Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-18 Cooperative journals: 《计算机应用研究》
Abstract： This paper proposed an end-to-end deep reinforcement learning framework to improve the efficiency of solving the Multi-Depot Vehicle Routing Problem (MDVRP) . There is a novel formulation of the Markov Decision Process (MDP) for the MDVRP, including the definitions of its state, action, and reward. Then, this paper exploited an improved Graph Attention Network (GAT) as the encoder to perform feature embedding on the graph representation of MDVRP, and designed a Transformer-based decoder. Meanwhile, this paper used the improved REINFORCE algorithm to train the proposed encoder-decoder model. Furthermore, the designed encoder-decoder model is not bounded by the size of the graph. That is, once the framework is trained, it can be used to solve MDVRP instances with different scales. Finally, the results on randomly generated and published standard instances verify the feasibility and effectiveness of the proposed framework. Significantly, even on solving MDVRP with 100 customer nodes, the trained model takes only two milliseconds on average to obtain a very competitive solution compared with existing methods.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the single description and the noise sensitive problems of the local binary pattern(LBP) , a multi-scale adaptive threshold local ternary pattern(MSALTP) algorithm is proposed. The algorithm first enlarges the original images . Secondly, divides the images into several regions equally and calculates the mean value of the pixels. Then calculates the deviation between the center and the mean pixel value of each region. Finally, extract the ALTP features and the resulting statistical features histograms are used to classify the images. Experiments show that the proposed algorithm recognition rates are higher than the current anti-noise algorithms under different noise.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-09-12 Cooperative journals: 《计算机应用研究》
Abstract： Reliable scene analysis, which commonly involves vehicle identification, pedestrian detection, and collision avoidance, u, is an essential technique in realizing automatic driving of intelligent vehicle. This paper proposes a stereo vision multi-object tracking method based on joint probabilistic data association (JPDA) filter for intelligent vehicle: the method collects the images and videos of vehicles and pedestrians with the help of stereo camera mounted on top of a vehicle, model the uncertainty of sensor in Lie group, and adopt the Euclidean algorithm to implement state-filtering for the preprocessed images. Utilize the improved JPDA to rectified the tracking of vehicles and pedestrians. Experiment results show that the proposed method can settle the tracking of multi-objects effectively, and improve the level of automation and intelligence for driving system dramatically. Compared with other new target tracking methods, the proposed method has obvious advantages in tracking precision and speed. It does not produce obvious offset in tracking the vehicle and will not miss the tracking of pedestrians.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： This paper researched Image matching issue and proposed a feature point matching algorithm to improve the BRIEF algorithm. The proposed algorithm generated the feature point description operator according to the difference signs and the difference magnitudes relation between the random point and the feature point. BRIEF was sensitive to noise because the small difference magnitude is more susceptible to the noise. To solve this problem, This paper determined a small pixel difference threshold by the neighborhood mean value of the BRIEF. Comparing the Hamming distance between descriptions realizes the feature points matching. Compared with BRIEF and ORB algorithm, the experiments proved that the operator has higher discriminant, simple calculation and good noise suppression performance. And the matching accuracy is higher.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-17 Cooperative journals: 《计算机应用研究》
Abstract： As extant method wrongly highlights backgrounds in salient object detection from complex background images, this paper proposed a new algorithm of saliency detection to suppress background in combination with sparse reconstruction and energy optimization, which extracted foreground more accurate instantaneously. Firstly, it decomposed the input image into superpixel within Abstract: ng unnecessary detail. Then, it selected image boundary superpixels as background templates, which used as sparse dictionaries to calculate reconstruction errors that as superpixels initial saliency. Finally, it introduced the energy equation to optimize the initial saliency, and generated the final saliency map after the foreground of optimized saliency was enhanced. It tested the proposed algorithm and other 10 algorithms on MSRA10K and ECSSD1000 dataset with ground truths. The PR Curve, Precision (P) and F-measure(F) of the proposed algorithm had better performance than other 10 algorithms. The experimental results show that the proposed algorithm is more robust to suppress background effectively, and the extraction of salient object is more precise.
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 high complexity, strong nonlinearity and strong time characteristics of intermittent process, this paper proposed a new method based on kernel entropy component analysis (KECA) to reduce the dimensionality of the KECA characteristic variables, and used the fireworks algorithm (FWA) to optimize the support vector machine (SVM) parameters for the intermittent process of division fault diagnosis method. Firstly, it carried out multi-directional kernel principal component analysis (MKPCA) for the on-line fault monitoring and output the fault data. Second, it used K-means method to divid the batch process into several sub-periods. It used KECA to reduce characteristic variable dimensionality according to the contribution rate of entropy to determine the number of selected elements and extracted feature information in depth. Finally, constructed FWA optimized SVM parameter fault diagnosis model in each sub-period, put the reduced dimension processed fault data into their own sub-period FWA-SVM diagnostic model for fault diagnosis. Through a variety of comparative experimental study based on penicillin simulation data, verified the feasibility and effectiveness of this method.
Hits
Hits
Downloads
Comment