ISSN 0253-2778

CN 34-1054/N

2017 Vol. 47, No. 7

Display Method:
Original Paper
Ocean remote sensing image auto-annotation based on DBNMI model
HUANG Dongmei, XU Qiongqiong, DU Yanling, HE Qi
2017, 47(7): 541-546. doi: 10.3969/j.issn.0253-2778.2017.07.001
Bridge the semantic gap between low-level visual feature and high-level semantic concepts has been the subject of intensive investigation on large scale remote sensing image management for years in order to improve the accuracy of automatic image annotation. An ocean remote sensing image auto-annotation method based on DBNMI model was proposed for contributions of semantic similarity about different regions of ocean remote sensing images. Initial remote sensing images were adaptively segmented, ocean remote sensing images were divided into background and the object region by means of a coarse-grained method, the relationship between low-level visual feature and high-level semantics label of the object region was modeled automatically, using DBN model, and the co-occurrence relations and adversarial relations between semantic concepts for improving image annotation results were calculated. The proposed approach is evaluated on a public remote sensing image dataset. The experimental results show a satisfactory improvement on accuracy.
Tag-based personalized travel recommendation
LI Yamei, WANG Changdong
2017, 47(7): 547-555. doi: 10.3969/j.issn.0253-2778.2017.07.002
The disparity between the huge number of tourist attraction and the limited number of trigs made by tourists has resulted in the sparseness of tourist travel data, which seriously affects the accuracy of the recommendation results. To solve this problem, four kinds of tags area, time, topic, type were extracted from a mass of travel notes to enrich the data. On the one hand, travel attractions were recommended to users by tag-content-based recommendation algorithm. On the other hand, user interest features were described with attractions feature tags. Then, similar users were found according to the interest tags of users and attractions were recommended through collaborative filtering. The tag-based collaborative filtering algorithm by 63.7% compared with the collaborative filtering recommendation algorithm and by 22.5% compared with the attraction-heat-based recommendation algorithm. Tag-content-based recommendation algorithm can improve the accuracy by 27.6% compared with the attraction-heat-based recommendation algorithm. The two algorithms were further combined with linear weight so that the two algorithms complement each other, resulting in better recommendation results. Our tag-based hybrid algorithm can make a significant improvement, i.e. increasing the accuracy by 61.3% over the tag-based collaborative filtering algorithm and 54.7% over the tag-content-based recommendation algorithm. The improvement of recommendation accuracy will enhance the user experience and make online travel websites more competitive.
Visualization of multi-dimensional sparse spatial-temporal data
2017, 47(7): 556-568. doi: 10.3969/j.issn.0253-2778.2017.07.003
Multi-dimensionality and sparseness of spatial-temporal data are major challenges for data analysis. Data visualization can effectively address certain data analysis challenges and has increasingly drawn attention from both industry and academia. A hybrid approach for the visualization of multi-dimensional sparse spatial-temporal data was proposed. The method combined multiple data view models and human-machine interaction mechanisms in order to intuitively express the multi-dimensional features, statistical group features, as well as typical individual behavior patterns. Furthermore, a visual analysis method was introduced for the identification and detection of abnormal individual behaviors. A data visualization system based on gas filling data gathered from gas stations in Xinjiang Province was implemented. By using different view models (parallel coordinates, map view, calendar matrix, Sankey
Mobile user propagation capability evaluation and coverage optimization algorithm
XU Wanru, YANG Panlong
2017, 47(7): 569-574. doi: 10.3969/j.issn.0253-2778.2017.07.004
The distribution efficiency of mobile advertising is extremely important for both advertisers and users. Few studies have been conducted on efficient ad delivery, especially user tracing and the budget. In order to obtain a feasible and effective mobile advertising distribution policy, the concept of location-centric mobile crowd sourcing network was presented to replace the traditional user-centric networks and platforms, in which the location information for advertizing distribution plays a crucial role. Therefore, the user selection under the interested area coverage(interested area coverage, IAC) region was mainly focused upon. However, research centering on location information we need requires the consideration of the temporal characteristics of each user, and effective calculation of the ICA. The problem will be more difficult to solve when considering the budget constraint. To address these challenges, considering the location sensitive mobile advertising applications, and a user selection solution was proposed, which was proved to be an NP-hard budget-constrained problem. Then, the submodularity problem was explored and a simple and effective heuristic was presented whose approximate ratio is(1-
Adaptive ensemble classification algorithm for data streams based on information entropy
SUN Yange, WANG Zhihai, YUAN Jidong, BAI Yang
2017, 47(7): 575-582. doi: 10.3969/j.issn.0253-2778.2017.07.005
The processing of streaming data implies new requirements concerning limited amount of memory, small processing time, and one scan of incoming instances. Most of the approaches in the literature to deal with concept drift only focus on gradual or abrupt concept drift and have not addressed the problem of recurring concepts. Motivated by this challenge, an ensemble with internal change detection was proposed to enhance performance by exploring the recurring concepts. It is done by maintaining a pool of classifiers, which dynamically adds and removes classifiers in response to the change detector. The algorithm adopts a two window change detection model, which adopts the Jensen-Shannon divergence to measure the distance of the distributions between two consecutive windows. When a change is detected, the repository of stored historical concepts is checked for reuse. The proposed algorithm has been experimentally compared with the state-of-the-art algorithms on synthetic and real datasets. The results show the suitability of the proposed algorithm for different types of drift as well as static environments.
An effective memetic algorithm for heterogeneous vehicle CARP
ZHANG Yuzhou, LIU Xiaofei, HUANG Shihua, MEI Jun
2017, 47(7): 583-593. doi: 10.3969/j.issn.0253-2778.2017.07.006
In view of wider application background,heterogeneous vehicle capacitated arc routing problem (HVCARP) and approach for it are investigated in this paper. First, the cost of any route in HVCARP is divided into two parts, i.e., variable cost and fixed cost, and penalty coefficients for vehicles keep the routes and their vehicles close. In view of the characteristic of HVCARP, we propose a local search operator, namely exchanging vehicles among same group routes (EVSGR) for the vehicle,which adjusts the vehicles for routes based on the loads and vehicles of the routes so as to minimize the cost. Then,the EVSGR operator is integrated into memetic algorithm (MA), and the resultant algorithm is used to solve HVCARP. Finally, the proposed algorithm is run on the instances modified from the instances of the benchmark data sets for CARP, and a large number of experimental results show that the EVSGR operator based memetic algorithm is effective for HVCARP.
Study on the feature velocity in ultrasonic guided wave of long bone based on the S transforms
SHAO Xingming, ZHENG Chichao, ZHAO Wei, PENG Hu
2017, 47(7): 594-601. doi: 10.3969/j.issn.0253-2778.2017.07.007
Ultrasonic guided wave can reflect the material properties of the long bone. The S transform with high-resolution features was proposed to extract the feature velocity in ultrasonic guided wave, and then to reflect the mechanical properties of long bone. Firstly, S transform was performed on the long bone guided wave. Then the maximum method was used to determine the propagation time corresponding to maximum energy extreme point in time-frequency domain. According to the propagation distance, the feature velocities of different modes in ultrasonic guided wave were calculated. The simulation results show that the feature velocities are linearly correlated to bone Poisson's ratio, and is insensitive to the thickness of cortical bones. Therefore, it can be used to evaluate the bone Poisson's ratio. In addition, the method does not rely on the prior knowledge of the guided wave so that it has a certain practical value.
Quadratic residue codes over Fp+u Fp+ v Fp+uv Fp+v2 Fp+uv2 Fp
QIAN Liqin, SHI Minjia, SOK Lin, PING Jingshui
2017, 47(7): 602-609. doi: 10.3969/j.issn.0253-2778.2017.07.008
Let R=Fp+u Fp+v Fp+uv Fp+v2 Fp+uv2 Fp, where u2=1, v3=v, and p is an odd prime. Quadratic residue codes of prime length n=q over the ring R was investigated, where q (q≠p) is an odd prime such that p is a quadratic residue modulo q. The cyclic codes of length n over R were studied, and then the quadratic residue codes over R in terms of idempotent generators were difined. Moreover, the relation between these codes and their extended codes are discussed. Finally, two specific forms of idempotent generators of quadratic residue codes over  Fp+u Fp+v Fp+uv Fp+v2 Fp+uv2 Fp were given to illustrate some results.
The Gray image of a class of constacyclic codes over the ring Fpm[u]/
DING Jian, LI Hongju
2017, 47(7): 610-620. doi: 10.3969/j.issn.0253-2778.2017.07.009
Let R(pm,k)=Fpm[u]/, where pj-1+1≤k≤pj and uk=0 for some positive prime number p and positive integer j. A new Gray map from R(pm,k) to Fpmpj was defined. It was proved that the Gray image of a linear (1+u+…+uk-1) constacyclic code of an arbitrary length N over R(pm,k) is a distance invariant linear cyclic code of length pjN over Fpm. Moreover, the generator polynomial of the Gray image of such a constacyclic code was determined, and some optimal linear cyclic codes over F3, F5 and F7 were constructed via the Gray map.
Simultaneous improvement of Jsc, Voc and FF of polymer solar cells with CuI as hole transport layer
SU Hetang, ZHAO Yuxia, DING Jian, DONG Kexiu, YU Wenjuan, HE Ye
2017, 47(7): 621-626. doi: 10.3969/j.issn.0253-2778.2017.07.010
A novel buffer layer copper iodide (CuI) was introduced into PCDTBT, PC70BM polymer solar cells (PSCs), where the CuI acts simultaneously as of hole transport layer (HTL) and electron block layer (EBL). Through depositing the CuI between the polymer and Ag and Au anodes, the hole collection ability has been increased. By optimizing the thickness of CuI and top anode, simultaneous enhancement of short-circuit current density (Jsc), open-circuit voltage (Voc), and fill factor (FF) has been achieved, leading to a dramatic increase of device efficiency, from 0.67% to 5.47% with 3 nm CuI and Au anode in comparison with the device without CuI. With thicker CuI, the efficiency decreases noticeably both for Au and Ag anodes due to the block effect. The result indicates that the CuI is an effective buffer layer for PCDTBT, PC70BM PSCs.