Liang Xu

Prof. Dr. Liang Xu

orcid

Research Focus & Metrics

Academic Career Length

17Years (2010-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 June 2021 in Forests
Verified
Journal Article
Open access
PDF file
JPJin Pan
Jin Pan
XOXiaoming Ou
Xiaoming Ou
LXLiang Xu
Liang Xu
Forest fires are serious disasters that affect countries all over the world. With the progress of image processing, numerous image-based surveillance systems for fires have been installed in forests. The rapid and accurate detection and grading of fire smoke can provide useful information, which helps humans to quickly control and reduce forest losses. Currently, convolutional neural networks (CNN) have yielded excellent performance in image recognition. Previous studies mostly paid attention to CNN-based image classification for fire detection. However, the research of CNN-based region detection and grading of fire is extremely scarce due to a challenging task which locates and segments fire regions using image-level annotations instead of inaccessible pixel-level labels. This paper presents a novel collaborative region detection and grading framework for fire smoke using a weakly supervised fine segmentation and a lightweight Faster R-CNN. The multi-task framework can simultaneously implement the early-stage alarm, region detection, classification, and grading of fire smoke. To provide an accurate segmentation on image-level, we propose the weakly supervised fine segmentation method, which consists of a segmentation network and a decision network. We aggregate image-level information, instead of expensive pixel-level labels, from all training images into the segmentation network, which simultaneously locates and segments fire smoke regions. To train the segmentation network using only image-level annotations, we propose a two-stage weakly supervised learning strategy, in which a novel weakly supervised loss is proposed to roughly detect the region of fire smoke, and a new region-refining segmentation algorithm is further used to accurately identify this region. The decision network incorporating a residual spatial attention module is utilized to predict the category of forest fire smoke. To reduce the complexity of the Faster R-CNN, we first introduced a knowledge distillation technique to compress the structure of this model. To grade forest fire smoke, we used a 3-input/1-output fuzzy system to evaluate the severity level. We evaluated the proposed approach using a developed fire smoke dataset, which included five different scenes varying by the fire smoke level. The proposed method exhibited competitive performance compared to state-of-the-art methods.

2 March 2020 in IEEE Access
Verified
Journal Article
Open access
LXLiang Xu
Liang Xu
SLShuai Lv
Shuai Lv
YDYong Deng
Yong Deng
XLXiuxi Li
Xiuxi Li
Surface defect detection is a critical task in product quality assurance for manufacturing lines. The deep learning-based methods recently developed for defect detection are typically trained using a supervised learning strategy and large defect sample sets. Conventional methods often require additional pixel-level labeling or bounding boxes to predict the location of defects. However, the number of required samples and the time-intensive annotation process limits the practical use of these algorithms. As such, this study proposes a weakly supervised detection framework in which a CNN model is trained to identify surface cracks in motor commutators. The model was trained using small subsets of defect samples (~5-30) and does not require a pre-trained network. This approach consists of localization and decision networks that simultaneously predict both the location and probability of defects. A new loss function was also developed to identify abnormal regions in a sample with accessible image-level labels. A collaboration learning strategy was then applied to utilize the loss function and compensate for imbalances at the pixel level. Experimental results using a small number of image-level training labels from a real industrial dataset exhibited a 99.5% recognition accuracy, which is comparable to relevant methods using pixel-level labels.

24 July 2019 in IEEE Access
Verified
Journal Article
Open access
LXLiang Xu
Liang Xu
YWYuxi Wang
Yuxi Wang
XLXiuxi Li
Xiuxi Li
MPMing Pan
Ming Pan
Many deep-learning character recognition methods have been developed over the past few years. Chinese characters are widely used in many countries; however deep learning-based Chinese character recognition methods are faced with various problems, such as a large amount of data required for training, numerous parameters, and a large consumption of computing resources. Concept learning is a hominine learning approach. Unlike existing deep learning models, conceptual model learning can be realized by using as little as one sample. This study is the first to propose a handwritten Chinese character recognition method based on concept learning. Different from the existing image representation-based character recognition methods, the proposed method builds a meta stroke library with prior knowledge, and then presents a Chinese character conceptual model based on stroke relationship learning using a character stroke extraction method and Bayesian program learning. During character recognition, Monte Carlo Markov chain sampling is utilized to obtain the character generation model for each character conceptual. This generation model can calculate the probability of the target and training characters being the same classification, and thereby determines the classification of the target character. The experimental results indicate that, with the proposed method, the conceptual model of each character can be built for character classification prediction using as few as one character sample. Our approach obtains better performance than the state-of-the-art methods on ICDAR-2013 competition dataset.

26 April 2019 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
LXLiang Xu
Liang Xu
HXHaibo Xu
Haibo Xu
XLXiuxi Li
Xiuxi Li
MPMing Pan
Ming Pan
Surface defects of explosive cartridge in the automatic sorting process are of small area, irregular shape, and random distribution, all problematic characteristics that hinder surface defect detection. To address these issues, a new multi-defect detection method has been proposed in this study based on a combination of an improved visual attention model and image partitioning-weighted eigenvalue. Firstly, image pre-processing is carried out by a background estimation algorithm. Then, a new fusion operator based on defects discrimination is implemented in a visual attention model to integrate intensity, orientation and edge conspicuity into a saliency graph, which a saliency effect of defects is considered during the fusion process. Thirdly, a saliency map is divided into image blocks based on image variance. This allows for the extraction of image blocks including defects, the calculation of the weighted eigenvalue, and the determination of regions containing multi-defect. The image partitioning-weighted eigenvalue is used to make a decision for multi-defect. The experimental results show this method’s detection accuracy as 98.2%, with a less computation time and quickly detection speed. Therefore, this method could be adopted for on-line detection systems for explosive cartridge surface defects.

1 June 2016 in Measurement
Verified
Journal Article
Closed access
LXLiang Xu
Liang Xu
XHXiao-Min He
Xiao-Min He
XLXiu Xi Li
Xiu Xi Li
MPMing Pan
Ming Pan

14 July 2020 in IEEE Access
Verified
Journal Article
Open access
LXLiang Xu
Liang Xu
ZLZhiliang Li
Zhiliang Li
XLXiuxi Li
Xiuxi Li
Accurate localization is critical in the internet of things (IOT), especially for wireless sensor networks (WSNs). Location estimation can be affected by factors such as node density, topological diversity, and sensor coverage. As such, we propose a hybrid approach using multistage collaborative calibration for wireless sensor network localization, specifically in 3D environments. This technique integrates a Modified version of Light Gradient Boosting Model (MLGB), which is based on a regression scheme, a cooperative methodology, and a fine calibration model for collaborative fusion. These techniques were combined with quadrilateral shrunk centroid (QSC) and distance vector hop algorithms, using a multi-communication radius and an improved frog-leaping algorithm (DVMFL). In the first step, MLGB was used to correct for inhomogeneous localization estimation errors and RSSI data sparsity. As a result, the model is capable of adapting to high topological diversity (i.e., C-shape, H-shape, S-shape, and O-shape).Successive steps further improved prediction accuracy by using a screening cooperative anchor node strategy to increase node density and enhance the QSC-DVMFL fusion framework for fine position estimation. The proposed methodology was assessed in a series of validation, comparing it to other techniques. The results demonstrated a clear effectiveness and adaptability across a variety of factors that typically affect WSN localization.