20 September 2021 in IEEE Journal of Biomedical and Health Informatics
The combination of Raman spectroscopy and deep learning technology provides an automatic, rapid, and accurate scheme for the clinical diagnosis of pathogenic bacteria. However, the accuracy of existing deep learning methods is still limited because of the single and fixed scales of deep neural networks. We propose a deep neural network that can learn multi-scale features of Raman spectra by using the automatic combination of multi-receptive fields of convolutional layers. This model is based on the expert knowledge that the discrimination information of Raman spectra is composed of multi-scale spectral peaks. We enhance the interpretability of the model by visualizing the activated wavenumbers of the bacterial spectrum that can be used for reference in related work. Compared with existing state-of-the-art methods, the proposed method achieves higher accuracy and efficiency for bacterial identification on isolate-level, empiric-treatment-level, and antibiotic-resistance-level tasks. The clinical bacterial identification task requires significantly fewer patient samples to achieve similar accuracy. Therefore, this method has tremendous potential for the identification of clinical pathogenic bacteria, antibiotic susceptibility testing, and prescription guidance.
15 November 2023 in Knowledge-Based Systems
18 March 2024 in IEEE Transactions on Instrumentation and Measurement
Due to underwater light absorption and scattering, underwater images often suffer from color distortion and low contrast. However, existing underwater image enhancement methods can only solve the problem of underwater image degradation in a specific scene, and do not consider the impact of dynamic changes in water depth and water quality on underwater image degradation. In order to solve the above problems, this paper proposes a multi-channel adaptive fusion underwater image enhancement algorithm. Firstly, in order to alleviate the color distortion issue caused by light attenuation, this paper proposes a gridded adaptive channel compensation algorithm. Subsequently, the compensated image is used for multi-channel image enhancement. The first channel image is the enhanced result of the compensated image using the local entropy-constrained gray world algorithm proposed in this paper and the second channel image is the enhanced result of the first channel image using the contrast limited adaptive histogram equalization(CLAHE) algorithm. Next, the saliency weight image and brightness weight image of the two-channel images are calculated respectively. Finally, in order to combine the information advantages of different channel images, this paper adopts Laplacian-Gaussian pyramid to fuse the two-channel enhanced images and their corresponding weight images to obtain the final enhanced image. Experiments on three datasets, LSUI, UIEB and RUIE, show that the underwater images processed by the algorithm proposed in this paper have a good improvement in color and contrast, and robustness is better than the comparative methods. The UIQM value of the enhanced images is 126.844% higher than the original images, the UCIQE value is 66.370% higher than the original images, and the number of ORB feature points is 95.862% higher than the original images. Results available at https://drive.google.com/file.
21 June 2024 in IEEE Transactions on Instrumentation and Measurement
Clear images captured under transformer oil are crucial data sources for internal fault diagnosis of oil-immersed transformer. However, images captured under transformer oil suffer from color distortion, low brightness and detail distortion. In order to solve the above problems, a novel multi-channel feature fusion algorithm is proposed in this paper. The first channel image is derived from the gridded gray world white balance algorithm and the adaptive dynamic color channel compensation algorithm proposed in this paper. The second channel image is derived from the standard deviation weighted multi-scale retinex with color restoration (MSRCR) proposed in this paper. The third channel image is derived from the Laplace sharpening based on the second channel image. Then, the saliency weight and brightness weight of the three channel images are calculated respectively, and the weights are normalized. Finally, the three channel images and the corresponding weightes are fused using the multi-scale fusion Gaussian pyramid principle to obtain a clear image under the transformer oil. The experimental results show that the proposed algorithm can better solve the problems of color distortion, detail distortion and low brightness under transformer oil than the comparison methods. The UIQM value of the enhanced images is 115.483% higher than the original images, the UCIQE value is 33.517% higher than the original images, the NIQE value of the enhanced images is 16.420% lower than the original images.
1 March 2022 in Journal of Electronic Imaging
Loop closure detection (LCD) is an indispensable part of simultaneous localization and mapping (SLAM) systems, making it significant for eliminating accumulative trajectory error and constructing a consistent map. Focusing on visual SLAM, we propose a LCD method using local-global similarity measurement strategies. As with most deep learning-based methods, we treat the LCD problem as a classification problem. We use a pretrained VGG-16 network to obtain advanced features and a multiscale feature extraction module to integrate the features of different receptive fields. Finally, we overcome the problems of dynamic object occlusion and large-area repetitive texture using a local-global similarity measurement module to measure the similarity between current and historical frames in all dimensions. This approach decouples feature extraction from classification determination, which enables us to use the advantages of both supervised and unsupervised learning to cope with the imbalance in the number of nonloop and loop pairs. Experimental results on three different datasets showed that the proposed method is significantly better than the state-of-the-art appearance-based LCD methods in dynamic urban, suburban, and highway scenarios.
21 August 2025 in ISA Transactions