Abstract
Mineral image classification technology based on machine vision is an efficient system for ore sorting. With the development of artificial intelligence and computer technology, the deep learning-based mineral image classification system is gradually applied to ore sorting. However, there is a bottleneck in improving classification accuracy, and the feature extraction ability of the CNNs model is relatively limited for multi-category mineral image classification tasks. Therefore, four visual attention blocks are designed and embedded in the existing CNNs model, and new mineral image classification models based on the visual attention mechanism and CNNs are proposed. Then, referring to the building strategies of the different depth ResNet, we build various CNNs model embedding with attention blocks for mineral image classification and visualize the models by Grad-CAM to observe the change in classification weight distributions and classification weight values. Finally, by using the confusion matrices, this experiment systematically evaluates the classification performance of the proposed models and analyzes the misjudgment rate.
Highlights
At this stage, the exploitation and application of mineral resources have entered a new era since the inventory of their mineral resources has declined rapidly with the growth of industrial development, which raises new demands for ore mining and application technology
According to the evaluation results of ResNet in the anthracite data set, we find that the improvement of classification performance and the change of training time caused by embedding attention block are similar to that in gas coal and coking coal data sets
CONCLUSION & OUTLOOK In order to solve the problems of low classification accuracy in multi-category mineral image classification tasks and low efficiency of mineral image feature extraction in convolution neural networks (CNNs) models, combining with the visual attention mechanism, four construction strategies of the visual attention module are proposed, including the Squeeze and Excitation (SE) block, Channel Attention (CA) block, Spatial Attention (SA) block, and Mixed Attention (MA) attention block, and all of them can be flexibly embedded into the existing general CNNs models
Summary
The exploitation and application of mineral resources have entered a new era since the inventory of their mineral resources has declined rapidly with the growth of industrial development, which raises new demands for ore mining and application technology. The intelligent ore sorting equipment put into production is mainly based on ray sensors and used in large-grain particle identification and separation, including XRT and XRF, which has a high classification accuracy and fast classification speed [2]–[5] The problems such as high cost and high radiation still limit their further application and development. In contrast to ray sensor-based sorting equipment, machine vision-based ore sorting equipment extracts the ore feature information from the images collected through optical components and completes the image classification task in static or dynamic scenes It has the advantages of low cost, high efficiency, no radiation, and easy installation. In order to solve the above difficulties and improve the application potentials of the deep learning-based ore sorting equipment, this paper takes the multi-category ore image classification task as the research aspect and proposes to embed the visual attention mechanism in the deep learning-based mineral image classification model. (4) How the visual attention modules influence the distributions and values of model classification weight?
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
