In this paper, we address the problem of person reidentification (re-id), which remains to be challenging due to view point changes, pose variations, different camera settings, etc. Different from common methods that ...
详细信息
In this paper, a multi-scale bias field estimation is proposed to carry out the aero-thermal radiation correction. The bias field is estimated at scales from coarse to fine by an alternative minimization, after which,...
详细信息
Spiking neural P systems with astrocytes (SNPA, for short) are a class of distributed parallel computing devices inspired from the way spikes pass through the synapses between the neurons. In the present work, we disc...
详细信息
Electroencephalography (EEG) is widely used in the field of neural engineering. EEG signals can describe the brain activities while the subjects with para/tetraplegia perform movement with their limbs. This paper revi...
详细信息
In this paper, a modified method for landslide prediction is presented. This method is based on the back propagation neural network(BPNN), and we use the combination of genetic algorithm and simulated annealing algori...
详细信息
The license plate location technique is an important imageprocessing step in license plate recognition system. Vehicle license plates are distinguished from backgrounds using features proposed in existing literatures...
详细信息
The license plate location technique is an important imageprocessing step in license plate recognition system. Vehicle license plates are distinguished from backgrounds using features proposed in existing literatures. However, the effect of location is quite affected by feature selection. In this paper, we propose a method of precise license plate location fusing salient features. The method is mainly divided into three steps. First, candidate license plate regions are detected using improved Harris corner feature with much less time than traditional method. Then, candidates are sifted to only retain license plates based on two salient features named color combination and mean difference which are first proposed in this paper. Finally, the license plates are located precisely according to the projection feature. In experiment, the proposed algorithm was tested with 1942 real images captured in different environment and the license plates are successfully located as 97.6% in average with only 109ms. The experiment results demonstrates the effectiveness and efficient of our algorithm.
This paper proposes a dependency-enhanced pre-reordering method for Chinese-English statistical machine translation(SMT).Firstly,two kinds of dependency structure-based rules are extracted based on the source-side dep...
详细信息
ISBN:
(纸本)9781509009107
This paper proposes a dependency-enhanced pre-reordering method for Chinese-English statistical machine translation(SMT).Firstly,two kinds of dependency structure-based rules are extracted based on the source-side dependency tree and corresponding word alignments between the source-side and the target-side *** a maximum entropy classifier is used to calculate the orientation probability in terms of swap or monotone between two *** a result,a reordering rule set is *** different ways are proposed to filter out the rule ***,the dependency parsing trees of the training data,development set and the test set are traversed,and if the syntactic sub-tree structure matches the rules in the rule set,the word orders will be ***,a reordered source-side sentence is generated and then fed into an SMT system for *** conducted on NIST Chinese-English MT data sets show that the proposed method significantly improves translation performance by 0.46 BLEU compared to the baseline system.
Nonlinear system identification is one of the most important topics all over the word. Until now, there are many of off-line identification methods which exhibit well performance. The online approach with non-Gaussian...
详细信息
ISBN:
(纸本)9781467374439
Nonlinear system identification is one of the most important topics all over the word. Until now, there are many of off-line identification methods which exhibit well performance. The online approach with non-Gaussian noise, however, is still a challenge. For a class of nonlinear systems where all of the candidate parameters are contained in a definite parameter set, an online parameters and sates estimation method is proposed based on particle filter and Bayes theorem as the following steps. Firstly, regarding all of the candidates, the states are estimated by particle filter(PF) algorithm. Secondly the posterior probabilities of all of candidates are calculated according to the Bayes theorem;then the weights of all of the candidates are obtained through normalization. Lastly, the parameters and sates are estimated ultimately according to the weighted sum of all of the candidates and states. Numerical illustrations are presented to exhibit the application of the method proposed herein, and the performance of the method is examined.
The protocol stack plays a critical role in determining the performance of networked control system (NCS), which governs the communication activities and directly affects the communication quality of service (QoS). Fu...
详细信息
To achieve better performance with various load and system parameters in controlling a current-source rectifier (CSR) with less computing cost, a neural-network-based implementation of three-logic space-vector modulat...
详细信息
暂无评论