Search

Current low-mounting-height luminaires (LMHLs) have low effective luminous flux, high surface brightness, and many stray lights, resulting in high energy consumption and light interference, and the flicker effect may threaten driver safety. Currently, the flicker effect in LMHL lighting has been relatively poorly studied. However, in prior research, a Flicker Index (FI) quantification model was developed. In this study, we have modified the FI model to incorporate visual persistence, visual masking, and driving speed of the driver, and proposed a dynamic energy ratio model for quantifying flicker effects. The notion of effective luminous flux in road lighting was introduced, revealing that the reasons for high energy consumption, severe light pollution, and strong flicker effects in regular LMHLs were low effective luminous flux and high surface brightness. We recommend a low-mounting-height lighting technique that functions in both forward and reverse directions, which can significantly enhance the effective luminous flux of lighting fixtures, reducing stray light. The lighting fixtures created were tested on a road. The findings suggest that the energy consumption of both forward and reverse lighting fixtures is only 40% of the international standard requirements, resulting in 81% more energy savings than in existing LMHLs. The surface brightness of forward and reverse lighting fixtures is only 1.18 cd/m2, which is equivalent to 0.001% of the surface brightness of existing lighting fixtures. This indicates that 99.89% less invalid light is projected into the environment, indirectly reducing light pollution. Additionally, the flicker index of these fixtures is only 2% of that of existing lighting fixtures, reducing the impact of the flicker effect on drivers. This study provides insights into quantifying and mitigating the flicker effect in LMHLs, as well as presenting novel approaches and strategies for enhancing the efficiency and reducing light pollution of streetlamps.

Although there is abundant of investigations on fine-grained image retrieval, it is still an extremely challenging task in the field of computer vision, due to the character of small diversity in inter-class but large diversity within intra-class. To handle this task, loss functions are critical to the performance of a deep convolutional neural network in extracting the discriminative feature of the fine-grained image for retrieval. Recent studies showed that the global structure loss functions help to extract more discriminative features. In this paper, we introduce a novel global structure loss function, named Hard Decorrelated Centralized Loss, for further improving the representation for fine-grained image retrieval. The proposed loss is available in extracting the discriminative feature for dividing the most similar categories. In our experiments, we employ the proposed loss to train the convolutional neural network, which shows state-of-the-art performances on six classical fine-grained image retrieval benchmarks, e.g. CUB-200-2011 and Stanford Cars.

Similar to fingerprints and irises, pore-scale facial features can be used to distinguish human identities effectively. However, without pore-to-pore correspondences dataset, there are no deep learning based methods for pore-scale facial features. Actually, it is hard to establish a large pore-to-pore correspondences dataset due to the existing high-resolution face databases are uncalibrated and nonsynchronous. In this paper, we employ a constraint based on 3D facial model and construct a large pore-to-pore correspondences dataset. This dataset is then used to train a Convolutional Neural Network (CNN) to generate the novel pore-scale facial features - Deeply Learned Pore-scale Facial Features (DLPFF). The experiments show that our learning based method achieves the state-of-the-art matching performance on the Bosphorus facial database and has good generalization.

Generating a distinguishable feature for local patch is a main task in computer vision which aims at matching local patches. Recently, local patch descriptors from deep convolutional neural network (CNN) with a triplet loss have achieved promising performance. In this paper, we design a quadruplet loss, which can achieve a better result than other pairwise loss and triplet loss methods. Our loss is inspired by the thoughts of uniform distribution. It separates non-matching examples by using the hard sampled non-matching pairs in a batch, and simultaneously uses the random sampled non-matching examples to keep non-matching pairs to obey uniform distribution. A compact descriptor named QuadrupletNet is generated by combining the proposed quadruplet loss and L2Net CNN architecture. From our experiment, QuadrupletNet shows better performance on the Brown dataset and Hpatches dataset than Triplet loss methods on the same training set. The pre-trained QuadrupletNet is publicly available.

This paper studies the cluster output regulation problem of heterogeneous linear multi-agent systems under fixed directed topology, where dynamics of all agents are allowed to be different. Compare...

Descriptors with learned features, other than descriptors with handcrafted features, have been popular in the research community. In this paper, we propose a novel piecewise loss function to improve the performance of learned descriptor, which is inspired by the nearest neighbor distance ratio matching strategy (the matching criterion for SIFT). In our strategy, if the distance ratio between positive and closest negative example is larger than a threshold, the loss will be chosen as HardNet loss; If the ratio is smaller than the threshold, the loss will be confusion. Applying the proposed loss function to the L2Net and HardNet architecture, the proposed descriptor is named ConfusionNet. Specifically, we find that adding the confusion to train local descriptors can achieve better performance. Comparing with previous works, we show in the experiments that ConfusionNet achieves state-of-the-art results in patch retrieval, patch verification, and image matching.

This paper proposes a new home energy management system (EMS) based on the adaptive dynamic programming (ADP) algorithm for economic dispatch problem of microgrid with distributed generations. The electricity price, maximum power output of renewable generation, and load are considered as uncontrollable state variables in this paper. Then two neural networks are used to implement the ADP to minimize a given utility function over a certain time horizon. In this work, the fluctuating amounts of available renewable energy are trained and evaluated in this ADP approach. EMS reduces the daily household energy costs, maximizing the PV self-consumption and increase customer benefits through economically managing the operation of the battery. Then demonstrated via computer simulation that the EMS can reduce the overall energy cost.

Facial expression is the main research content of acquiring human psychology and developing human-computer interaction. The expression is made up of deformation and movement of each facial muscle, especially the un-posed micro expression is difficult to study because of the small degree of deformation. How to find an intuitive and interpretable method to visualize the deformation of facial expressions is an important and neglected task. In this paper, we present a new method of facial expression deformation visualization based on pore-scale facial feature matching. We use PSIFT (Pore Scale-Invariant Feature Transform) to extract a large number of facial pore-scale keypoints. Then GMS (Grid-based Motion Statistics) and VFC (Vector Field Consensus) are used to get robust matching pairs of facial feature points, so that to obtain multi-frame pore-scale keypoints tracking trajectories with time. Based on Munsell Color System, we transform matching groups into displacement vector field, Gaussian filter kernel and grid-based principal direction vector are computed to visualize the amplitude and direction of facial muscle movement. Compared with other methods of displacement field visualization (such as Horn-Schunck Optical Flow, SIFT Flow and Line Integral Convolution), our method achieves the best visualization effect in the field of facial expressions deformation on BP4D database and the collected high resolution facial expression video, the research on this subject conducive to the further development of others, such as facial expression recognition, expression modeling, expression generation and so on.

Regular inspection is important for ensuring safe operation of the power lines. Point cloud segmentation is an efficient way to carry out these inspections. Most of the existing methods depend on priori knowledge from a paticular power line corridor, which is not applicable for other unknown power line corridors. To address this problem, we propose the first end-to-end deep learning based framework for power line corridor point cloud segmentation. Specifically, we design an effective channel presentation for Light Detection and Ranging (LiDAR) point clouds and adapt a general convolutional neural network as our basic network. To evaluate the effectiveness and efficiency of our method, we collect and label a dataset, which covers a 720,000 square meter area of power line corridors. To verify the generalization ability of our method, we also test it on KITTI dataset. Experiments shows that our method not only achieves high accuracy with fast runtime on power line corridor dataset, but also performs well on KITTI dataset.

Burn-in is an effective and widely used means to improve product reliability by eliminating weak units before they are distributed in the market. Traditional burn-in that distinguishes weak units by failure during testing is inefficient and incompetent for degradation-failed products in which weak units degrade faster than normal individuals. Hence, the manufacturers have to turn to the degradation-based method. The mean lifetime to failure (MTTF) of a burnt-in population is diminished because of this type of burn-in increases the degradation level of all tested units. Ignoring the impact of burn-in leads to inferior test decisions. This study develops a multi-objective burn-in method that can simultaneously minimize the burn-in cost and maximize the burnt-in population's MTTF. We employ the time-transformed Wiener process with random effects to model the nonlinear degradation path of products and develop a burn-in scheme with two decision variables, namely, test duration and screening cutoff level. Cost expression and lifetime-based optimal objective are analytically developed. The optimal test policy is determined using the multi-objective evolutionary algorithm based on decomposition. A simulation study is conducted to demonstrate the usage and effectiveness of the multi-objective burn-in method. Published version

This paper proposes a new less rare earth permanent magnet motor (LREPMM) with hybrid ferrite PMs and rare earth PMs. The special superiority of the motor is the utilization of low priced nonrare earth PMs and the artfully combination of two kinds of PMs. In accordance with the concept of the reduction of the rare earth permanent magnet materials, the design and optimization of the motor are investigated based on the finite-element method (FEM) analysis. Moreover, in order to verify the validity of the motor, a conventional rare earth permanent magnet motor (REPMM) with identical stator is introduced, and the compares between the two motors are investigated. The results not only shows the low cost characteristics of the motor, but also shows that the LREPMM features the competitive torque performances, including preferable output torque, cogging torque, and good anti-demagnetization capacity.

Face recognition is a practical and important problem. In our life, face recognition problems exist everywhere. How to identify the face in the image accurately and quickly is an important and urgent topic. In this paper, an efficient and accurate face recognition method based on keypoints feature matching is proposed to extract the keypoints features of the face image, and then the nearest neighbor method is used to obtain a large number of keypoints rough matching. Then the correct matching of the keypoints is determined by the motion smoothness constraint of the original keypoints matching. Finally, the matching of the face is determined according to the number of keypoints matching between the two input face images. Some experiments show that this method is suitable for face recognition with expression change and pose change.

Pore-scale facial features, similar to fingerprints and irises, are effective to distinguish human identities. Nonetheless, there is a few of pore-scale facial feature database, which constrains deep learning methods to be employed of generating pore-scale facial features. In this paper, we propose a novel method by merging motion smoothness constraint and 3D model constraint, to generate a large and complex pore-scale facial feature database. The proposed method uses a powerful motion smoothness constraint in feature matching, rather than the standard ratio-test, and utilizes 3D model constraint to eliminate the incorrect matching. In the experiment, by using the proposed method, the matched numbers into high matched quality are two times higher than state-of-the-art with the same local features.

This paper performs the design and analysis of a new asymmetrical-rotor hybrid permanent magnet(PM) motor with special topology for improving torque performance base on the finite-element method(FEM). Different from traditional motors, the asymmetrical rotor structures is employed to allow the reluctance torque and the magnetic torque reach the maximum at the same current phase angle. Namely, the utilization of all torque components are significantly improved. In order to have a clear view of contribution, the frozen permeability method is applied to divide the torque into the reluctance torque and the magnetic torque. To clarify the merits of the proposed asymmetrical structure, a conventional motor with symmetrical rotor is presented for comparison under the same conditions. The results suggest the proposed design is effective in boosting torque performance. Besides, the anti-demagnetization analysis demonstrates that the topology of the proposed AR-HPM motor exhibits the advantage of avoiding the risk of demagnetization.

Similar to irises and fingerprints, pore-scale facial features are effective features for distinguishing human identities. Recently, the local feature extraction based on deep network architecture has been proposed, which needs a large dataset for training. However, there are no large databases for pore-scale facial features. Actually, it is hard to set up a large pore-scale facial-feature dataset, because the images from existing high-resolution face databases are uncalibrated and nonsynchronous, and human faces are nonrigid. To solve this problem, we propose a method to establish a large pore-to-pore correspondence dataset. We adopt Pore Scale-Invariant Feature Transform (PSIFT) to extract pore-scale facial features from face images, and use 3D Dense Face Alignment (3DDFA) to obtain a fitted 3D morphable model, which is constrained by matching keypoints. From our experiments, a large pore-to-pore correspondence dataset, including 17,136 classes of matched pore-keypoint pairs, is established.