Figure - available from: Scientific Reports
This content is subject to copyright. Terms and conditions apply.
Source publication
Using image recognition technology to realize coal gangue recognition is one of the development directions of intelligent fully mechanized caving mining. Aiming at the problem of low accuracy of coal gangue recognition in fully mechanized caving mining, the extraction method of Coal and gangue images features is proposed, and the corresponding coal...
Citations
... 21 proposed a model based on thermal infrared imaging technology and improved You Only Look Once version 5 (YOLOv5) algorithm based on the heatsensitive data map of coal and gangue mixed in the dark and wet working conditions to achieve coal and gangue under dark and damp working conditions-accurate identification of mixed situations. Zhang et al. 22 designed a multilight source image acquisition system and introduced the least squares SVM to analyze the images of coal and gangue and further obtain the parameters that distinguish them. Li et al. 23 proposed a coal gangue recognition and positioning method based on lightweight mixed-domain attention to solving high model complexity problems, strenuous training, and poor coal gangue recognition and positioning effects under complex conditions in traditional target detection algorithms. ...
... Cao et al. 27 proposed a group convolution and channel shuffling enhanced ghost network for coal and gangue image classification tasks. Many studies [16][17][18][19][20][21][22][23][24][25][26][27] use image processing and deep learning to identify coal gangue. Still, deep learning algorithms are inseparable from expensive hardware devices and long-time costs, whether in the model training or inference process. ...
Real‐time monitoring of the coal caving process in fully mechanized mining is crucial for achieving intelligent and efficient top‐coal caving. While the coal gangue identification method, employing vision and deep learning, has advanced in the realm of intelligent monitoring, it exhibits a dependency on high‐performance hardware. This reliance poses challenges for deploying identification equipment on mobile terminals, hindering the widespread application of this method. To address the issues above, the paper presents a lightweight algorithm, utilizing You Only Look Once version 5s (YOLOv5s), utilizing YOLOv5s for the real‐time perception of the top‐coal caving state in fully mechanized caving mining. We replace the backbone network of YOLOv5s with the ShuffleNetv2 structure that is more suitable for lightweight deployment, and add the Simple Attention Mechanism attention mechanism to the network structure to enhance the model's receptive field and feature expression ability, and reduce the impact of falling debris on the detection results. A dynamic experimental platform for top‐coal caving in fully mechanized caving mining for thick coal seams is set up, and preprocessing operations such as brightness, sharpening, and denoising are performed on the image data sets collected by high‐speed industrial cameras. Research results show that compared with the traditional YOLOv5s, the improved model's P, mAP, F1 score, and other indicators have increased by 3.4%, 2.1%, and 1.1%, respectively, the model size is 70% of the original, and the detection frames per second value has increased by 48.1%. The lightweight algorithm stabilizes the accuracy of coal gangue identification dramatically in real time. It dramatically reduces the computing pressure on the mobile terminal, providing basic theory and practice for real‐time monitoring of fully mechanized coal caving mining.
... Furthermore, the performance of a feature on disease detection can be affected by various constraints or settings. For example, factors such as lighting and the distance of the camera from the plant leaves can make feature extraction not consistent and challenging [37]. Also, diseases may produce different symptoms in different plant growth stages. ...
... Area is a measure of the total amount of space occupied by an object in two-dimensional space. In plant disease detection, an area can help identify the size of the diseased region and measure the progress of the disease [37]. In plant disease detection, the perimeter can help identify the edges of the infected region and measure the extent of the disease [78]. ...
... However, there are still some challenges in practical applications [1][2][3][4] . The coal gangue sorting robot identifies the target coal gangue and obtains its pose information through the recognition system during the sorting process [5][6][7] . The robot control system combines pose information, belt speed, and time to calculate the real-time position of the target coal gangue, thereby achieving robot sorting. ...
The coal gangue sorting robot may encounter variations in the pose of the target coal gangue due to belt slippage, deviation, and speed fluctuations, leading to failed or missed grasping attempts during the sorting process of coal gangue. In response to this issue, we propose a special two-stage network coal gangue image fast matching method to re-obtain the target coal gangue pose information, further improving the grasping accuracy and efficiency of the robot. In the first stage, we use SuperPoint to enhance the scene adaptability and credibility of feature point extraction. The improved Multi-scale Retinex with Color Restoration enhancement algorithm is used to further enhance the ability of Superpoint to detect feature points. In the second stage, we introduce SuperGlue for feature matching to improve the robustness of the matching network. Integrating the Progressive Sample Consensus algorithm to further eliminate erroneous feature matching point pairs and improve the accuracy of image matching. We conducted matching experiments of coal gangue under different object distances, scales, and rotation angles using various methods on the double-manipulator truss-type coal gangue sorting robot experimental platform independently developed by our team. The results showed that the matching rate of the proposed method was 98.2%, with a matching time of 84.6ms. It has the characteristics of a high matching rate, good real-time performance, and strong robustness, and can meet the requirements of efficient and accurate matching between coal gangue recognition images and sorting images.
... The ImageNet [1] dataset, which was produced and released by Li Feifei's team in two and a half years in 2009, was the key to a new milestone in visual recognition technology. Since AlexNet [2] won the ImageNet large scale visual recognition challenge (ILSVRC) competition in 2012, far surpassing second place, the era of deep learning has arrived, and its application in the field of image processing is becoming increasingly widespread [3][4][5][6]. The effectiveness of neural network models is closely related to the scale and quality of training data, and the production of datasets has become an important factor in improving model performance. ...
To address the issue of the lack of specialized data filtering algorithms for dataset production, we proposed an image filtering algorithm. Using feature fusion methods to improve discrete wavelet transform algorithm (DWT) and enhance the robustness of image feature extraction, a weighted hash algorithm was proposed to hash features to reduce the complexity and computational cost of feature comparison. To minimize the time cost of image filtering as much as possible, a fast distance calculation method was also proposed to calculate the similarity of images. The experimental results showed that compared with other advanced methods, the algorithm proposed in this paper had an average accuracy improvement of 3% and a speed improvement of at least 30%. Compared with traditional manual filtering methods, while ensuring accuracy, the filtering speed of a single image is increased from 9.9s to 0.01s, which has important application value for dataset production.
... In recent years, many scholars have conducted extensive research on coal gangue sorting based on image recognition technology [6][7][8][9][10][11]. Early image recognition methods were based on differences in grayscale and the texture of the surface natural images of coal and gangue, which were collected using CCD cameras. These methods utilized thresholds to distinguish coal and gangue by extracting statistical information about grayscale and texture as image features, achieving high accuracy in coal gangue recognition in laboratory settings [12][13][14][15][16][17]. However, it has been gradually found to be difficult to generalize these methods in practical industrial production, partly due to the stability issues associated with these surface image features. ...
Coal gangue image recognition is a critical technology for achieving automatic separation in coal processing, characterized by its rapid, environmentally friendly, and energy-saving nature. However, the response characteristics of coal and gangue vary greatly under different illuminance conditions, which poses challenges to the stability of feature extraction and recognition, especially when strict illuminance requirements are necessary. This leads to fluctuating coal gangue recognition accuracy in industrial environments. To address these issues and improve the accuracy and stability of image recognition under variable illuminance conditions, we propose a novel coal gangue recognition method based on laser speckle images. Firstly, we studied the inter-class separability and intra-class compactness of the collected laser speckle images of coal and gangue by extracting gray and texture features from the laser speckle images, and analyzed the performance of laser speckle images in representing the differences between coal and gangue minerals. Subsequently, coal gangue recognition was achieved using an SVM classifier based on the extracted features from the laser speckle images. The fusion feature approach achieved a recognition accuracy of 94.4%, providing further evidence of the feasibility of this method. Lastly, we conducted a comparative experiment between natural images and laser speckle images for coal gangue recognition using the same features. The average accuracy of coal gangue laser speckle image recognition under various lighting conditions is 96.7%, with a standard deviation of the recognition accuracy of 1.7%. This significantly surpasses the recognition accuracy obtained from natural coal and gangue images. The results showed that the proposed laser speckle image features can facilitate more stable coal gangue recognition with illumination factors, providing a new, reliable method for achieving accurate classification of coal and gangue in the industrial environment of mines.
... Using machine learning technology to establish a coal gangue recognition classifier, the precise classification of coal and gangue can be realized by extracting the image features of coal and gangue. 19,20 However, the machine learning method needs to manually extract various coal and gangue features, and the process is complicated, which is not conducive to the rapid identification of coal and gangue. In contrast, a convolutional neural network can automatically extract high-level features of images and respond quickly. ...
A coal gangue image recognition method based on complex conditions is proposed to address the current issue of image-based coal gangue recognition being greatly affected by complex conditions. First, complex conditions such as different shooting backgrounds, shooting distances, and lighting intensities are set to simulate the underground coal mining environment. Then, based on three convolutional neural network algorithms, the coal gangue recognition model is established, and the influence of different complex conditions on coal gangue image recognition is analyzed. At the same time, a network model with a strong generalization ability is determined. The results show that the accuracy of coal gangue image recognition has no obvious regularity under different shooting background conditions, and complex environments should be the primary factor affecting the accuracy of coal gangue image recognition. The accuracy of coal gangue image recognition is negatively correlated with the increase in shooting distance, and strong light conditions are conducive to improving the accuracy of coal gangue image recognition. The LeNet network model has the strongest generalization ability, which can meet the requirements of recognition accuracy and respond quickly. The accuracy of coal gangue image recognition under different complex conditions can reach more than 0.99, and the average single image recognition time is only 177 ms. This article studies the influence law of different complex conditions on the recognition of coal and gangue images and confirms that the LeNet network has strong generalization ability, achieving accurate and fast recognition of coal gangue images under complex conditions and providing a reference basis for the deployment of underground coal gangue sorting.
The KDR (karst development rate) of rocks and their PCR(porosity of carbonate rocks) are common research topics in Jinfo Mountain. The use of traditional carbonate research methods (TCRMs) for karst studies has been shown to be costly and time-consuming. Therefore, this study attempted to find a new, reliable, low-cost, and time-saving method for karst research. The Jinfo Mountain area is a typical carbonate rock area that is suitable for karst research. In this study, many images of rock samples from the Jinfo Mountain were obtained using rock-polarizing microscopes, which provided a good basis for the karst study of Jinfo Mountain. Furthermore, in this study, image analysis technology was used to find the karst development rate of rocks and their porosity. To ensure the accuracy of these research results, we compared the research results obtained using the image analysis techniques with those obtained using TCRM. The comparison showed that the image analysis technology is a feasible research techniques for studying karst in the Jinfo Mountain area. Furthermore, it has good reference significance for other karst study outside the Jinfo Mountain area.
