Chengyi Wang

Chengyi Wang


Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

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

Representative Publications

View All Publications
13 February 2018 in Remote Sensing
Verified
Journal Article
Open access
PDF file
JCJingbo Chen
Jingbo Chen
CWChengyi Wang
Chengyi Wang
MZMa Zhong
Ma Zhong
JCJiansheng Chen
Jiansheng Chen
DHDongxu He
Dongxu He
SAStephen Ackland
Stephen Ackland
Semantic-level land-use scene classification is a challenging problem, in which deep learning methods, e.g., convolutional neural networks (CNNs), have shown remarkable capacity. However, a lack of sufficient labeled images has proved a hindrance to increasing the land-use scene classification accuracy of CNNs. Aiming at this problem, this paper proposes a CNN pre-training method under the guidance of a human visual attention mechanism. Specifically, a computational visual attention model is used to automatically extract salient regions in unlabeled images. Then, sparse filters are adopted to learn features from these salient regions, with the learnt parameters used to initialize the convolutional layers of the CNN. Finally, the CNN is further fine-tuned on labeled images. Experiments are performed on the UCMerced and AID datasets, which show that when combined with a demonstrative CNN, our method can achieve 2.24% higher accuracy than a plain CNN and can obtain an overall accuracy of 92.43% when combined with AlexNet. The results indicate that the proposed method can effectively improve CNN performance using easy-to-access unlabeled images and thus will enhance the performance of land-use scene classification especially when a large-scale labeled dataset is unavailable.

30 June 2017 in Sensors
Verified
Journal Article
Open access
PDF file
DBDebojit Biswas
Debojit Biswas
Hongbo SuHongbo Su
Hongbo Su
CWChengyi Wang
Chengyi Wang
JBJason Blankenship
Jason Blankenship
Aleksandar StevanovicAleksandar Stevanovic
Aleksandar Stevanovic
Automatic car counting is an important component in the automated traffic system. Car counting is very important to understand the traffic load and optimize the traffic signals. In this paper, we implemented the Gaussian Background Subtraction Method and OverFeat Framework to count cars. OverFeat Framework is a combination of Convolution Neural Network (CNN) and one machine learning classifier (like Support Vector Machines (SVM) or Logistic Regression). With this study, we showed another possible application area for the OverFeat Framework. The advantages and shortcomings of the Background Subtraction Method and OverFeat Framework were analyzed using six individual traffic videos with different perspectives, such as camera angles, weather conditions and time of the day. In addition, we compared the two algorithms above with manual counting and a commercial software called Placemeter. The OverFeat Framework showed significant potential in the field of car counting with the average accuracy of 96.55% in our experiment.

1 June 2016 in Photogrammetric Engineering & Remote Sensing
Verified
Journal Article
Closed access
MYMing Yan
Ming Yan
CWChengyi Wang
Chengyi Wang
JMJianglin Ma
Jianglin Ma
ZWZhiyong Wang
Zhiyong Wang
BYBingyang Yu
Bingyang Yu
When an optical remote sensing satellite is imaging the Earth in-orbit, the propagation direction of the Line of Sight (LOS) will be changed because of atmospheric refraction. This will result in a geolocation deviation on the collinear rigorous geometric model for direct georeferencing, pushbroom images. To estimate and correct the atmospheric refraction geolocation error, the LOS vector tracking algorithm is introduced and a weighted mean algorithm is used to simplify the ISO standard atmospheric model into a troposphere and stratosphere, i.e., two layers spherical atmosphere. The simulation result shows the atmospheric refraction will introduce about 2 m and 7.5 m geometric displacement when the spacecraft is off-pointed view at 30 and 45 degree angle, respectively. For a state-of-the-art high resolution satellite, the atmospheric refraction displacement shall be corrected. The method has been practiced in the DMC3/TripleSat Constellation to remove the atmospheric refraction geolocation error without ground control points.

21 January 2019 in Remote Sensing
Verified
Journal Article
Open access
PDF file
YYYuan Yuan
Yuan Yuan
LLLei Lin
Lei Lin
JCJingbo Chen
Jingbo Chen
HSHichem Sahli
Hichem Sahli
YCYixiang Chen
Yixiang Chen
BWBin Wu
Bin Wu
Large amounts of farmland loss caused by urban expansion has been a severe global environmental problem. Therefore, monitoring urban encroachment upon farmland is a global issue. In this study, we propose a novel framework for modelling and monitoring the conversion of cultivated land to built-up land using a satellite image time series (SITS). The land-cover change process is modelled by a two-level hierarchical hidden semi-Markov model, which is composed of two Markov chains with hierarchical relationships. The upper chain represents annual land-cover dynamics, and the lower chain encodes the vegetation phenological patterns of each land-cover type. This kind of architecture enables us to represent the multilevel semantic information of SITS at different time scales. Specifically, intra-annual series reflect phenological differences and inter-annual series reflect land-cover dynamics. In this way, we can take advantage of the temporal information contained in the entire time series as well as the prior knowledge of land cover conversion to identify where and when changes occur. As a case study, we applied the proposed method for mapping annual, long-term urban-induced farmland loss from Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) time series in the Jing-Jin-Tang district, China from 2001 to 2010. The accuracy assessment showed that the proposed method was accurate for detecting conversions from cultivated land to built-up land, with the overall accuracy of 97.72% in the spatial domain and the temporal accuracy of 74.60%. The experimental results demonstrated the superiority of the proposed method in comparison with other state-of-the-art algorithms. In addition, the spatial-temporal patterns of urban expansion revealed in this study are consistent with the findings of previous studies, which also confirms the effectiveness of the proposed method.

16 January 2017 in Remote Sensing
Verified
Journal Article
Open access
PDF file
LLLu Li
Lu Li
CWChengyi Wang
Chengyi Wang
JCJingbo Chen
Jingbo Chen
JMJianglin Ma
Jianglin Ma
This paper proposes a novel method of segment-tree filtering to improve the classification accuracy of hyperspectral image (HSI). Segment-tree filtering is a versatile method that incorporates spatial information and has been widely applied in image preprocessing. However, to use this powerful framework in hyperspectral image classification, we must reduce the original feature dimensionality to avoid the Hughes problem; otherwise, the computational costs are high and the classification accuracy by original bands in the HSI is unsatisfactory. Therefore, feature extraction is adopted to produce new salient features. In this paper, the Semi-supervised Local Fisher (SELF) method of discriminant analysis is used to reduce HSI dimensionality. Then, a tree-structure filter that adaptively incorporates contextual information is constructed. Additionally, an initial classification map is generated using multi-class support vector machines (SVMs), and segment-tree filtering is conducted using this map. Finally, a simple Winner-Take-All (WTA) rule is applied to determine the class of each pixel in an HSI based on the maximum probability. The experimental results demonstrate that the proposed method can improve HSI classification accuracy significantly. Furthermore, a comparison between the proposed method and the current state-of-the-art methods, such as Extended Morphological Profiles (EMPs), Guided Filtering (GF), and Markov Random Fields (MRFs), suggests that our method is both competitive and robust.

27 July 2012 in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Verified
Journal Article
Open access
PDF file
SZS. Zheng
S. Zheng
YZY. Zhou
Y. Zhou
RHR. Huang
R. Huang
LZLangming Zhou
Langming Zhou
XXX. Xu
X. Xu
CWChengyi Wang
Chengyi Wang
Three-dimensional measurement and reconstruction during conservation and restoration of cultural relics have become an essential part of a modem museum regular work. Although many kinds of methods including laser scanning, computer vision and close-range photogrammetry have been put forward, but problems still exist, such as contradiction between cost and good result, time and fine effect. Aimed at these problems, this paper proposed a structure-light based method for 3D measurement and reconstruction of cultural relics in museums. Firstly, based on structure-light principle, digitalization hardware has been built and with its help, dense point cloud of cultural relics' surface can be easily acquired. To produce accurate 3D geometry model from point cloud data, multi processing algorithms have been developed and corresponding software has been implemented whose functions include blunder detection and removal, point cloud alignment and merge, 3D mesh construction and simplification. Finally, high-resolution images are captured and the alignment of these images and 3D geometry model is conducted and realistic, accurate 3D model is constructed. Based on such method, a complete system including hardware and software are built. Multi-kinds of cultural relics have been used to test this method and results prove its own feature such as high efficiency, high accuracy, easy operation and so on.