NWNing Wei

Ning Wei

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Ning Wei has a total of 31 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • MiT-Unet: Mixed Transformer Unet for Transmission Line Segmentation in RAV Images

    by Ning Wei, Jianwei Chen, Shuifa Sun

    The segmentation of power lines in drone images is one of the challenging tasks in the field of computer vision. Although power lines share the same difficulties with tiny object segmentation, occupying only a very small proportion of pixel areas in the images, the greater challenge is that they also have a very large visual perspective field. Therefore, the results obtained by traditional convolutional neural network-based segmentation methods are still unsatisfactory. To tackle these problems, we propose MiT-Unet (Mixed Transformer Unet), which has an analogous multi-level convolutional neural network structure similar to Unet for encoding and decoding transmission line detailed features. Meanwhile, dealing with excessive scales, we employ an Efficient Self-Attention-based module to enhance the global features of straight lines. Furthermore, in the level transition of this architecture, we propose simple yet efficient upsampling and downsampling modules, TLFM and TLFE, to fuse the global attention feature of power lines. These modules effectively extract and preserve the vulnerable features of power lines during the level transition process. Experimental results demonstrate that our proposed method achieves state-of-the-art performance in transmission line segmentation on the public TTPLA dataset and PLDU dataset. Moreover, the computational efficiency of the proposed model makes it potentially deployable on mobile platforms. The testing code has been released at: https://github.com/xiachengs/MiT-Unet.


  • Power Line Segmentation by Multilevel Attention From Hough Domain

    by Ning Wei, Yuyang Jin, Zhiping Dan, Na Liu, Fangzheng Peng, Shuifa Sun

    Power line segmentation is one of the fundamental and important research topic in unmanned aerial vehicles (UAV)-based smart grid inspection. However, due to the dilemma that the power line feature is vulnerable in complex backgrounds but its extension is huge in spatial, even when using CNN-based models, the accuracy of its segmentation remains unsatisfactory. To this end, in this paper, we propose a novel power line segmentation network PLNet. In this network, after the back-bone feature is reinforced by the Multi-Channel Feature Fusion (MCFF) module, the Line Feature Extracting (LFE) module is proposed to robustly encode the global straight-line features by leveraging the Hough domain. Moreover, the output of LEF guides the encoding of lower-level features through the attention mechanism in the proposed Line Feature Attention(LFA) module. Finally, through the multi-level top-down manner structure, the segmentation detail is progressively refined. Experimental results demonstrate that our proposed method achieves state-of-the-art performance in power line segmentation on the public TTPLA dataset. The testing code is available online at https://github.com/JYYMALL/PLNet.


  • Detecting Insulator Strings as Linked Chain Structure in Smart Grid Inspection

    by Ning Wei, Xiangyang Li, Jiaqi Jin, Peng Chen, Shuifa Sun

    In high-voltage power systems, insulators are essential components in transmission lines for increasing shooting distance and securing wires. Unmanned aerial vehicle imaging becomes a common way of inspecting the state of the insulators. However, the automatic detection of insulators with complex backgrounds is still a challenging task. Most of the existing object detection methods are based on anchors, which do not have sufficient ability to describe objects that have a string-like structure. To tackle it, inspired by the keypoints-based object detection method, we propose a novel chain structure framework to detect insulators. First, we model the string-like object as a chain structure consisting of keypoints and their linkage, by which the insulator strings features are efficiently encoded and trained in the proposed ChainNet. Then, an assembling algorithm is proposed to assemble the estimated keypoints and linkages into chains, by which the insulator strings can be tightly enclosed in rotational bounding boxes. We evaluate the proposed approach on our collected large-scale rigorous dataset under the Directed Intersection over Union metric. The extensive experimental results show that the proposed method achieves 75.7% mAP, which yields 3.4, 15, and 10.6 improvements to the state-of-the-art rotational anchor, axial-aligned anchor, and anchor-free detection methods, respectively. Moreover, the proposed framework can easily be extended to detect other string-like manmade objects in the industrial area.


  • Open Access

    Mixed-type data generation method based on generative adversarial networks

    by Ning Wei, Longzhi Wang, Guanhua Chen, Yirong Wu, Shunfa Sun, Peng Chen

    Data-driven based deep learing has become a key research direction in the field of artificial intelligence. Abundant training data is a guarantee for building efficient and accurate models. However, due to the privacy protection policy, research institutions are often limited to obtain a large number of training data, which would lead to a lack of training sets circumstance. In this paper, a mixed-type data generation model based on generative adversarial networks is proposed to synthesize fake data that have the same distribution with the real data, so as to supplement the real data and increase the number of available samples. The model first pre-trains the autoencoder which maps given dataset into a low-dimensional continuous space. Then, the generator constructed in the low-dimension space is obtained by training it adversarially with discriminator constructed in the original space. Since the constructed discriminator not only consider the loss of the continuous attributes but also the labeled attributes, the generator nets formed by the generator and the decoder can effectively learn the intrinsic distribution of the mixed data. We evaluate the proposed method both in the independent distribution of the attribute and in the relationship of the attributes, and the experiment results show that the proposed generate method has a better performance in preserve the intrinsic distribution compared with other generation algorithms based on deep learning.


  • Open Access

    Mixed-Type Data Generation Method Based On Generative Adversarial Networks

    by Ning Wei, Longzhi Wang, Guanhua Chen, Yirong Wu, Shuifa Sun, Peng Chen

    Data-driven based deep learing has become a key research direction in the field of artificial intelligence. Abundant training data is a guarantee for building efficient and accurate models. However, due to the privacy protection policy, research institutions are often limited to obtain a large number of training data, which would lead to a lack of training sets circumstance. In this paper, a mixed data generation model (mixGAN) based on generative adversarial networks (GANs) is proposed to synthesize fake data that have the same distribution with the real data, so as to supplement the real data and increase the number of available samples. The model first pre-trains the autoencoder which maps given dataset into a low-dimensional continuous space. Then, the Generator constructed in the low-dimension space is obtained by training it adversarially with Discriminator constructed in the original space. Since the constructed Discriminator not only consider the loss of the continuous attributes but also the labeled attributes, the generator nets formed by the Generator and the decoder can effectively learn the intrinsic distribution of the mixed data. We evaluate the proposed method both in the independent distribution of the attribute and in the relationship of the attributes, and the experiment results show that the proposed generate method has a better performance in preserve the intrinsic distribution compared with other generation algorithms based on deep learning.

  • Open Access

    Surface Saliency Detection Based on Curvature Co-Occurrence Histograms

    by Ning Wei, Kaiyuan Gao, Rongrong Ji, Peng Chen

    This paper presents a 3-D surface saliency feature detection method, which can measure the importance geometry region of the point clouds. Different from existing approaches that are based on 3-D filter banks, the new method first constructs the curvature co-occurrence histogram (CCH), which encodes not only the global curvature occurrence, but also the co-occurrence of local distinctive features. And then, the mesh saliency is extracted from CCH through our mapping function. The proposed method is easy to implement and has high computation efficiency, which makes it especially suitable for large-scale 3-D point cloud preprocessing. The effectiveness of the saliency is demonstrated by point clouds’ registration and mesh simplification. Experimental results, from visual and quantification, demonstrate that our saliency contains more local geometrical details information and has more stable globally measurement comparing with curvature only described feature and center-surround saliency.


  • Volume Rendering of Ultrasonic Sequential Images Based on Resampled Data

    by Yihao Yang, Ning Wei, Fangmin Dong, Peng Chen, Jian Wang, Xiaolei Wang, Hong Yao, Changqing Cai, Yue Zhang, Jingan Ding, et al.

    Buoyancy correction of air for weight volume is very important in the mass measurement. Traditional hydrostatic method for measuring weight volume is the most accurate method currently, but it is a complicated process requiring the weight to be wetted in a reference fluid. To eliminate the influence of liquid against weight mass in the determination of weight volume of the hydrostatic method, the measurement method of weight volume based on acoustic method is presented in the paper. According to the principle of sound voltage balance, the measurement system of weight volume is designed for measuring the weight volume. The results from 10 g to 100 g show accuracy of the measurement system is enough to measuring volume of class E weight. The relative extend uncertainty of system is less than 36.7×10 -4 .