Ke Xu

Dr. Ke Xu

Nanjing Agricultural University
orcid

Research Focus & Metrics

Academic Career Length

16Years (2011-Present)

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

Representative Publications

View All Publications
12 June 2020 in IEEE Access
Verified
Journal Article
Open access
Ke XuKe Xu
Ke Xu
HLHuaimin Li
Huaimin Li
WCWeixing Cao
Weixing Cao
YZYan Zhu
Yan Zhu
RCRongjia Chen
Rongjia Chen
JNJun Ni
Jun Ni
Due to the low recognition rate of weeds in wheat fields and the inability to accurately locate weeds, we propose a recognition method for weeds in natural wheat fields based on the fusion of RGB image features and depth features. The method breaks through the limitations of the two-dimensional spatial features extracted from RGB images when recognizing grass weeds similar to wheat. According to the species, distribution of weeds in wheat fields, we extracted the color, position, texture, and depth features of weeds in wheat fields from RGB and depth images during the tillering and jointing stages. And then used the AdaBoost algorithm for the integrated learning of multiple classifiers, thereby achieving the recognition of weeds in wheat fields. The experimental results revealed that the recognition speed of weeds during the tillering stage was 0.2 s and the accuracy rate was 88%. The recognition speed of weeds during the jointing stage was 0.69 s, and the accuracy rate of weed recognition was 81.08%. These results are significantly higher than the weed recognition rate based on features extracted from RGB images.

29 March 2022 in Agronomy
Verified
Journal Article
Open access
PDF file
HLHuaimin Li
Huaimin Li
DLDonghang Li
Donghang Li
Ke XuKe Xu
Ke Xu
WCWeixing Cao
Weixing Cao
XJXiaoping Jiang
Xiaoping Jiang
Jun NiJun Ni
Jun Ni
Canopy spectral reflectance can indicate both crop nutrient and canopy structural information. Differences in canopy structure can affect spectral reflectance. However, a non-imaging spectrometer cannot distinguish such differences while monitoring crop nutrients, because the results are likely to be influenced by the canopy structure. In addition, nitrogen application rate is one of the main factors influencing the canopy structure of crops. Strong correlations exist between indices of canopy structure and leaf nitrogen, and thus, these can be used to compensate for the spectral monitoring of nitrogen content in wheat leaves. In this study, canopy structural indices (CSI) such as wheat coverage, height, and textural features were obtained based on the RGB and height images obtained by the RGB-D camera. Moreover, canopy spectral reflectance was obtained by an ASD hyperspectral spectrometer, based on which two vegetation indices—ratio vegetation index (RVI) and angular insensitivity vegetation index (AIVI)—were constructed. With the vegetation indices and CSIs as input parameters, a model was established to predict the leaf nitrogen content (LNC) and leaf nitrogen accumulation (LNA) of wheat based on partial least squares (PLS) and random forest (RF) regression algorithms. The results showed that the RF model with RVI and CSI as inputs had the highest prediction accuracy for LNA, the coefficient of determination (R2) reached 0.79, and the root mean square error (RMSE) was 1.54 g/m2. The vegetation indices and coverage were relatively important features in the model. In addition, the PLS model with AIVI and CSI as input parameters had the highest prediction accuracy for LNC, with an R2 of 0.78 and an RMSE of 0.35%, among the vegetation indices. In addition, parts of both the textural and height features were important. The results suggested that PLS and RF regression algorithms can effectively integrate spectral and canopy structural information, and canopy structural information effectively supplement spectral information by improving the prediction accuracy of vegetation indices for LNA and LNC.

10 December 2020 in Remote Sensing
Verified
Journal Article
Open access
PDF file
Ke XuKe Xu
Ke Xu
JZJingchao Zhang
Jingchao Zhang
HLHuaimin Li
Huaimin Li
WCWeixing Cao
Weixing Cao
YZYan Zhu
Yan Zhu
Jun NiJun Ni
Jun Ni
The accurate estimation of nitrogen accumulation is of great significance to nitrogen fertilizer management in wheat production. To overcome the shortcomings of spectral technology, which ignores the anisotropy of canopy structure when predicting the nitrogen accumulation in wheat, resulting in low accuracy and unstable prediction results, we propose a method for predicting wheat nitrogen accumulation based on the fusion of spectral and canopy structure features. After depth images are repaired using a hole-filling algorithm, RGB images and depth images are fused through IHS transformation, and textural features of the fused images are then extracted in order to express the three-dimensional structural information of the canopy. The fused images contain depth information of the canopy, which breaks through the limitation of extracting canopy structure features from a two-dimensional image. By comparing the experimental results of multiple regression analyses and BP neural networks, we found that the characteristics of the canopy structure effectively compensated for the model prediction of nitrogen accumulation based only on spectral characteristics. Our prediction model displayed better accuracy and stability, with prediction accuracy values (R2) based on BP neural network for the leaf layer nitrogen accumulation (LNA) and shoot nitrogen accumulation (SNA) during a full growth period of 0.74 and 0.73, respectively, and corresponding relative root mean square errors (RRMSEs) of 40.13% and 35.73%.

3 September 2025 in Sensors
Verified
Journal Article
Open access
PDF file
XZXiaoyu Zhang
Xiaoyu Zhang
RCRui Cai
Rui Cai
NJNing Jiang
Ning Jiang
MXMinwen Xing
Minwen Xing
Ke XuKe Xu
Ke Xu
YHYaocong Hu
Yaocong Hu
Person re-identification aims to match images of the same individual across non-overlapping cameras by analyzing personal characteristics. Recently, Transformer-based models have demonstrated excellent capabilities and achieved breakthrough progress in this task. However, their high computational costs and inadequate capacity to capture fine-grained local features impose significant constraints on re-identification performance. To address these challenges, this paper proposes a novel Toward Efficient Transformer-based Person Re-identification (TE-TransReID) framework. Specifically, the proposed framework retains only the former L-th layer layers of a pretrained Vision Transformer (ViT) for global feature extraction while combining local features extracted from a pretrained CNN, thus achieving the trade-off between high accuracy and lightweight networks. Additionally, we propose a dual efficient feature-fusion strategy to integrate global and local features for accurate person re-identification. The Efficient Token-based Feature-Fusion Module (ETFFM) employs the gate-based network to learn fused token-wise features, while the Efficient Patch-based Feature-Fusion Module (EPFFM) utilizes a lightweight Transformer to aggregate patch-level features. Finally, TE-TransReID achieves a rank-1 of 94.8%, 88.3%, and 85.7% on Market1501, DukeMTMC, and MSMT17 with a parameter of 27.5 M, respectively. Compared to existing CNN–Transformer hybrid models, TE-TransReID maintains comparable recognition accuracy while drastically reducing model parameters, establishing an optimal equilibrium between recognition accuracy and computational efficiency.

1 December 2011
Verified
Conference Paper
Open access
Ke XuKe Xu
Ke Xu
LQLei Qin
Lei Qin
LYLin Yang
Lin Yang
Recently announced RGB-D cameras like Microsoft Kinect are attractive sensing systems providing RGB images along with registered depth information at a high frame rate. This advantage makes such cameras greatly useful in the context of robotics, especially for dense 3D mapping in home environment. However, depth information they provide is usually incomplete and imprecise. This paper presents a method that effectively rectifies depth maps, particularly for missing regions, via fusion of 2D-3D images. Segmentation for plane fitting as well as Markov random field based optimization are applied successively to achieve our objective. We test our algorithm on a range of data with varying kinds and amount of incompleteness.