An automatic cleaning, labeling, and scanning device for rock and mineral specimens and its control method.

Abstract

This application relates to an automated cleaning, labeling, and scanning device and its control method for rock and mineral specimens. The device mainly includes a control system and, sequentially arranged along the conveying direction, a three-dimensional image scanning and recognition module, a directional spray cleaning module, a drying module, and a labeling and spectral imaging module. The three-dimensional image scanning and recognition module includes a first image scanner; the directional spray cleaning module includes a cleaning tank; the drying module is used to dry the cleaned specimens by blowing air; and the labeling and spectral imaging module includes a second image scanner, an automatic rotating arm, and a third image scanner. This application employs a method that deeply integrates three-dimensional visual perception technology with automated production line execution mechanisms, achieving a comprehensive effect of adaptive and precise cleaning of rock and mineral specimens with arbitrary irregular shapes, automatic facet finding and standardized labeling, and multimodal digital archiving, eliminating the need for manual intervention.

Description

Technical Field

[0001] This application relates to the field of rock and mineral collection technology, and in particular to an automatic cleaning, labeling, and scanning device for rock and mineral specimens and its control method. Background Technology

[0002] Rock and mineral specimens are important carriers for geological teaching, scientific research experiments, and popular science displays. Through observation, touch, and comparison of typical rock and mineral samples, the color, structure, mineral assemblage, and genetic types of different rock types can be visually demonstrated. They are indispensable basic data for basic geological education and public science communication. Currently, geology departments in universities, geological research institutes, natural history museums, and other institutions collect a large number of field rock and mineral specimens every year. The surfaces of specimens collected in the field are often covered with a large amount of rock debris, mud, and weathering products, requiring multiple steps such as cleaning, drying, numbering and labeling, photographing, and information registration before they can be used.

[0003] Existing technologies for handling the above processes are highly fragmented and have a very low degree of automation, heavily relying on manual labor. Because natural rock and mineral specimens have highly irregular three-dimensional shapes and varying degrees of surface contamination, traditional automated equipment can only perform indiscriminate global rinsing and single-stage air blowing. This not only results in poor cleaning effectiveness and easily damages fragile specimens, but also means that in the subsequent labeling stage, mechanical equipment cannot automatically find suitable flat areas for writing on irregular rocks. This makes tedious manual review and handwritten labeling indispensable, ultimately leading to low specimen processing efficiency and the acquisition of two-dimensional photographic information that is insufficient to meet the archiving requirements of modern digital specimen libraries for three-dimensional morphology and spectral characteristics.

[0004] Given the above situation, how to solve the technical problem of the extremely fragmented process and low degree of automation when processing field rock and mineral specimens with irregular three-dimensional morphology, which relies heavily on manual labor, is a problem to be solved in this technical field. Summary of the Invention

[0005] In view of the above-mentioned defects or improvement needs of the existing technology, this application provides an automatic cleaning, labeling and scanning device and its control method for rock and mineral specimens. By deeply linking three-dimensional vision perception technology with automated production line execution mechanism, it achieves the comprehensive effect of adaptive and precise cleaning of rock and mineral specimens with arbitrary irregular shape, automatic surface finding and standardized writing of labels, and multi-modal digital archiving, thus eliminating the need for manual intervention.

[0006] The embodiments of this application adopt the following technical solutions: In the first aspect, this application provides an automatic cleaning, labeling and scanning device for rock and mineral specimens, including a control system and a three-dimensional image scanning and recognition module, a directional spray cleaning module, an air drying module and a labeling and spectral imaging module arranged sequentially along the conveying direction; The three-dimensional image scanning and recognition module includes a first image scanner, used to acquire the initial three-dimensional spatial image of the specimen and identify the three-dimensional coordinates and surface normal of the contaminant area on the specimen surface; The directional spray cleaning module includes a cleaning tank, which is equipped with at least one nozzle with adjustable spray angle and pressure. The control system controls the nozzle to perform directional spray cleaning on the contaminated area based on the three-dimensional coordinates and surface normal of the contaminated area. The air-drying module is used to dry the cleaned specimens by blowing air. The annotation and spectral imaging module includes a second image scanner, an automatic rotating arm, and a third image scanner. The second image scanner is used to acquire the outline and curvature of each visible surface of the dried specimen to select the target annotation surface. The automatic rotating arm has a writing head at its end. The control system converts the spatial coordinates of the target annotation surface into joint rotation commands to drive the writing head to complete the annotation on the target annotation surface. The third image scanner is used to perform spectral scanning imaging on the annotated specimen.

[0007] By adopting the above technical solution, the first image scanner is used for pre-positioned three-dimensional perception and linked with the cleaning box for directional spraying. At the same time, the second image scanner is used to automatically find the surface and link with the robotic arm for annotation. Finally, the third image scanner is used to perform spectral scanning imaging on the annotated specimens and store them in the database. This effectively solves the technical problem that traditional methods cannot automatically adapt to irregular rock morphologies. It realizes the unmanned and integrated processing of field rock specimens from a dirty state to thorough cleaning, standardized numbering, and digital scanning and storage, which greatly improves processing efficiency and standardization.

[0008] In some embodiments, when controlling the nozzle to perform directional spraying, the control system calculates the angle between the nozzle spraying direction and the normal to the surface of the dirt, controls the sprayed water flow direction to be consistent with the normal to the surface of the dirt or within a preset deviation threshold; and adaptively adjusts the spraying time and spraying pressure of the nozzle according to the dirt coverage area and dirt thickness indicators identified by the first image scanner.

[0009] By adopting the above technical solution, controlling the nozzle to be aligned with the surface of the dirt and dynamically adjusting the pressure and time according to the thickness of the dirt, the impact force of the high-pressure water flow can be precisely concentrated on the attached mud and dirt. This significantly improves the efficiency of removing stubborn mud, while minimizing the consumption of clean water and avoiding excessive rinsing damage to the dirt-free areas of the specimen.

[0010] In some embodiments, a high-frequency vibrating conveyor belt is provided at the bottom of the cleaning tank. The high-frequency vibrating conveyor belt performs high-frequency vibration as it supports and transports the specimen through the cleaning tank, so as to accelerate the removal of dirt and grime from the specimen surface and the discharge of surface water film.

[0011] By adopting the above technical solution and introducing a high-frequency vibrating conveyor belt, the mechanical high-frequency oscillation breaks the adhesion stress between the dirt and the rock surface. This, combined with the directional high-pressure spraying, forms a dual effect of physical oscillation and hydraulic stripping, which greatly accelerates the cleaning process and helps to quickly drain surface water, reducing the pressure of subsequent air drying.

[0012] In some embodiments, the air-drying module includes an air-drying box and a lifting intelligent air knife disposed inside the air-drying box. The control system acquires the three-dimensional contour model of the specimen established by the first image scanner and controls the lifting intelligent air knife to move above the specimen, performing contour-following lifting and lowering along the three-dimensional contour of the specimen, so that the air outlet of the lifting intelligent air knife maintains a dynamic constant distance from the specimen surface.

[0013] By adopting the above technical solution, the lifting intelligent air knife can move up and down according to the three-dimensional contour of the specimen, ensuring that the shear stress of the strong wind on the irregular rock surface is always in the best state. This avoids the problem of slow drying caused by the traditional fixed air outlet being too far away from some rock surfaces, and significantly improves the drying efficiency.

[0014] In some embodiments, the annotation and spectral imaging module further includes an annotation worktable for placing and transporting specimens. The control system acquires multi-angle images of the specimens through a second image scanner, calculates the flatness and area of ​​each visible surface of the specimen using edge detection and curvature analysis algorithms, removes the bottom surface in contact with the annotation worktable, and selects the visible surface with flatness greater than a preset flatness threshold and the largest area as the target annotation surface.

[0015] By adopting the above technical solution, the optimal target annotation surface is automatically calculated and selected on irregular three-dimensional specimens using curvature and area algorithms, ensuring that the specimen numbers written by the robotic arm are flat, clear, and continuous, and are not easily worn during long-term preservation.

[0016] In some embodiments, the labeling worktable is provided with an adaptive posture adjustment mechanism, which includes a multi-axis rotating tray and a flexible clamping member; after the target labeling surface is selected, the control system controls the multi-axis rotating tray to rotate so that the target labeling surface is rotated to a horizontally upward posture, and controls the flexible clamping member to fix the specimen from the side so that the automatic rotating arm can perform writing and labeling.

[0017] By adopting the above technical solution, the adaptive posture adjustment mechanism can rotate the selected tilted target marking surface to a horizontal position and hold it firmly, avoiding the problem of the rock slipping or rolling when writing under force, and greatly improving the neatness of the handwriting and the accuracy of the strokes when the automatic rotating arm is writing.

[0018] In some embodiments, along the conveying direction, a feeding and conveying module for carrying and transporting the specimen to be processed is provided at the three-dimensional image scanning and recognition module, and a discharging and storage module for carrying and transporting the labeled specimen is provided behind the labeling and spectral imaging module; the feeding and conveying module, the three-dimensional image scanning and recognition module, the directional spray cleaning module, the air drying module, the labeling and spectral imaging module, and the discharging and storage module are arranged in a U-shape; the device also includes an environmentally friendly water circulation module, which includes a water tank and a sedimentation tank that are connected to the bottom of the cleaning tank and the air drying module, and the supernatant outlet of the sedimentation tank is connected to the nozzle through a water pump.

[0019] By adopting the above technical solutions, the U-shaped production line layout significantly shortens the longitudinal footprint of the equipment, making it easier to deploy in laboratories or specimen banks with limited space. At the same time, the environmentally friendly recycling system composed of water tanks and sedimentation tanks purifies and reuses the wastewater that washes away the mud, significantly reducing the water consumption and wastewater discharge pressure of the equipment, and possessing excellent green and environmentally friendly attributes.

[0020] Secondly, this application provides a control method for an automatic cleaning, labeling, and scanning device for rock and mineral specimens, applied to the automatic cleaning, labeling, and scanning device for rock and mineral specimens described in the first aspect. When controlling the directional spray cleaning module, the control system executes an adaptive anti-damage cleaning control strategy. The control system calculates the curvature abruptness rate and edge thickness of the specimen surface based on the three-dimensional spatial image acquired by the first image scanner, and establishes a spatial vulnerability distribution model of the specimen. Based on the color depth and volume difference of the three-dimensional spatial image, the sludge adhesion thickness in the contaminated area is estimated. The control system integrates the vulnerability distribution model with the sludge adhesion thickness to dynamically generate nozzle control commands. When the nozzle sweeps over an edge area where the vulnerability is higher than the preset threshold, it automatically reduces the water pump pressure and switches the constant water flow to a high-frequency atomization pulse mode. When it sweeps over a sludge area where the thickness is higher than the preset threshold, it increases the water pump pressure and controls the nozzle to perform spiral micro-motion cross-spraying along the sludge normal.

[0021] By adopting the above technical solution, it is possible to accurately identify the fragile edge areas and sludge adhesion of rock and mineral specimens based on three-dimensional images, and adaptively adjust the spray pressure and spray mode by zone. Low-pressure atomized pulse is used to gently clean highly fragile edges to avoid specimen damage and edge breakage, while pressurized spiral cross-spray is used to efficiently remove dirt from thick dirt areas. While effectively removing dirt from the surface of rock and minerals, the risk of specimen edge damage is significantly reduced, achieving non-destructive, efficient, and differentiated adaptive directional cleaning.

[0022] In some implementations, the control system executes a force-potential hybrid control strategy with microscopic topological compensation when controlling the annotation and spectral imaging modules. After the control system selects the target annotation surface through the second image scanner, it extracts the microscopic three-dimensional topological undulation map of the target annotation surface. The control system generates a two-dimensional writing trajectory for the target annotation surface and overlays the microscopic three-dimensional topological undulation map onto the two-dimensional writing trajectory to generate a three-dimensional writing instruction that includes dynamic compensation for the Z-axis. The three-dimensional writing instruction includes joint rotation instructions. During the process of the automatic rotating arm executing three-dimensional writing instructions, the control system obtains the current feedback value of each joint of the automatic rotating arm in real time to estimate the end contact force and executes the impedance control algorithm. When the contact force is detected to increase instantaneously due to the hard protrusion of the rock, it actively generates a compliant yielding displacement in the Z-axis direction to ensure the uniformity and continuity of ink adhesion when the pen tip writes on the rough rock surface.

[0023] By adopting the above technical solution, combining the microscopic three-dimensional topological undulations of the annotation surface for Z-axis dynamic trajectory compensation, and performing real-time force feedback impedance compliance control during the writing process, the annotation automatically and smoothly yields when encountering hard protrusions in the rock, effectively solving the problems of ink breakage, uneven handwriting, and pen tip scratching of specimens on rough and uneven rock surfaces, ensuring continuous and uniform handwriting and accurate and reliable position of automated annotation, and improving the quality and stability of automated annotation of rock specimens.

[0024] In some embodiments, the control system further includes a closed-loop evaluation feedback mechanism for cleaning quality based on multispectral features: After acquiring the multispectral image of the labeled specimen, the control system uses the multispectral image features to invert the material composition of the specimen surface. If residual clay mineral spectral characteristics exceeding the area threshold are detected in a non-depression area of ​​the specimen, it is determined that the cleaning is substandard. The control system feeds back the specimen's contour data and residual location coordinates as negative samples to the directional spray cleaning module. Through machine learning algorithms, the water pressure reference value and spray residence time parameters for subsequent cleaning of similar rocks and minerals are automatically adjusted to achieve self-supervised optimization of the parameters of the entire cleaning process.

[0025] By adopting the above technical solution, multispectral features are used to accurately identify clay mineral residues that are invisible to the naked eye, enabling quantitative closed-loop evaluation of cleaning quality. Data on unqualified samples are fed back to the cleaning system for machine learning self-optimization, automatically iteratively adjusting parameters such as water pressure and spraying time. This allows for continuous improvement of the cleaning effect on similar samples without manual intervention, achieving self-monitoring, self-adaptation, high-quality, and stable operation of the cleaning production line.

[0026] In summary, this application includes at least the following beneficial technical effects: 1. The first image scanner is used for pre-positioned three-dimensional perception and linked with the cleaning box for directional spraying. At the same time, the second image scanner is used to automatically find the surface and link with the robotic arm for annotation. Finally, the third image scanner is used to perform spectral scanning imaging on the annotated specimen and put it into the database. This effectively solves the technical problem that traditional methods cannot automatically adapt to irregular rock morphology. It realizes the unmanned and integrated processing of field rock specimens from dirty state to thorough cleaning, standardized numbering and digital scanning and storage, which greatly improves processing efficiency and standardization. 2. By controlling the nozzle to align with the surface of the dirt and dynamically adjusting the pressure and time according to the thickness of the dirt, the impact force of the high-pressure water flow can be precisely concentrated on the attached mud and dirt. This significantly improves the efficiency of removing stubborn mud while minimizing the consumption of clean water and avoiding excessive rinsing damage to the dirt-free areas of the specimen. 3. A high-frequency vibrating conveyor belt is introduced, which uses mechanical high-frequency vibration to break the adhesion stress between the dirt and the rock surface. This, combined with directional high-pressure spraying, forms a dual effect of physical vibration and hydraulic stripping, which greatly accelerates the cleaning process and helps to quickly drain surface water, reducing the pressure of subsequent air drying. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly described below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A top view of the overall production line layout of an automatic cleaning, labeling, and scanning device for rock and mineral specimens provided in this application embodiment; Figure 2 This is a schematic diagram of the internal operation of the directional spray cleaning module provided in the embodiments of this application; Figure 3 This is a schematic diagram of the internal workings of the air-drying module provided in an embodiment of this application; Figure 4 This is a schematic diagram of the internal workings of the annotation and spectral imaging module provided in the embodiments of this application; Figure 5 This is a data flow logic block diagram of the control system provided in an embodiment of this application.

[0029] In the diagram: 11. Feeding and conveying module; 12. 3D image scanning and recognition module; 13. Directional spray cleaning module; 14. Drying module; 15. Labeling and spectral imaging module; 16. Discharge and storage module; 17. Conveyor belt; 181. Water tank; 182. Sedimentation tank; 183. Water pump; 121. First image scanner; 131. Cleaning box; 132. Spray nozzle; 133. High-frequency vibrating conveyor belt; 134. Eccentric wheel vibrating motor; 141. Drying box. ; 142. Lifting intelligent air knife; 143. Guide rail; 144. Bottom hollow structure; 151. Second image scanner; 152. Third image scanner; 153. Top slide rail; 154. Automatic rotating arm; 155. Writing head; 156. Multi-axis rotating tray; 157. Flexible clamping component; 158. Universal ball hinge; 20. Specimen; 21. Dirt; 22. Target marking surface; A. Direction of water jet; B. Normal direction of dirt surface; C. Contour lifting trajectory. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the technical features involved in the various embodiments described below can be combined with each other as long as they do not conflict with each other.

[0031] The present application will now be described in detail with reference to the accompanying drawings and embodiments. Example 1

[0032] like Figure 1 As shown, Embodiment 1 of this application provides an automatic cleaning, labeling, and scanning device for rock and mineral specimens, including a control system and a three-dimensional image scanning and recognition module 12, a directional spray cleaning module 13, a drying module 14, and a labeling and spectral imaging module 15 arranged sequentially along the conveying direction. The device is suitable for rock and mineral specimens with dimensions ranging from 5cm×5cm×5cm to 30cm×30cm×30cm. Furthermore, along the conveying direction, a feeding and conveying module 11 for carrying and transporting the specimens to be processed is provided at the three-dimensional image scanning and recognition module 12, and a discharging and storage module 16 for carrying and transporting the labeled specimens is provided behind the labeling and spectral imaging module 15. The overall structure of the device is arranged in a U-shape.

[0033] It should be noted that traditional linear specimen processing lines occupy a large area, while geological laboratories or specimen repositories typically have limited vertical space. Furthermore, rock washing consumes a huge amount of water and easily generates large amounts of muddy wastewater, which, if directly discharged, can lead to environmental pollution and pipe blockage. In this embodiment, the feeding and conveying module 11, the three-dimensional image scanning and recognition module 12, the directional spray cleaning module 13, the air-drying module 14, the labeling and spectral imaging module 15, and the discharging and storage module 16 are connected in a U-shape via a conveyor belt 17. Simultaneously, the device also includes an environmentally friendly water circulation module, which includes a water tank 181 and a sedimentation tank 182 that are connected to the bottom of the directional spray cleaning module 13 and the air-drying module 14. The supernatant outlet of the sedimentation tank 182 is recirculated to the nozzle 132 of the directional spray cleaning module 13 via a water pump 183. With the above setup, the U-shaped layout allows operators to complete feeding and unloading in the same area, greatly saving laboratory space; the environmentally friendly water circulation module collects wastewater by gravity at the bottom and performs multi-stage sedimentation, then pumps the supernatant back into the nozzle 132, saving more than 80% of cleaning water while avoiding silt clogging the sewer.

[0034] In some embodiments, the feeding and conveying module 11 includes a feeding conveyor belt for carrying the rock specimen to be processed, and a three-dimensional image scanning and recognition module 12 can be fixed about 30cm above the conveyor belt for image acquisition and three-dimensional reconstruction of the specimen 20 passing by.

[0035] In some embodiments, the three-dimensional image scanning and recognition module 12 includes a first image scanner 121. In one specific embodiment, the first image scanner 121 may employ an industrial-grade RGB-D depth camera or a line laser 3D scanner, with a built-in image recognition program capable of identifying the spatial location of debris 21 such as rock fragments and mud on the surface of the specimen 20, and establishing a three-dimensional spatial image of the specimen 20 surface. The recognition program extracts mud and attached areas based on color features, texture features, and grayscale differences, and outputs the three-dimensional coordinates and normal information of these areas in the device coordinate system. Based on the above settings, the first image scanner 121 is used to acquire the initial three-dimensional spatial image of the specimen 20, obtain the three-dimensional point cloud data of the specimen 20 surface, and, combined with image recognition algorithms (including threshold segmentation, texture analysis, and region growing, etc.), automatically identify the mud, attached rock fragments, and other debris areas on the surface of the specimen 20. The recognition program can project the debris areas into the device coordinate system to obtain the three-dimensional spatial coordinates (x, y, z) and surface normal direction of each debris area.

[0036] In some embodiments, the directional spray cleaning module 13 includes a cleaning tank 131, which is a closed or semi-closed box formed by wrapping a plastic sheet. Its internal dimensions are approximately 80cm long, 50cm wide, and 60cm high. The cleaning tank 131 contains at least one (e.g., 5-8) high-pressure nozzles 132 with adjustable spray angle and pressure (e.g., high-pressure water jet nozzles mounted on a two-axis gimbal). It should be noted that traditional cleaning devices blindly spray the entire surface, resulting in significant water waste. Furthermore, the fixed-angle water flow is easily refracted when hitting the inclined mud, greatly reducing the cleaning impact. In this embodiment, when the control system controls the nozzle 132 to perform directional spraying, it selects one or more high-pressure nozzles 132 as target nozzles 132 based on the location and quantity of the dirt 21, calculates the angle between the spray direction of the nozzle 132 and the normal to the surface of the dirt 21, and drives the nozzle 132 to deflect via a motor, controlling the direction of the sprayed water flow to remain consistent with or within a preset deviation threshold (e.g., ±5 degrees) of the normal to the surface of the dirt 21. Figure 2 The dashed line A indicates the spray direction of nozzle 132, and the dashed line B indicates the normal direction of the surface of the dirt 21. During cleaning, dashed lines A and B need to coincide or be within a preset deviation threshold. Furthermore, the control system adaptively adjusts the spraying time of nozzle 132 and the spraying pressure provided by water pump 183 based on the dirt 21 coverage area and dirt 21 thickness indicators identified by the first image scanner 121. For example, it increases the spraying time and spraying pressure for dirt 21 with large area and thickness, and decreases the spraying time and spraying pressure for dirt 21 with small area and thickness (such as a small amount of dust or fine adhering matter). The above scheme ensures that the high-pressure water flow always hits the core of dirt 21 directly with a vertical tangent (i.e., normal direction), maximizing kinetic energy utilization, multiplying the efficiency of removing stubborn dirt, and ensuring cleaning effect while avoiding unnecessary erosion of the rock specimen.

[0037] To further improve the cleaning effect, a high-frequency vibrating conveyor belt 133 is provided at the bottom of the cleaning tank 131. It should be noted that some clay-type dirt 21 is extremely tightly bound to the pores of the rock surface, and it is difficult to completely remove it by water flow alone. In this embodiment, the high-frequency vibrating conveyor belt 133, driven by the eccentric wheel vibration motor 134 below, performs high-frequency vibration (e.g., 50-100Hz) while supporting and transporting the specimen 20 through the cleaning tank 131, thereby forming a dual peeling effect of physical vibration and water impact. The high-frequency vibration can quickly cut off the adhesion stress between the dirt and the micropores of the rock, accelerate the shedding of dirt 21, and at the same time accelerate the rupture and discharge of the surface water film.

[0038] In some embodiments, the air-drying module 14 is used to dry the cleaned specimen 20 by blowing air. It includes an air-drying box 141 with an internal cavity approximately 160cm long, 50cm wide, and 60cm high. The air-drying box 141 has multiple air inlets and outlets around its perimeter. The air inlets are connected to a warm air blower, and the air outlets are used to expel hot and humid air. The bottom of the air-drying box 141 has a hollow structure 144 to facilitate the drainage of water. A water tank 181 is provided at the bottom of the box to collect dripping water, which can be recycled after sedimentation. The specimen 20 moves with a conveyor belt inside the air-drying box 141, and the high-speed inflow of warm air blows on all sides of the specimen 20 to achieve rapid air drying.

[0039] Optionally, in some embodiments, a lifting intelligent air knife 142 is installed inside the drying box 141. The lifting intelligent air knife 142 is movable left and right inside the drying box 141 via a guide rail 143. It should be noted that rocks in the wild have extremely varied shapes and heights. Traditional fixed-height warm air blowers, if set too high, will not be effective at drying low rocks; if set too low, they may bump into taller rock specimens. (Reference) Figure 3 As shown, in this embodiment, the control system acquires the three-dimensional contour model of the specimen 20 established by the first image scanner 121, and controls the lifting intelligent air knife 142 to move above the specimen 20, performing contour-following lifting along the three-dimensional contour of the specimen 20 (for example, driving the air knife to move up and down via a lead screw guide rail). Figure 3 As shown by the dashed line C, the air outlet of the lifting intelligent air knife 142 maintains a dynamic and constant distance from the surface of the specimen 20 (e.g., maintaining an optimal blowing distance of 3-5 cm). This design perfectly balances safety and drying efficiency, ensuring that the specimen can receive strong winds to sweep away water regardless of the rock's elevation, thus avoiding drying dead zones.

[0040] In some embodiments, the annotation and spectral imaging module 15 includes a second image scanner 151, an automatic rotating arm 154, a third image scanner 152, and an annotation worktable. The air-dried specimen 20 enters the annotation and spectral imaging module 15 via a conveyor belt 17. The module's overall dimensions are approximately 50cm high × 50cm long × 40cm wide. Its interior bottom is equipped with a thick, soft rubber tray with high friction, serving as an annotation worktable to fix the specimen 20 and provide short-distance transport capability. The tray automatically stops after the specimen 20 reaches the center of the device. The top of the module is equipped with a second image scanner 151 and a third image scanner 152. The second image scanner 151 has automatic image recognition capability and is used to take top-down and multi-angle pictures of the specimen 20 placed on the rubber tray. Through edge detection, curvature analysis and plane fitting algorithms, it calculates the flatness and area of ​​each visible surface of the specimen 20, and selects the surface with the largest flatness and area after removing the bottom as the target annotation surface 22. The module is equipped with an automatic rotating arm 154, which has multi-degree-of-freedom joints and an end-effector writing head 155, and can achieve fixed-point writing at any preset position in three-dimensional space. The control system calculates the target point coordinates required by the rotating arm based on the spatial position and posture of the annotation surface determined by the second image scanner 151. The automatic rotating arm 154 can be converted into the rotation angle command of each joint through inverse kinematics solution, driving the end-effector writing head 155 to write the specimen 20 number at the preset position on the annotation surface. The third image scanner 152 has the capabilities of taking pictures and spectral scanning, including a visible light imaging module and a spectral imaging module. By moving along the top slide rail 153 of the device and / or coordinating with the micro-rotation of the specimen 20, it can achieve multi-angle and multi-directional imaging and spectral scanning of the specimen 20. The spectral imaging module can collect the reflectance spectral information of the rock surface in multiple bands, with typical bands covering the visible to near-infrared range. The control system automatically associates the spectral imaging data with the specimen 20 number and stores them uniformly in the database. Specifically, the control system can register the multi-view images with the three-dimensional model of the specimen 20 constructed in the first image scanner 121, mapping the spectral data onto the surface of the three-dimensional model, thereby forming a three-dimensional spectral model with geometric shape and spectral features. At the same time, using the specimen 20 number as an index, the system uniformly stores information such as images, spectra, three-dimensional models, and cleaning parameters in the database.

[0041] It should be noted that when automated equipment writes on irregular rocks, the rough and uneven surfaces make it difficult for the robotic arm to find the flattest spot to start writing, which can easily lead to distorted and illegible numbers. In this embodiment, the control system acquires multi-angle images of the dried specimen 20 through the second image scanner 151, calculates the flatness and area of ​​each visible surface of the specimen 20 using edge detection and curvature analysis algorithms, removes the bottom surface in contact with the labeling platform, and selects the visible surface with a flatness greater than a preset flatness threshold and the largest area as the target labeling surface 22. This solution achieves intelligent surface finding using machine vision, fundamentally solving the problem of difficult-to-write standardized characters on irregular surfaces and ensuring the clarity and readability of the labels.

[0042] Furthermore, after finding a flat surface, the issue of maintaining the writing posture still needs to be addressed. It should be noted that even if a flat marking surface is found, it is often tilted, and the rock is prone to slipping and rolling when pressed by the robotic arm, leading to broken strokes or misaligned writing. (Reference) Figure 4 As shown, in some embodiments, the annotation worktable is equipped with an adaptive posture adjustment mechanism, which includes a multi-axis rotating tray 156 and a flexible clamping member 157. After selecting the target annotation surface 22, the control system controls the multi-axis rotating tray 156 to rotate (e.g., driven by a universal ball hinge 158), causing the target annotation surface 22 to rotate to an absolutely horizontal upward posture, and controls the pneumatically driven flexible clamping member 157 (such as a silicone gripper) to wrap around and fix the specimen 20 from the side. In this way, the optimal writing surface of any spatial posture is forcibly converted into a standard two-dimensional horizontal working surface, and sufficient clamping rigidity is provided, providing an extremely stable physical platform for the high-precision writing of the automatic rotating arm 154.

[0043] Furthermore, after the specimen 20 is labeled, the third image scanner 152 (such as a hyperspectral camera) is activated to perform multispectral scanning imaging on the labeled specimen, extract the spectral characteristics of the rock's material composition, fuse them with the three-dimensional contour data, and finally output a complete digital twin file for storage. Finally, after labeling and scanning are completed, the rubber tray of the labeling and spectral imaging module 15 restarts, sending the specimen 20 to the exit position, where it is transported to the preset storage area via the discharge conveyor belt of the discharge and storage module 16.

[0044] Through the above solution, this application has at least the following effects: 1. High cleaning efficiency and high degree of automation: The spatial distribution information of dirt 21 is obtained by the first image scanner 121 and image recognition program. With the help of adjustable angle high pressure nozzle 132 and high frequency vibration belt, mud, rock debris and other dirt 21 can be sprayed and cleaned in a directional manner, which greatly improves cleaning efficiency and reduces water waste and manual intervention.

[0045] 2. Uniform, accurate and reliable labeling: The second image scanner 151 automatically selects a flat and large surface as the labeling surface, and the automatic rotating arm 154 writes the number at the designated position, ensuring uniform labeling position, clear font and continuous numbering, which significantly reduces problems such as missing numbers, duplicate numbers and non-standard writing caused by manual labeling.

[0046] 3. Abundant spectral imaging information provides support for digital management: The third image scanner 152 enables multi-angle visible light imaging and spectral scanning of specimen 20. Combined with 3D model and database management, it forms a complete specimen 20 information system of number, 3D morphology and spectral characteristics, which facilitates subsequent teaching demonstrations, remote sharing and intelligent identification research based on spectral characteristics.

[0047] 4. Integrated process, small footprint, and good environmental protection: The entire set of equipment is arranged in a U-shaped conveyor, integrating the cleaning, drying, labeling and scanning processes into a continuous production line, saving space and manpower; a water tank 181 and a sedimentation tank 182 are set below the cleaning and drying area to realize centralized collection and sedimentation treatment of wastewater, and the supernatant can be partially reused, reducing wastewater discharge and meeting green environmental protection requirements. Example 2

[0048] Based on the automatic cleaning, labeling, and scanning device for rock and mineral specimens provided in Embodiment 1, Embodiment 2 of this application provides a control method for the automatic cleaning, labeling, and scanning device for rock and mineral specimens. This method is executed by the aforementioned control system (e.g., a computing platform including an industrial computer, PLC, and a vision computing motherboard).

[0049] refer to Figure 5 As shown, the control system can include a three-layer architecture: a visual perception layer, a core algorithm and control layer, and a physical execution and data layer. The visual perception layer uses first, second, and third image scanners 152 to acquire three-dimensional images, multi-angle images, and multispectral data of specimen 20, respectively. The core algorithm and control layer relies on shared three-dimensional model data to execute algorithms such as dirt coordinate and normal recognition, flatness analysis and surface selection, and geometric and spectral feature registration. The physical execution and data layer, based on the algorithm output, realizes directional spray cleaning by nozzle 132, tray posture adjustment, and robotic arm reverse motion writing, ultimately generating a 3D multispectral digital twin file containing cleaning parameters and label numbers, completing the fully automated processing and closed-loop management of specimen 20.

[0050] In some implementations, the control system further includes the following intelligent control strategies: I. Adaptive Damage Prevention and Cleaning Control Strategy: The technical problem this strategy aims to solve is that many valuable geological specimens (such as shale, mica, or crystalline minerals) have very fragile edges. If high-pressure water jets are used to clean the dirt, the natural edges or crystals of the specimens could easily be broken, causing irreversible scientific damage.

[0051] Based on the above, the control system performs the following steps: Based on the three-dimensional spatial image acquired by the first image scanner 121, the control system extracts the local features of the point cloud, calculates the curvature abrupt change rate and edge thickness of the specimen 20 surface, and establishes a spatial vulnerability distribution model of the specimen 20 (for example, sharp edges with a thickness of less than 5 mm and a very high curvature abrupt change rate are marked as high vulnerability areas).

[0052] The thickness of sludge adhesion in contaminated areas is estimated based on the color depth and volume difference of three-dimensional spatial images (compared with a historical clean rock sample library).

[0053] The control system integrates the above-mentioned distribution model and dynamically generates control commands for the nozzle 132: when the nozzle 132 sweeps over an edge area with a fragility higher than a preset threshold, it automatically reduces the pressure of the water pump 183 and switches the constant water flow to a high-frequency atomization pulse mode (using a gentle pulse water mist instead of a high-pressure water column); when it sweeps over a stubborn dirt area with a thickness higher than a preset threshold, it increases the pressure of the water pump 183 and controls the two-axis motor to drive the nozzle 132 to perform spiral micro-motion cross-spraying along the normal direction of the dirt 21.

[0054] The above strategy endows the cleaning device with targeted cleaning intelligence, ensuring that stubborn sludge is completely removed while greatly avoiding hydraulic cutting and damage to the fragile structure of the rock specimen.

[0055] II. Deep Pit Drying Control Strategy Based on Gaussian Curvature: The technical problem this strategy aims to solve is that when there are deep pits on the rock surface, if the wind knife keeps blowing vertically downwards, the strong wind will create local pressure standing waves (vortices) at the bottom of the pit, causing the water at the bottom of the pit to swirl inside and not be blown out.

[0056] Based on the above, when controlling the lifting intelligent air knife 142, the control system extracts the local Gaussian curvature of the three-dimensional contour of the specimen 20; when it is determined that there is a deep pit feature in the current drying area with a depth greater than a preset threshold, the control system controls the lifting intelligent air knife 142 to tilt at a yaw angle, so that the high-pressure air jet is injected into the inner wall of the deep pit at an oblique angle (such as an angle of 30 degrees to 45 degrees).

[0057] By employing the above strategy and utilizing the wall-attachment effect (Coanda effect) in fluid mechanics, the obliquely cut high-pressure air will form a unidirectional flow along the pit wall, forcibly removing the residual water stains at the bottom of the deep pit from the other side, thus completely eliminating the problem of water accumulation in dead corners during the drying process of irregular rocks.

[0058] III. Force-potential hybrid control strategy for micro-topology compensation: The technical problem this strategy aims to solve is that although a macroscopically smooth marking surface was found in Example 1, natural rocks are still microscopically covered with millimeter-sized gravel protrusions or tiny pits. Traditional robotic arms only write along two-dimensional planar trajectories, which inevitably leads to the pen tip being suspended in the air at pits (ink breakage) and obstructed at protrusions (pen tip breakage), seriously affecting the marking quality and equipment lifespan.

[0059] Based on the above, after selecting the target annotation surface 22 through the second image scanner 151, the control system further extracts a microscopic three-dimensional topological undulation map of the surface: that is, high-resolution three-dimensional point cloud acquisition is performed on the target annotation surface 22, and after point cloud meshing and height map generation algorithms, a height distribution matrix in pixels is obtained. Each pixel in the matrix corresponds to the two-dimensional coordinates (x, y) of the annotation surface and stores the actual height value z at that location. For example, after acquiring the granite annotation surface, the map will clearly show that there is a quartz sand protrusion with a height of +0.3mm at coordinates (x1, y1) and a tiny pit with a depth of -0.2mm at coordinates (x2, y2).

[0060] After generating a two-dimensional writing trajectory (e.g., number "R-2023-01"), a microscopic topological undulation map is overlaid to generate a three-dimensional writing instruction that includes dynamic compensation along the Z-axis (height direction). This instruction is then sent to the robotic arm, causing the pen tip height to change in real time with the trajectory: it automatically raises when encountering protrusions and automatically lowers when encountering depressions, always maintaining a constant contact distance with the rock surface.

[0061] When the automatic rotating arm 154 performs writing, the control system collects the current feedback values ​​of each joint of the robotic arm in real time to estimate the end contact force and executes the impedance control algorithm. When it detects that the contact force increases instantaneously due to an unforeseen local hard protrusion on the rock, the control system actively generates a compliant yielding displacement in the Z-axis direction to avoid damage to the pen tip or scratching the surface of the specimen 20.

[0062] The impedance control algorithm establishes a dynamic relationship between the contact force and displacement at the end of the robotic arm (similar to a spring-damped mechanical model). The control system estimates the end-effector contact force by real-time acquisition of joint current. When the contact force exceeds a preset threshold, it actively corrects the trajectory, achieving a smooth adjustment of position according to force changes, rather than forcibly executing a preset trajectory. The smooth yielding displacement refers to the reverse displacement compensation generated by the end of the robotic arm in the Z-axis direction under the impedance control algorithm, which is positively correlated with the magnitude of the contact force. When the contact force increases instantaneously (e.g., the pen tip encounters a hard protrusion), the control system actively generates a small yielding displacement in the Z-axis direction to reduce the contact force, achieving flexible contact and preventing pen tip breakage due to overload, scratches on the specimen surface, or handwriting misalignment. For example, if the pen tip encounters a 0.5mm hard protrusion not marked on the topology map, and the contact force suddenly increases, the robotic arm will automatically retract 0.5mm to maintain stable pen pressure, preventing damage to the pen or deep scratches on the stone.

[0063] Through the above strategies, true biomimetic smooth writing is achieved. The robotic arm can automatically adjust the pressure of the pen when it senses the rough resistance of the stone surface, just like a human wrist. This ensures both the uniformity of ink adhesion and the continuity of the writing, and effectively extends the service life of the robotic arm's writing end.

[0064] IV. Cleaning quality closed-loop evaluation feedback mechanism based on multispectral characteristics: The technical problem this mechanism aims to solve is that production lines typically operate in an open-loop, unidirectional manner. If a new type of ore, which is extremely difficult to clean, such as sticky red soil ore, is encountered, the equipment cannot clean it properly using the default parameters. The equipment can only wait for manual intervention at the end of the production line to discover the problem and rework it, lacking the system's self-learning and closed-loop optimization capabilities.

[0065] Based on the above, after acquiring the multispectral image of the labeled specimen, the control system uses the multispectral image features to invert the material composition of the surface of the specimen 20 (for example, there is a significant difference between the spectral characteristics of the rock body minerals and the spectral characteristics of the attached clay minerals). If the spectral characteristics of residual clay minerals exceeding the area threshold are detected in the non-natural depression area of ​​the specimen 20, it is determined that the cleaning is substandard; the control system automatically labels the contour data and residual position coordinates of the specimen 20 as negative samples and feeds them back to the directional spray cleaning module 13. Through the built-in machine learning algorithm, the water pressure benchmark value and spray residence time parameters are automatically improved when similar rocks and ores are subsequently cleaned in the warehouse.

[0066] Through this mechanism, the digital scanning results from the backend are fed back to the frontend cleaning module, giving the entire production line a self-monitoring and self-evolving AI closed-loop optimization capability. The longer the equipment runs and the more types of rocks it processes, the more accurate the matching of its cleaning parameters becomes, eliminating the reliance on manual inspection and debugging.

[0067] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An automatic cleaning, labeling, and scanning device for rock and mineral specimens, characterized in that, It includes a control system and a three-dimensional image scanning and recognition module, a directional spray cleaning module, a drying module, and a labeling and spectral imaging module arranged sequentially along the conveying direction; The three-dimensional image scanning and recognition module includes a first image scanner, used to acquire the initial three-dimensional spatial image of the specimen and identify the three-dimensional coordinates and surface normal of the contaminant area on the specimen surface; The directional spray cleaning module includes a cleaning tank, which is equipped with at least one nozzle with adjustable spray angle and pressure. The control system controls the nozzle to perform directional spray cleaning on the contaminated area based on the three-dimensional coordinates and surface normal of the contaminated area. The air-drying module is used to dry the cleaned specimens by blowing air. The annotation and spectral imaging module includes a second image scanner, an automatic rotating arm, and a third image scanner. The second image scanner is used to acquire the outline and curvature of each visible surface of the dried specimen to select the target annotation surface. The automatic rotating arm has a writing head at its end. The control system converts the spatial coordinates of the target annotation surface into joint rotation commands to drive the writing head to complete the annotation on the target annotation surface. The third image scanner is used to perform spectral scanning imaging on the annotated specimen.

2. The automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 1, characterized in that, When controlling the nozzle to perform directional spraying, the control system calculates the angle between the nozzle spray direction and the normal to the surface of the dirt, and controls the spray water flow direction to be consistent with the normal to the surface of the dirt or within a preset deviation threshold; and adaptively adjusts the spraying time and spraying pressure of the nozzle according to the dirt coverage area and dirt thickness indicators identified by the first image scanner.

3. The automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 1, characterized in that, The bottom of the cleaning tank is equipped with a high-frequency vibrating conveyor belt. The high-frequency vibrating conveyor belt performs high-frequency vibration while supporting and transporting the specimen through the cleaning tank, so as to accelerate the removal of dirt and water film from the specimen surface.

4. The automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 1, characterized in that, The air-drying module includes an air-drying box and a lifting intelligent air knife set inside the air-drying box. The control system acquires the three-dimensional contour model of the specimen established by the first image scanner and controls the lifting intelligent air knife to move above the specimen, performing contour-following lifting and lowering along the three-dimensional contour of the specimen, so that the air outlet of the lifting intelligent air knife maintains a dynamic constant distance from the specimen surface.

5. The automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 1, characterized in that, The annotation and spectral imaging module also includes an annotation worktable for placing and transporting specimens. The control system acquires multi-angle images of the specimens through a second image scanner, calculates the flatness and area of ​​each visible surface of the specimen using edge detection and curvature analysis algorithms, removes the bottom surface that is in contact with the annotation worktable, and selects the visible surface with a flatness greater than a preset flatness threshold and the largest area as the target annotation surface.

6. The automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 5, characterized in that, The labeling platform is equipped with an adaptive posture adjustment mechanism, which includes a multi-axis rotating tray and a flexible clamping component. After selecting the target labeling surface, the control system controls the multi-axis rotating tray to rotate so that the target labeling surface is rotated to a horizontally upward posture, and controls the flexible clamping component to fix the specimen from the side so that the automatic rotating arm can perform writing and labeling.

7. The automatic cleaning, labeling, and scanning device for rock and mineral specimens according to any one of claims 1-6, characterized in that, Along the conveying direction, a feeding and conveying module for carrying and transporting the specimen to be processed is provided at the three-dimensional image scanning and recognition module, and a discharging and storage module for carrying and transporting the labeled specimen is provided behind the labeling and spectral imaging module; the feeding and conveying module, the three-dimensional image scanning and recognition module, the directional spray cleaning module, the air drying module, the labeling and spectral imaging module, and the discharging and storage module are arranged in a U-shape; the device also includes an environmentally friendly water circulation module, which includes a water tank and a sedimentation tank that are connected to the bottom of the cleaning tank and the air drying module, and the supernatant outlet of the sedimentation tank is connected to the nozzle through a water pump.

8. A control method for an automatic cleaning, labeling, and scanning device for rock and mineral specimens, applied to the automatic cleaning, labeling, and scanning device for rock and mineral specimens as described in any one of claims 1-7, characterized in that, When controlling the directional spray cleaning module, the control system executes an adaptive damage prevention cleaning control strategy: The control system calculates the curvature abruptness rate and edge thickness of the specimen surface based on the three-dimensional spatial image acquired by the first image scanner, and establishes a spatial vulnerability distribution model of the specimen. Based on the color depth and volume difference of the three-dimensional spatial image, the sludge adhesion thickness in the contaminated area is estimated. The control system integrates the vulnerability distribution model with the sludge adhesion thickness to dynamically generate nozzle control commands. When the nozzle sweeps over an edge area where the vulnerability is higher than the preset threshold, it automatically reduces the water pump pressure and switches the constant water flow to a high-frequency atomization pulse mode. When it sweeps over a sludge area where the thickness is higher than the preset threshold, it increases the water pump pressure and controls the nozzle to perform spiral micro-motion cross-spraying along the sludge normal.

9. The control method for the automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 8, characterized in that, When controlling the annotation and spectral imaging modules, the control system executes a force-potential hybrid control strategy with microscopic topological compensation. After the control system selects the target annotation surface through the second image scanner, it extracts the microscopic three-dimensional topological undulation map of the target annotation surface. The control system generates a two-dimensional writing trajectory for the target annotation surface and overlays the microscopic three-dimensional topological undulation map onto the two-dimensional writing trajectory to generate a three-dimensional writing instruction that includes dynamic compensation for the Z-axis. The three-dimensional writing instruction includes joint rotation instructions. During the process of the automatic rotating arm executing three-dimensional writing instructions, the control system obtains the current feedback value of each joint of the automatic rotating arm in real time to estimate the end contact force and executes the impedance control algorithm. When the contact force is detected to increase instantaneously due to the hard protrusion of the rock, it actively generates a compliant yielding displacement in the Z-axis direction to ensure the uniformity and continuity of ink adhesion when the pen tip writes on the rough rock surface.

10. The control method for the automatic cleaning, labeling, and scanning device for rock and mineral specimens according to claim 8, characterized in that, The control system also includes a closed-loop evaluation feedback mechanism for cleaning quality based on multispectral features: After acquiring the multispectral image of the labeled specimen, the control system uses the multispectral image features to invert the material composition of the specimen surface. If residual clay mineral spectral characteristics exceeding the area threshold are detected in a non-depression area of ​​the specimen, it is determined that the cleaning is substandard. The control system feeds back the specimen's contour data and residual location coordinates as negative samples to the directional spray cleaning module. Through machine learning algorithms, the water pressure reference value and spray residence time parameters for subsequent cleaning of similar rocks and minerals are automatically adjusted to achieve self-supervised optimization of the parameters of the entire cleaning process.

Images