of about 11% over NDCG@N for the same query compared to the original method. Experimental results show that the proposed method can effectively improve the quality of citation recommendation and get better recommendation results.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-12 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the bottleneck of traditional extraction method of topographic factor on soil erosion in processing massive data, we propose a method based on MapReduce. This method combined the parallel computing model MapReduce with the revised universal soil loss equation(RUSLE) . The method established the flow relationship search tree by the principle of steep slope and the B+ tree, and represented the correlation of topographic data; The method used the MapReduce model to search flow path and converge grid instead of the traditional forward and backward traversal algorithm, and solved the efficiency of the flow accumulation and slope length calculation in topographic factor calculation. The experimental results show that for the extraction of topographic factor from massive digital elevation model, the method can effectively improve the extraction efficiency in the range of the allowable accuracy.
Hits
Hits
Downloads
Comment