Efficient image segmentation based on deep learning for mineral image classification

Ore image classification based on small deep learning model: Evaluation and optimization of model depth, model structure and data size

Deep learning-based damage detection of mining conveyor belt

Deep learning-based image classification for online multi-coal and multi-class sorting

The results of the work have been obtained within the framework of the research-and-development work “Development of advanced technologies for the complete mining of steam coal with the accumulation of waste rocks in the underground space” (No. 0120U101099). The authors are grateful to the specialists of PJSC “DTEK Pavlohradvuhillia” for the information on the mine provided for research-and-technology analysis.

/pdf/research-into-balance-of-rocks-and-underground-cavities-3ii6178f2o.pdf

Research into balance of rocks and underground cavities formation in the coal mine flowsheet when mining thin seams

Multi-information online detection of coal quality based on machine vision

Sensor-based intelligent sorting technology is a mineral separation technology with the merits of high-efficiency, energy-saving and water-saving. However, the prediction accuracy of conventional machine learning methods is unstable in multi-class selection of ores. The purpose of this study is to propose a competitive voting method to improve the multi-class prediction accuracy of ores in machine vision-based sorting system by combining the classification advantages of various machine learning methods. The operations of image segmentation, feature extraction and feature selection are presented to obtain the multi-class datasets. Three ones of traditional machine learning models with higher classification accuracies are used to establish competitive voting classification models. A case study using the image data of a gas coal shows the merits of the proposed approach. Results derived using this competitive voting approach reveal that it outperforms preexisting approaches.

/pdf/competitive-voting-based-multi-class-prediction-for-ore-4tqypnu6ft.pdf

Competitive Voting-based Multi-class Prediction for Ore Selection

Adversarial attack and adversarial detection have become a hot issue in the field of deep learning based image forensics. However, current researches mainly focus on optical images. Synthetic aperture radar (SAR) images are quite different from the optical images in both imaging mechanism and data structure. This paper aims to study adversarial attack and adversarial detection for SAR images. Firstly, we analyze the distribution characteristics of SAR adversarial examples (AEs) in both output space and feature space by transferring optical attacks. In order to match the digital perturbation with the scattering energy of target, we then propose a generation method of SAR AEs with regional constraint. Experiments show that the proposed method generates SAR AEs that can evade current adversarial detection at the cost of attack success rate. Finally, we point out an open issue that decreasing the perturbation scale leads to the degradation of adversarial detection against both optical AEs and SAR AEs.

An Empirical Study Towards SAR Adversarial Examples

Adversarial examples (AEs) bring increasing concern on the security of deep-learning-based synthetic aperture radar (SAR) target recognition systems. SAR AEs with perturbation constrained to the vicinity of the target have been recently in the spotlight due to the physical realization prospects. However, current adversarial detection methods generally suffer severe performance degradation against SAR AEs with region-constrained perturbation. To solve this problem, we treated SAR AEs as low-probability samples incompatible with the clean dataset. With the help of energy-based models, we captured an inherent energy gap between SAR AEs and clean samples that is robust to the changes of the perturbation region. Inspired by this discovery, we propose an energy-based adversarial detector, which requires no modification to a pretrained model. To better distinguish the clean samples and AEs, energy regularization was adopted to fine-tune the pretrained model. Experiments demonstrated that the proposed method significantly boosts the detection performance against SAR AEs with region-constrained perturbation.

/pdf/energy-based-adversarial-example-detection-for-sar-images-3qijcgd4.pdf

Energy-Based Adversarial Example Detection for SAR Images

The invention belongs to the field of detection technology for tooth direction of a gear ring of a wind turbine bearing and provides a detector and a detection method for tooth direction of the gear ring of the wind turbine bearing; the detector for tooth direction comprises a measuring datum block (1), a measuring rod (2) and a measuring instrument, wherein the lower end surface of the measuring datum block is a measuring datum plane arranged on a datum end surface (5) processed by being placing on a bearing tooth to be measured, the measuring rod (2) is parallel to a processing surface (6) of the bearing tooth to be measured, and the measuring rod (2) is in inserting combination with the measuring datum block (1) through a chute (4) of the measuring datum block; the lower end of the measuring rod (2) is provided with a level indicator (3) for detecting tooth direction error, and the head of the level indicator (3) is contacted with the processing surface (6) of the bearing tooth to be measured In the invention, the difficult problem that the tooth direction of a gear of the wind turbine bearing is detected is solved, so that the tooth direction deviation of the gear of the wind turbine bearing can be visually reflected and the final tooth direction precision of the gear of the bearing can be judged accurately

Zhiwei Zhang

Papers

Multi-information online detection of coal quality based on machine vision

Competitive Voting-based Multi-class Prediction for Ore Selection

An Empirical Study Towards SAR Adversarial Examples

Energy-Based Adversarial Example Detection for SAR Images

Detector and detection method for tooth direction of gear ring of wind turbine bearing