A flexible adaptive intelligent tool cabinet based on visual recognition and a use method thereof

Abstract

The application relates to the technical field of intelligent management of hydropower station maintenance tools, and discloses a flexible self-adaptive intelligent tool cabinet based on visual identification and a use method, aiming to solve the problems of high dependence on manual operation, poor adaptability and low efficiency of traditional tool management. The tool cabinet comprises a sheet metal shell and a welded framework, the welded framework is separated into a middle stacker running channel and double-sided storage areas, the storage areas are divided into small parts, large parts and tool box areas, and are matched with an industrial computer, a visual detection component, a stacker component and a box locking component; the use method comprises identity verification, instruction matching, fork self-adaptive adjustment, material box transfer, visual verification and inventory closed-loop management. The application realizes flexible taking and placing of tools of multiple specifications, reduces manual intervention, realizes tool full-life-cycle traceability, and meets the tool management requirements of hydropower station maintenance.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for hydropower station maintenance tools, specifically to a flexible adaptive intelligent tool cabinet based on visual recognition and its usage method. Background Technology

[0002] With the rapid development of my country's hydropower industry, the demand for operation and maintenance of hydro-turbine generator units is increasing. The types and sizes of tools used at maintenance sites are numerous, and efficient tool management directly affects the overall efficiency of unit maintenance. Currently, tool storage management at hydropower stations still relies primarily on traditional manual methods, using single-size tool cabinets to store tools of different sizes, which has many drawbacks: Firstly, the handling of tools relies entirely on manual labor, which is labor-intensive and can easily lead to tool loss or damage, creating potential safety hazards. Secondly, traditional tool cabinets lack sophisticated zoned storage design, and existing stacker crane-style smart cabinets can only pick up and put in material frames of the same size, which cannot adapt to the storage needs of tools of different sizes, resulting in high space waste and long time to find and put in tools. Third, the lack of automated tools for identification and registration means that the entry and exit of goods rely on manual ledgers, which is prone to omissions and errors, resulting in inaccurate inventory data. Fourth, there is no access control or full-process traceability mechanism. There are no clear records of tool issuance and return. Overdue issuance or loss cannot be traced. Inventory management is also lax. Summary of the Invention

[0003] To address the aforementioned technical pain points in the management of tools for hydro turbine maintenance, there is an urgent need to develop a highly intelligent and adaptable tool storage device and usage method. This would enable flexible and adaptive retrieval and placement of tools of different sizes, automated registration via visual recognition, and closed-loop inventory management. Simultaneously, it would reduce manual intervention, improve tool storage and retrieval efficiency, and ensure the safety and efficiency of tool management at hydropower station maintenance sites.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by this invention is as follows: A flexible adaptive intelligent tool cabinet based on visual recognition includes a cabinet body, an industrial control computer, and an operation control system. The cabinet body includes a sheet metal shell and a welded frame. The welded frame divides the internal space of the cabinet into a central stacker crane operating channel and storage areas symmetrically distributed on both sides of the operating channel. The storage area is divided into a toolbox area, a small tools area, and a large tools area; The intelligent tool cabinet is also equipped with a vision detection component for automatic tool identification and registration, and a stacker crane component for adaptive retrieval and placement of tool boxes; Stacker crane components include stacking forks, fork adjustment components, stacking lifting components, and stacking translation components; both the large tool area and the toolbox area are equipped with box locking components; The above-mentioned tool cabinet is used in accordance with the following steps: S1. The operator completes the identity verification through the operation control system. After the system verification is successful, the operation permission is unlocked, an operation log with a global timestamp is created, and the tool basic database, storage location mapping database and stacker crane dynamic operation parameter database pre-stored in the industrial control computer are retrieved. S2. When the operator initiates a tool requisition or return instruction, the system automatically matches the corresponding toolbox type and storage location coordinates according to the instruction, and parses and obtains the size topology parameters and load threshold of the target material box in real time. S3. Based on the size and topology parameters of the target material box, the system controls the fork adjustment component to adaptively adjust the spacing between the two sets of stacking forks to the optimal fit width through closed-loop feedback; at the same time, it uses an improved path planning algorithm to dynamically generate obstacle avoidance trajectory, controls the stacking lifting component and stacking translation component to link multiple axes, and drives the stacker crane component to run to the target storage position with high precision. S4. Upon reaching the target storage location, if the target material box is located in the large tool area or toolbox area, the system controls the box locking component to perform multi-state interactive unlocking. After confirming that the lock is completely released, the system controls the stacking forks to extend and complete the material box grabbing, and smoothly transports the material box to the tool cabinet retrieval port for the operator to pick up and put away. S5. After the pick-up and drop-off are completed and a return confirmation instruction is received, the stacker crane component moves the material box back to its original storage position and triggers the cargo box locking component to lock it. The system simultaneously triggers the vision detection component, which uses a multi-view fusion strategy to perform a full-domain 3D point cloud scan of the inside of the material box. Through point cloud feature matching and calculation, the system completes the accurate verification of tool types and quantities, and then automatically updates the tool inventory database to form a closed-loop management.

[0005] In the preferred embodiment, the closed-loop control method for adaptive adjustment of fork spacing in step S3 is as follows: The system retrieves the target toolbox's width calibration value, i.e., its theoretical width. The ideal opening of the stacking forks is calculated based on the adaptive adaptation formula. : ; in, This is the allowance for structural interference compensation on one side of the forks. For dynamically set safety operating intervals, This is a fine-tuning compensation function set based on historical long-term wear and tear. The industrial control computer sends control commands to the fork adjustment assembly, driving the two sets of stacking forks to move relative to each other, and also collects and feeds back the real-time inner spacing of the stacking forks. When the error When the arrival signal is triggered, the next path operation is started.

[0006] In the preferred embodiment, the multi-state interactive unlocking and security monitoring method for the cargo box locking component in step S4 is as follows: The system sends an unlocking command to the cargo box locking component, executes the unlocking action, and performs real-time high-frequency detection of the unlocking status; Define a safety judgment strategy: Set a first threshold time t1 and a second threshold time t2 (t1 < t2). If a fully unlocked signal is detected within t1, it is determined to be a normal unlock. If no unlocked signal is detected during (t1, t2), the system controls the cargo box locking component to perform a reciprocating unblocking action. If no valid unlocked signal is received after t2, the system immediately cuts off the power supply of the stacker crane component to execute the safety brake, triggers an audible and visual alarm, and reports the fault code to the operation control system.

[0007] In the preferred embodiment, the method for global scanning and point cloud acquisition of the visual inspection component in step S5 is as follows: The system sends a scan trigger command to the vision detection component, which performs a uniform scanning motion and dynamically adjusts the parameters of the reflective and shadow interference areas inside the material box of the large tool area to suppress scan ghosting. During the scanning process, the collected image features and depth information are registered through multi-view fusion. After removing the outer boundary noise, the fusion generates a 3D point cloud dataset Praw that covers the entire environment inside the material box.

[0008] In the preferred embodiment, step S5, the specific calculation steps of the tool recognition algorithm based on the 3D point cloud dataset Praw, include: 5.1. The point cloud dataset Praw is downsampled using a spatial voxel mesh, and the RANSAC plane fitting algorithm is incorporated to remove the point clouds on the bottom and side walls of the hopper, and the isolated tool target point cloud clusters Ci are extracted. 5.2 For each main cluster, calculate its FPFH multidimensional geometric features and perform low-frequency coarse registration with the standard point cloud topology template in the tool base database of the industrial control computer. 5.3. The optimal spatial transformation matrix is ​​solved using the Weighted Iterative Closest Point (Weighted-ICP) algorithm. The objective function of the Weighted-ICP algorithm is defined as follows: ; in, and Let R and T be the coordinates of the point cloud cluster to be matched and the standard template, respectively, and R and T be the rotation matrix and translation vector to be solved. These are dynamic weighting coefficients set based on point curvature and normal vector similarity; when E(R,T) is lower than the preset classification convergence threshold... At that time, the system outputs a matrix of the types of tools that were successfully identified and their corresponding quantities.

[0009] In the preferred embodiment, the closed-loop management and status traceability mechanism for the tool inventory database in step S5 is as follows: The industrial control computer reads the tool array set Arraycurrent from the point cloud recognition calculation output and performs a difference calculation with the historical data Arraylast from the previous working cycle cached locally; If a negative difference set exists, the tool's independent ID status corresponding to the difference set will be changed to "outbound", and the operation timestamp will be strongly bound to the currently verified operator's permission account. If a positive difference set exists, the account binding of the tool's independent ID will be removed, and the tool's status will be refreshed to "in reserve". All operation logs will be encrypted and compressed and pushed to the superior management system on a regular basis to achieve long-term traceability management of the tool's entire lifecycle.

[0010] In the preferred embodiment, the flexible and simplified access method for the small tool area is as follows: When the system recognizes a toolbox whose selected target is the small tool area, it does not assign a scanning task to the vision inspection component and skips the vision scanning and recognition process. During the process of tool retrieval and placement and stacker crane components transferring the material box back to the warehouse, the system collects the static weight difference value ΔW before and after retrieval and placement through the weight detection module of the material box bearing structure; divides ΔW by the single-item calibrated weight of the dedicated small tool to calculate the integer value n of the tool quantity change value. At the same time, the manual confirmation and error-proofing mechanism of the operation control system interface is used to ensure that the inventory data of small tools has no drift deviation.

[0011] In the preferred embodiment, the improved path planning and dynamic storage scheduling method in step S3 has a time decay mechanism, specifically: Regarding path planning: the optimized A is adopted. The search algorithm introduces the energy consumption time difference of the mechanical movement of the stacking lifting component and the stacking translation component as a variable coefficient heuristic function parameter to reduce mechanical fatigue loss caused by high-frequency start and stop. Regarding storage location scheduling: The industrial control computer periodically calculates the comprehensive heat index of individual tool storage locations. The calculation formula is: ; in, , These represent the total frequency of entry and exit for the storage tool, respectively, and the third item is the exponential decay factor in the dimension of operation time. The attenuation constant is used; during the set idle period, the system automatically reorganizes all storage locations in the cabinet into fragmented units, and... The most active high-frequency toolbox is iteratively transported by stacker crane components to the central physical layer with the shortest three-dimensional straight-line distance from the picking port.

[0012] In the preferred embodiment, the method also features end-to-end anti-omission and correction, as well as timeout interception and early warning functions, specifically: After the stacker crane component moves the material box to the front picking port of the tool cabinet, the system's built-in timer starts counting down. If the operator does not trigger the return confirmation command in the operation control system within the set Twarn period, the system will lock the control page and send a warning message to the operator. The system continuously performs bidirectional inventory threshold checks: when the quantity of any type of tool is less than the preset minimum safety stock level, a replenishment alarm is triggered; when it is detected that a tool has been issued and its status is bound to an operator's account for a long period of time, an overdue issuance alarm is sounded and copied to the administrator.

[0013] In a preferred embodiment, the tool cabinet also includes a cabinet body, an operation control system, an industrial computer, a vision inspection component, a stacker crane component, and multiple sets of cargo box locking components; the cabinet body includes a sheet metal shell and a welded frame, the welded frame dividing the interior space into a central stacker crane operating channel and storage areas symmetrically distributed on both sides of the operating channel, the storage areas being divided into a tool box area, a small tool area, and a large tool area; The operation and control system is located on the top of the sheet metal housing. The industrial computer is electrically connected to the operation and control system, vision inspection components, stacker crane components, and cargo box locking components. The stacker crane components are located in the middle stacker crane running channel, including stacking forks, fork adjustment components, stacking lifting components and stacking translation components. The fork adjustment components can adjust the spacing between the stacking forks, and the stacking lifting components and stacking translation components realize multi-axis linkage of the stacking forks. The visual inspection component is set up in the large tool area, and the cargo box locking component is set up in the storage locations of the large tool area and the toolbox area. The welded frame is a grid pallet support structure that can withstand the asymmetric self-weight moment of each storage area when fully loaded. The system adopts a modular integrated design. The industrial control computer sends multi-threaded control commands to the stacker crane component, vision inspection component and cargo box locking component through the industrial Ethernet bus and receives data feedback for processing.

[0014] A flexible, adaptive intelligent tool cabinet based on visual recognition and its usage method have the following beneficial effects during use: 1. Enables flexible and adaptive picking and placement of tools of various sizes and automated registration, significantly reducing manual intervention. Through the closed-loop adaptive adjustment of the forklift adjustment component, it can accurately adapt to the picking and placement needs of three types of toolboxes: small toolbox, large toolbox, and tool box. Combined with the 3D point cloud recognition technology of the vision inspection component, it can automatically identify and register large tools in and out of the warehouse, replacing manual ledgers and improving the efficiency and accuracy of tool storage, retrieval, and registration. 2. Adopting a zoned, refined storage and differentiated management strategy improves space utilization and access efficiency. The tool cabinet is divided into three storage areas: small tools, large tools, and toolboxes. For small tools, a simplified access method of weight detection + manual confirmation is used, while for large tools, a visually accurate registration method is used. This balances saving computing power and management accuracy. At the same time, the refined zoning design significantly improves the cabinet space utilization and shortens the time for finding and placing tools. 3. Achieve closed-loop management and traceability of the entire tool lifecycle, improving the precision of management. Identity verification enables access control, and the tool's entry and exit status is strongly linked to the operator's account. Combined with timestamped operation logs and dynamic updates to the inventory database, the entire process of tool issuance, return, and inventory status is traceable, effectively preventing issues such as tool loss and overdue issuance. 4. Integrating intelligent path planning and multiple safety early warning mechanisms ensures safe equipment operation and stable maintenance. Optimized path planning algorithms reduce stacker crane mechanical wear, while multiple early warning systems, including abnormal unlocking monitoring of cargo container locking components, prevention of tool loss during storage and retrieval, and inventory threshold inspections, achieve dual safety control for equipment operation and inventory management. This adapts to the complex operating conditions at hydropower stations and ensures a stable supply of maintenance tools. Attached Figure Description

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is an external view of the intelligent tool cabinet of the present invention; Figure 2 This is a diagram of the internal skeleton of the intelligent tool cabinet of the present invention; Figure 3 This is a top view of the frame of the intelligent tool cabinet of the present invention; Figure 4 This is a side view of the frame of the intelligent tool cabinet of the present invention; Figure 5 This is a flowchart illustrating the method of using the intelligent tool cabinet of the present invention.

[0016] In the diagram: 1. Sheet metal shell; 2. Welded frame; 3. Middle stacker crane running channel; 4. Storage area; 401. Small tool area; 402. Large tool area; 403. Tool box area; 5. Industrial control computer; 6. Vision inspection component; 7. Stacker crane component; 701. Stacking fork; 702. Fork adjustment component; 703. Stacking lifting component; 704. Stacking translation component; 8. Cargo box locking component. Detailed Implementation

[0017] like Figure 1 to Figure 4As shown, this embodiment discloses a flexible adaptive intelligent tool cabinet based on visual recognition. The tool cabinet includes a cabinet body, an industrial control computer 5, an operation control system, and is also equipped with a visual detection component 6 for automatic tool identification and registration, a stacker crane component 7 for adaptive retrieval and placement of tool boxes, and multiple sets of cargo box locking components 8. The electrical components are electrically connected and coordinated to realize flexible adaptive retrieval and placement of maintenance tools of various specifications, automatic registration by visual recognition, and refined inventory management, adapting to the tool storage and use needs of hydro-generator maintenance sites. The cabinet body is the basic load-bearing structure of the tool cabinet, including a sheet metal shell 1 and a welded frame 2. The sheet metal shell 1 is located on the outermost side of the cabinet and is made of painted steel plate. It not only protects the internal components and stored tools from moisture, but also has a decorative effect. Its top also provides a stable installation base for the operation control system, making it convenient for on-site operators to input commands and check the equipment operation status nearby. The welded frame 2 is the core load-bearing structure of the cabinet. It is made of square steel pipe and has a grid-like pallet support structure. It has high structural strength and can effectively withstand the asymmetric self-weight moment generated when each storage area is fully loaded with tools. It is suitable for the actual use conditions at the hydropower station site. The welded frame 2 scientifically divides the internal space of the cabinet, forming a central stacker crane running channel 3 and two storage areas 4 symmetrically distributed on both sides of the running channel 3. The two storage areas 4 have the same symmetrical structure, which can maximize the use of the internal space of the cabinet and increase the overall tool storage capacity. The storage areas 4 on both sides are divided into small tool area 401, large tool area 402 and toolbox area 403 according to the specifications, size and usage characteristics of the tools, so as to realize the refined storage of tools by area; among them, the toolbox area 403 is specially arranged in the position closest to the cabinet access port, which is suitable for the high-frequency access needs of complete toolboxes, and can effectively shorten the transfer path of stacker crane component 7 and improve storage and retrieval efficiency. The operation control system is fixedly installed at the top of the sheet metal housing 1. It serves as the interaction unit between the operator and the tool cabinet, and can receive instructions from the operator for issuing and returning tools. At the same time, it can display the operating status of the equipment, tool inventory information, and fault prompts in real time. The industrial control computer 5 is the core data processing and control unit of the tool cabinet. It is electrically connected to the operation control system, vision inspection component 6, stacker crane component 7, and cargo box locking component 8 through an industrial Ethernet bus. It can issue multi-threaded control instructions to each execution component, and receive operating status, detection data, and other information returned by each component. It can complete real-time calculation, processing, and storage of data, and provide core data support for the intelligent control of tool access. The stacker crane component 7 is located within the intermediate stacker crane operating channel 3 and is the core execution component for toolbox retrieval, placement, and transfer. It comprises stacking forks 701, fork adjustment components 702, stacking lifting components 703, and stacking translation components 704. These components work in tandem to achieve precise multi-axis movement. The stacking lifting components 703 and stacking translation components 704 work together to drive the stacking forks 701 to move precisely to any target storage location in three-dimensional space. The fork adjustment components 702 can adaptively adjust the relative distance between the two sets of stacking forks 701 according to the size parameters of the target toolbox, precisely adapting to the retrieval and placement needs of three different sizes of toolboxes: small toolboxes, large toolboxes, and tool boxes. The stacking forks 701 can extend and retract to grasp and place the toolboxes, and in conjunction with the overall movement of the component, complete the smooth transfer of the toolboxes between the target storage location and the cabinet's retrieval opening. The visual inspection component 6 is set up in the large tool area 402. Its position is directly opposite the material box storage position of the large tool area 402. It can perform full-area scanning of the inside of the large tool box, collect the three-dimensional point cloud data of the tool and transmit it back to the industrial control computer 5 in real time. The industrial control computer 5 completes the processing of point cloud data, feature matching and tool recognition, and finally realizes the automatic registration of large tools entering and leaving the warehouse, replacing the traditional manual ledger recording method. The cargo box locking assembly 8 is configured in multiple sets, and is deployed one by one for each storage location in the large tool area 402 and the toolbox area 403. The small tool area 401 does not need to be equipped with this assembly because the toolboxes are small and lightweight. The cargo box locking assembly 8 can lock and fix the large toolboxes and toolboxes stored in the corresponding storage locations, effectively preventing the toolboxes from sliding or shifting during cabinet operation and stacker crane operation, ensuring the stability of the toolbox storage. In addition, before the toolboxes are taken out of the warehouse, the cargo box locking assembly 8 can be automatically unlocked according to the instructions of the industrial control computer 5, without affecting the normal picking and placing operation of the toolboxes by the stacker crane assembly 7.

[0018] Example 1: Visual intelligent access and closed-loop management method for large tools: This embodiment mainly demonstrates how to utilize this flexible adaptive intelligent tool cabinet for automated storage and retrieval of large tools (such as large wrenches, special clamps, etc.) and inventory checks during the maintenance of hydro-generator units. Figure 5 (as shown) S11. Authentication and Access Control: Operators verify their identity using facial recognition or card swiping on the control panel at the front of the tool cabinet. After successful verification by the industrial control computer 5, the system assigns equipment operation permissions to the operator, generates an operation log with a global timestamp in the background, and retrieves the tool basic database and storage location mapping table pre-stored in the industrial control computer 5. S12. Access Task Matching and Parameter Parsing: The operator issues a command to retrieve large tool A via the screen. The system automatically matches the tool to a specific large toolbox stored in the large tool area 402, and analyzes the size parameters (such as width and depth) and current real-time weight of the large toolbox in real time. S13. Stacker crane adaptive adjustment and path operation: Based on the dimensions of the large toolbox obtained from the analysis, the industrial control computer 5 controls the fork adjustment assembly 702 to start, and adaptively adjusts the distance between the two sets of stacking forks 701 through the electronic control driver to precisely match the width of the bottom slot of the large toolbox. Subsequently, the system dynamically plans the running trajectory, controls the stacking lifting assembly 703 and the stacking translation assembly 704 to work together efficiently on multiple axes, and quickly travels along the middle stacker crane running channel 3 and accurately positions itself to the target storage location in the storage area 4; S14. Unlocking, storage, and transfer of material boxes: After the stacker crane reaches the target position in the large tool area 402, the system sends an unlocking command to the cargo box locking assembly 8. The cylinder inside the cargo box locking assembly 8 pulls out the locking pin, and the sensor confirms that it is fully unlocked. Subsequently, the stacker forks 701, which have been adjusted to the correct width, extend smoothly, lifting the large tool box from the welded frame 2, and it is transported by the stacker crane assembly 7 to the retrieval port for the operator to access the large tool A. S15. Material box placement and automatic visual registration: After the operator takes tool A and confirms its return, the stacker crane component 7 transports the large toolbox back to its original position. The cargo box locking component 8 immediately locks it back onto the welded frame 2 for safety. At this time, the system automatically triggers the vision inspection component 6 installed above the area to perform a full-area 3D point cloud scan of the large toolbox. The industrial control computer 5 matches the scanned point cloud data with the database model and finds that tool A is missing. It then automatically generates a system record stating "Tool A has been issued, recipient: XXX", achieving a closed-loop tool library storage system without the need for manual inventory checks.

[0019] Example 2: Efficient and Flexible Access and Timeout Prevention Method for Small Tool Areas: This embodiment primarily demonstrates a simplified scheduling process and safety fault-tolerance mechanism for retrieving consumable, high-frequency small tools (such as various screws, washers, multimeter probes, etc.) during routine maintenance. Figure 5 (as shown) S21. Authentication and Access Control: After the operator logs into their personal account and the industrial control computer verifies their identity, the operator unlocks the operation permissions for small tools and simultaneously retrieves the security protection protocol for the picking port. S22. Access Task Matching and Parameter Parsing: When an operator selects a certain high-frequency small tool (such as a nut of a specific size) in the system, the system quickly retrieves the material and finds that it is stored in the dedicated small tool box in the small tool area 401, and reads the ultra-small width dimension and light load threshold of the small tool box. S23. Stacker crane adaptive adjustment and path operation: The industrial control computer 5 issues a command to the fork adjustment component 702 to bring the spacing of the stacking forks 701 together to the minimum adaptable width. The stacking translation component 704 and the stacking lifting component 703, based on the system's built-in time decay scheduling algorithm, prioritize the optimal and least damaged mechanical movement path and smoothly move along the middle stacker crane running channel 3 to the small tool area 401; S24. Material box retrieval and placement (unlock-free mechanism): Because the individual toolboxes in the small tool area 401 are extremely lightweight, the area was not equipped with a box locking component 8 according to the initial settings. After the stacker crane component 7 arrives, the stacker forks 701 directly insert and pull out the small toolboxes, transferring and lowering them onto the storage / retrieval worktable at the front of the sheet metal housing 1. At this time, the system's built-in watchdog timer begins its countdown. S25, Simplified Home Relocation and Anti-Timeout Interception: If the operator leaves for any reason and does not click the return command within the set safety period, the system will immediately activate the anti-runaway mode to lock the control page and send an alarm prompt to the operator's mobile terminal through the reserved communication interface. If the toolbox is retrieved and confirmed to be returned normally, the stacker crane component 7 will transport the toolbox back to its original storage location. For dedicated toolboxes, the system does not utilize the computing power allocated by the vision inspection component 6, but instead records changes in inventory quantity through internal high-precision error-proof calculation logic (or direct system soft confirmation). The industrial control computer 5 quickly refreshes the inventory status. If it detects that the remaining quantity of the small tool is below the set minimum safety alarm line, the system will automatically sound a replenishment alarm to ensure sufficient spare parts.

Claims

1. A flexible adaptive intelligent tool cabinet based on vision recognition, comprising a cabinet body, an industrial control computer, and an operation control system, characterized in that: The cabinet body includes a sheet metal shell (1) and a welded frame (2). The welded frame (2) divides the cabinet space into a central stacker crane running channel (3) and storage areas (4) symmetrically distributed on both sides of the running channel (3). The storage area (4) is divided into a toolbox area (403), a small tool area (401), and a large tool area (402). The intelligent tool cabinet is also equipped with a vision detection component (6) for automatic tool identification and registration, and a stacker component (7) for adaptive retrieval and placement of tool boxes. The stacker crane assembly (7) includes a stacking fork (701), a fork adjustment assembly (702), a stacking lifting assembly (703), and a stacking translation assembly (704); both the large tool area (402) and the toolbox area (403) are equipped with a box locking assembly (8); The method of using the above tool cabinet includes the following steps: S1. The operator completes the identity verification through the operation control system. After the system verification is passed, the operation permission is unlocked, an operation log with a global timestamp is created, and the tool basic database, storage location mapping database and stacker crane dynamic operation parameter library stored in the industrial control computer (5) are retrieved at the same time. S2. When the operator initiates a tool requisition or return instruction, the system automatically matches the corresponding toolbox type and storage location coordinates according to the instruction, and parses and obtains the size topology parameters and load threshold of the target material box in real time. S3. Based on the size and topology parameters of the target material box, the system controls the fork adjustment component (702) to adaptively adjust the spacing between the two sets of stacking forks (701) to the optimal fit width through closed-loop feedback; at the same time, it uses an improved path planning algorithm to dynamically generate an obstacle avoidance trajectory, controls the stacking lifting component (703) and the stacking translation component (704) to work together on multiple axes, and drives the stacker component (7) to run to the target storage position with high precision. S4. After reaching the target storage location, if the target box is located in the large tool area (402) or toolbox area (403), the system controls the box locking component (8) to perform multi-state interactive unlocking. After confirming that the lock is completely released, the system controls the stacking fork (701) to extend and complete the box grabbing, and smoothly transports the box to the tool cabinet retrieval port for the operator to pick up and put away. S5. After the pick-up and drop-off are completed and the return confirmation instruction is received, the stacker crane component (7) moves the material box back to its original storage position and triggers the cargo box locking component (8) to lock it; the system synchronously triggers the vision detection component (6), adopts a multi-view fusion strategy to perform a full-domain three-dimensional point cloud scan of the inside of the material box, and completes the accurate verification of the tool type and quantity through point cloud feature matching calculation, thereby automatically updating the tool inventory database to form a closed-loop management.

2. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: In step S3, the closed-loop control method for adaptive adjustment of fork spacing is as follows: The system retrieves the target toolbox's width calibration value, i.e., its theoretical width. The ideal opening of the stacking forks (701) is calculated based on the adaptive adaptation formula. : ； in, This is the allowance for structural interference compensation on one side of the forks. For dynamically set safety operating intervals, This is a fine-tuning compensation function set based on historical long-term wear and tear. The industrial control computer (5) sends control commands to the fork adjustment assembly (702) to drive the two sets of stacking forks (701) to move relative to each other, and collects and receives feedback on the real-time inner spacing of the stacking forks (701). When error When the arrival signal is triggered, the next path operation is started.

3. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: In step S4, the multi-state interactive unlocking and security monitoring method of the cargo box locking component (8) is as follows: The system sends an unlocking command to the cargo box locking component (8), executes the unlocking action, and performs real-time high-frequency detection of the unlocking status; Define a security determination strategy: Set a first threshold time t1 and a second threshold time t2 (t1 < t2). If a fully unlocked signal is detected within t1, then the unlock is determined to be normal. If no unlocking signal is detected during (t1, t2), the system controls the cargo box locking component (8) to perform a reciprocating unblocking action; if no valid unlocking signal is received after t2, the system immediately cuts off the power supply of the stacker crane component (7) to perform a safety brake, triggers an audible and visual alarm, and reports the fault code to the operation control system.

4. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: In step S5, the global scanning and point cloud acquisition method of the visual inspection component (6) is as follows: The system sends a scan trigger command to the vision detection component (6), and the vision detection component (6) performs a uniform scanning motion. It dynamically adjusts the parameters of the reflective and shadow interference areas inside the material box of the large tool area (402) to suppress the scanning shadow. During the scanning process, the collected image features and depth information are registered through multi-view fusion. After removing the outer boundary noise, the fusion generates a 3D point cloud dataset Praw that covers the entire environment inside the material box.

5. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 4, characterized in that: In step S5, the specific calculation steps of the tool recognition algorithm based on the 3D point cloud dataset Praw include: 5.

1. The point cloud dataset Praw is downsampled using a spatial voxel mesh, and the RANSAC plane fitting algorithm is incorporated to remove the point clouds on the bottom and side walls of the hopper, and the isolated tool target point cloud clusters Ci are extracted. 5.2 For each main cluster, calculate its FPFH multidimensional geometric features and perform low-frequency coarse registration with the standard point cloud topology template in the tool base database of the industrial control computer (5); 5.

3. The optimal spatial transformation matrix is ​​solved using the Weighted Iterative Closest Point (Weighted-ICP) algorithm. The objective function of the Weighted-ICP algorithm is defined as follows: ； in, and Let R and T be the coordinates of the point cloud cluster to be matched and the standard template, respectively, and R and T be the rotation matrix and translation vector to be solved. These are dynamic weighting coefficients set based on point curvature and normal vector similarity; when E(R,T) is lower than the preset classification convergence threshold... At that time, the system outputs a matrix of the types of tools that were successfully identified and their corresponding quantities.

6. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: In step S5, the closed-loop management and status traceability mechanism for the tool inventory database is as follows: The industrial control computer (5) reads the tool array set Arraycurrent output from this point cloud recognition calculation and performs a difference calculation with the historical data Arraylast from the previous working cycle cached locally; If a negative difference set exists, the tool's independent ID status corresponding to the difference set will be changed to "outbound", and the operation timestamp will be strongly bound to the currently verified operator's permission account. If a positive difference set exists, the account binding of the tool's independent ID will be removed, and the tool's status will be refreshed to "in reserve". All operation logs will be encrypted and compressed and pushed to the superior management system on a regular basis to achieve long-term traceability management of the tool's entire lifecycle.

7. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: The flexible and minimalist access method for the small tool area (401) is as follows: When the system recognizes a toolbox whose selected target is the small tool area (401), it does not assign a scanning task to the visual inspection component (6) and skips the visual scanning recognition process; During the process of tool picking and placing and stacker crane component (7) transferring the material box back to the warehouse, the system collects the static weight difference value ΔW before and after picking and placing through the weight detection module of the material box bearing structure; divides ΔW by the single item calibration weight of the dedicated small tool to calculate the integer value n of the tool quantity change value, and at the same time, the manual confirmation error prevention mechanism of the operation control system interface is used to ensure that the inventory data of small tools has no drift deviation.

8. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: The improved path planning and dynamic storage scheduling method in step S3 has a time decay mechanism, specifically: Regarding path planning: the optimized A is adopted. The search algorithm introduces the energy consumption time difference of the mechanical movement of the stacking lifting component (703) and the stacking translation component (704) as a variable coefficient heuristic function parameter to reduce mechanical fatigue loss caused by high frequency start-stop. Regarding storage location scheduling: The industrial control computer (5) periodically calculates the comprehensive heat index of individual tool storage locations. The calculation formula is: ； in, , These represent the total frequency of entry and exit for the storage tool, respectively, and the third item is the exponential decay factor in the dimension of operation time. The attenuation constant is used; during the set idle period, the system automatically reorganizes all storage locations in the cabinet into fragmented units, and... The most active high-frequency toolbox is iteratively transported to the central physical layer with the shortest three-dimensional straight distance from the picking port by the stacker crane component (7).

9. The flexible adaptive intelligent tool cabinet based on visual recognition according to claim 1, characterized in that: This method also features end-to-end anti-omission and correction, as well as timeout interception and early warning functions, specifically: When the stacker crane component (7) moves the material box to the front picking port of the tool cabinet, the built-in timer of the system starts counting down. If the operator does not trigger the return confirmation command in the operation control system within the set Twarn period, the system will lock the control page and send a warning message to the operator. The system continuously performs bidirectional inventory threshold checks: when the quantity of any category of tools falls below the preset minimum safety stock level, a replenishment alert is triggered; When it is detected that a tool has been used and its status is permanently bound to a certain operator's account, an overdue use alarm will be triggered and copied to the administrator.

10. The flexible adaptive intelligent tool cabinet based on visual recognition according to any one of claims 1-9, characterized in that: The tool cabinet also includes a cabinet body, an operation control system, an industrial computer (5), a vision inspection component (6), a stacker crane component (7), and multiple sets of cargo box locking components (8); the cabinet body includes a sheet metal shell (1) and a welded frame (2), the welded frame (2) divides the internal space of the cabinet into a central stacker crane running channel (3) and storage areas (4) symmetrically distributed on both sides of the running channel (3), the storage area (4) is divided into a tool box area (403), a small tool area (401), and a large tool area (402); The operation control system is located at the top of the sheet metal shell (1), and the industrial computer (5) is electrically connected to the operation control system, the vision inspection component (6), the stacker crane component (7) and the cargo box locking component (8). The stacker crane assembly (7) is located in the middle stacker crane running channel (3) and includes a stacking fork (701), a fork adjustment assembly (702), a stacking lifting assembly (703) and a stacking translation assembly (704). The fork adjustment assembly (702) can adjust the spacing of the stacking forks (701). The stacking lifting assembly (703) and the stacking translation assembly (704) realize multi-axis linkage of the stacking forks (701). The visual inspection component (6) is set in the large tool area (402), and the cargo box locking component (8) is set in the storage locations of the large tool area (402) and the toolbox area (403); The welded frame (2) is a grid pallet support structure that can withstand the asymmetric self-weight moment of each storage area when fully loaded. The system adopts a modular integrated design. The industrial control computer (5) sends multi-threaded control commands to the stacker crane component (7), vision inspection component (6) and cargo box locking component (8) through the industrial Ethernet bus and receives data feedback for processing.

Images