Unmanned aerial vehicles (UAVs) are being employed in a rapidly increasing number of applications in mining, including the emerging area of mapping underground void spaces such as stopes, which are otherwise inaccessible to humans, automated ground vehicles and survey technologies. Void mapping can provide both visual rock surface and 3D structural information about stopes, supporting more effective planning of ongoing blast designs. Underground stope mapping by UAVs, however, involves overcoming a number of engineering challenges to allow flights beyond operator line-of-sight where there is no global positioning system (GPS), natural or artificial light, or existing communications infrastructure. This paper describes the construction of a UAV sensor suite that uses sound navigation and ranging (SONAR) data to create a rough 3D model of the underground UAV operational environment in real time to provide operators with high situational awareness for beyond line-of-sight operations. The system also provides a backup when dust obscures visual sensors to provide situation awareness and a coarser, but still informative, 3D model of the underground space. Typically, light detection and ranging (LIDAR) systems have superseded SONAR sensors for similar applications. LIDAR is much more accurate than SONAR, but has several disadvantages. SONAR sensor data is sparse, and therefore much easier to process in real time on-board the UAV than LIDAR. The SONAR sensor hardware is also lighter than current LIDAR systems, which is of importance regarding the constrained payload capacity of UAVs. However, the most important factor that makes SONAR stand out in this application is its ability to operate in dusty or smoke-filled environments. The UAV system was tested both above and below-ground using a predefined path with check point locations for the UAV to follow. Due to the lack of GPS, survey points in combination with photogrammetry allowed the UAV's location to be estimated. This allowed the system to be tested to determine how accurate the SONAR data is in comparison with 3D modelling via photogrammetry of images from a separate digital single-lens reflex camera. Comparing the shape of void surfaces determined by photogrammetry with that determined by SONAR provides quantifiable accuracy when the photogrammetry models are used as ground truth data. Above-ground and underground pilot studies have determined that SONAR sensors provide acceptable accuracy compared with modelling via photogrammetry, sufficient to provide effective situational awareness for human operation of the UAV beyond line-of-sight.
Surface cracks on buildings, natural walls and underground mine tunnels can indicate serious structural integrity issues that threaten the safety of the structure and people in the environment. Timely detection and monitoring of cracks is crucial to managing these risks, especially if the systems can be made highly automated through robots. Visionbased crack detection algorithms using deep neural networks have exhibited promise for structured surfaces such as walls or civil engineering tunnels, but little work has addressed highly unstructured environments such as rock cliffs and bare mining tunnels. To address this challenge, this paper presents PointCrack3D, a new 3D-point-cloud-based crack detection algorithm for unstructured surfaces. The method comprises three key components: an adaptive down-sampling method that maintains sufficient crack point density, a DNN that classifies each point as crack or non-crack, and a post-processing clustering method that groups crack points into crack instances.The method was validated experimentally on a new large natural rock dataset, comprising coloured LIDAR point clouds spanning more than 900 m 2 and 412 individual cracks. Results demonstrate a crack detection rate of 97% overall and 100% for cracks with a maximum width of more than 3 cm, significantly outperforming the state of the art. Furthermore, for crossvalidation, PointCrack3D was applied to an entirely new dataset acquired in different location and not used at all in training and shown to detect 100% of its crack instances. We also characterise the relationship between detection performance, crack width and number of points per crack, providing a foundation upon which to make decisions about both practical deployments and future research directions.
Pasar tenaga kerja mengalami kecenderungan perubahan karena transformasi struktural ekonomi, perubahan struktur penduduk, digitalisasi, perubahan iklim, pandemi COVID-19, serta ketidakpastian ekonomi. Fleksibilitas pasar tenaga kerja dibutuhkan untuk beradaptasi perubahan tersebut. Tujuan kajian ini adalah untuk menganalisis kondisi pekerja PKWT dan alih daya dalam konteks penerapan Undang-undang Nomor 11 Tahun 2020 tentang Cipta Kerja dan turunannya. Sumber Data yang digunakan berupa data Primer dengan melakukan FGD serta wawancara mendalam dan data Sekunder dari Survei Angkatan Kerja Nasional (Sakernas), BPS dan Wajib Lapor Ketenagakerjaan Perusahaan), Kemnaker RI. Hasil studi memperlihatkan bahwa secara umum UU Cipta Kerja sudah memberikan perlindungan yang lebih baik daripada peraturan sebelumnya. Namun, di dalam pelaksanaannya masih perlu peningkatan awareness dan pengawasan yang lebih baik. Pada intinya, apabila pekerja dilindungi dalam semua bentuk kontrak apa pun, perbedaan antara pekerja PKWT-PKWTT dan alih daya tidak akan terlalu berarti.
This thesis proposes a novel automated crack detection and characterisation method on unstructured surfaces using 3D point cloud. Crack detection on unstructured surfaces poses a challenge compared to flat surfaces such as pavements and concrete, which typically utilise image-based sensors. The detection method utilises a point cloud-based deep learning method to perform point-wise classification. The detected points are then automatically characterised to estimate the detected cracks’ properties such as width profile, orientation, and length. The proposed method enables the deployment of autonomous systems to conduct reliable surveys in environments risky to humans.
Deep learning is a sub-field of artificial intelligence that combines feature engineering and classification in one method. It is a data-driven technique that optimises a predictive model via learning from a large dataset. Digitisation in industry has included acquisition and storage of a variety of large datasets for interpretation and decision making. This has led to the adoption of deep learning in different industries, such as transportation, manufacturing, medicine and agriculture. However, in the mining industry, the adoption and development of new technologies, including deep learning methods, has not progressed at the same rate as in other industries. Nevertheless, in the past 5 years, applications of deep learning have been increasing in the mining research space. Deep learning has been implemented to solve a variety of problems related to mine exploration, ore and metal extraction and reclamation processes. The increased automation adoption in mining provides an avenue for wider application of deep learning as an element within a mine automation framework. This work provides a compact, comprehensive review of deep learning implementations in mining-related applications. The trends of these implementations in terms of years, venues, deep learning network types, tasks and general implementation, categorised by the value chain operations of exploration, extraction and reclamation are outlined. The review enables shortcomings regarding progress within the research context to be highlighted such as the proprietary nature of data, small datasets (tens to thousands of data points) limited to single operations with unique geology, mine design and equipment, lack of large scale publicly available mining related datasets and limited sensor types leading to the majority of applications being image-based analysis. Gaps identified for future research and application includes the usage of a wider range of sensor data, improved understanding of the outputs by mining practitioners, adversarial testing of the deep learning models, development of public datasets covering the extensive range of conditions experienced in mines.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.