Dynamic calibration of crane spreader extreme position non-contact measurement system and method

Abstract

The application discloses a kind of dynamic calibration crane sling limit position non-contact measurement system and method, system includes binocular vision unit, calibration unit, control unit;Crane includes lower crossbeam, trolley, wire rope and sling, trolley lower portion is connected with sling by wire rope, calibration unit includes sling feature marker and two groups of fixed calibration checkerboard, sling feature marker is fixed on sling;Binocular vision unit includes left industrial camera and right industrial camera, two industrial cameras are symmetrically arranged, optical axis is upwardly inclined and points to sling operation area, for synchronously collecting the image of sling feature marker and fixed calibration checkerboard;Control unit is used to realize image data real-time processing, three-dimensional coordinate solution and control crane operation.The application has strong light interference resistance and vibration resistance, can adapt to complex field working condition, significantly improves special equipment inspection and detection efficiency and data reliability.

Description

Technical Field

[0001] This invention belongs to the field of crane testing technology, specifically relating to a non-contact measurement system and method for dynamically calibrated crane lifting device limit positions. Background Technology

[0002] Gantry cranes are key lifting equipment widely used in ports, freight yards, and construction sites. Their main structure typically consists of a metal structure, a hoisting mechanism, a trolley traveling mechanism, and a trip trolley traveling mechanism. Among these, the accuracy of the limit positions of the spreader directly affects the crane's operational safety and efficiency.

[0003] Currently, the traditional method for measuring the limit positions of gantry crane spreaders mainly uses contact or semi-contact manual measurement methods. The specific operation procedure is as follows: first, mark the center point on the surface of the spreader, then control the spreader to the left and right limit positions of the trolley travel, and finally, the inspector uses a laser rangefinder or steel ruler to manually measure the horizontal distance between the center line of the spreader and the center line of the track.

[0004] However, the aforementioned traditional measurement methods have significant technical drawbacks in practical applications:

[0005] 1. Low work efficiency and cumbersome operation: This method relies on manual marking on the lifting device, which is time-consuming and labor-intensive; and each inspection requires repeated marking and resetting, which is difficult to meet the needs of large-scale and high-frequency inspection.

[0006] 2. Measurement accuracy is severely limited by the environment and tools:

[0007] When using a steel ruler for measurement, the measurement data is easily affected by the ruler's own weight bending, thermal expansion and contraction stretching, and human reading errors, resulting in large dispersion and difficulty in guaranteeing accuracy.

[0008] While laser rangefinders improve accuracy to some extent, their measuring beams are easily affected by strong ambient light and require an absolutely unobstructed measuring path, making them less adaptable to complex port or construction site environments.

[0009] 3. Significant safety hazards exist: Traditional methods require inspection personnel to climb onto the crane itself or work close to the track area. This not only increases the risk of falls from heights but also exposes personnel to the danger zone of moving machinery, seriously threatening their lives and failing to meet the "inherent safety" requirements of modern special equipment inspection.

[0010] In summary, there is an urgent need to develop a non-contact, high-precision, and highly safe method for measuring the limit position of lifting devices in order to solve the aforementioned problems existing in the current technology. Summary of the Invention

[0011] In view of the defects and shortcomings of the existing technology, the present invention proposes a non-contact measurement system and method for dynamically calibrated crane lifting device limit positions.

[0012] The technical solution adopted in this invention is:

[0013] A non-contact measurement system for the limit positions of a crane spreader with dynamic calibration includes a binocular vision unit, a calibration unit, and a control unit;

[0014] The crane includes a lower crossbeam, a trolley, a wire rope, and a lifting device. The trolley is located at the top and can move between left and right extreme positions. The lifting device is connected to the lower part of the trolley via a wire rope. The lower crossbeam includes a left lower crossbeam and a right lower crossbeam.

[0015] The calibration unit includes a lifting device feature marker and two sets of fixed calibration checkerboard grids. The lifting device feature marker is fixed on the lifting device. The lifting device feature marker adopts a composite structure of a central fluorescent circular area and an outer auxiliary checkerboard grid. The central fluorescent circular area is a high-precision measurement target surface, and the outer auxiliary checkerboard grid is used to provide auxiliary corner point features. The two sets of fixed calibration checkerboard grids are respectively fixedly installed on the opposite inner sides of the lower left crossbeam and the lower right crossbeam.

[0016] The binocular vision unit includes a left industrial camera and a right industrial camera. The two industrial cameras are symmetrically arranged, with their optical axes tilted upwards and pointing towards the working area of ​​the lifting device, for synchronously acquiring images of the lifting device's characteristic markers and the fixed calibration checkerboard pattern.

[0017] The control unit is used to realize real-time image data processing, three-dimensional coordinate calculation, and control the operation of the crane.

[0018] Furthermore, the control unit includes an industrial computer and a PLC controller. The industrial computer is used for real-time image data processing and three-dimensional coordinate calculation. The PLC controller is used to receive instructions from the industrial computer and control the operation of the crane. The industrial computer and the PLC controller establish a high-speed data link through an industrial communication protocol.

[0019] Furthermore, the left industrial camera is rigidly mounted on the lower left crossbeam of the crane, and the right industrial camera is rigidly mounted on the lower right crossbeam of the crane.

[0020] A measurement method for a non-contact measurement system for the limit positions of a crane spreader based on the above-described dynamic calibration includes the following steps:

[0021] Step 1: System initialization and parameter configuration;

[0022] Step 2: Online dynamic calibration;

[0023] Step 3: Image acquisition and validity verification at extreme positions;

[0024] Step 4: Mixed matching and center point pixel coordinate extraction;

[0025] Step 5: 3D reconstruction and limit position determination.

[0026] Furthermore, step 1 specifically includes:

[0027] Camera Intrinsic Calibration: An independent standard checkerboard grid was used to calibrate the intrinsic parameters of the left and right industrial cameras separately. During calibration, the independent standard checkerboard grid was placed in front of the left and right industrial cameras within a clear field of view, and their spatial orientation was varied to acquire multiple first images with different orientations. The intrinsic parameter matrices and radial distortion coefficients k1 and k2 were then solved for each camera. The intrinsic parameter matrices include the pixel focal length f. x ,f y Principal point coordinates (c x ,c y );

[0028] Left and right camera relative extrinsic parameter calibration: Determine the fixed spatial transformation relationship between two fixed calibration checkerboard grids on the left and right sides, and combine the industrial control computer to trigger the left industrial camera and the right industrial camera to synchronously acquire a second image containing the fixed calibration checkerboard grid on the opposite side, and calculate the fixed relative extrinsic parameter between the left industrial camera and the right industrial camera, wherein the fixed relative extrinsic parameter includes the baseline distance b.

[0029] World coordinate system definition: The world coordinate system is defined with the geometric center of the following beam as the origin: the X-axis is along the direction of the trolley track, that is, the left and right movement direction of the lifting device, the Y-axis is perpendicular to the direction of the trolley track, and the Z-axis is vertically upward. The three-dimensional coordinates of the corner points of the fixed calibration checkerboard (9) in the world coordinate system are pre-determined and stored. Template image pre-storage and feature library construction: During the system installation and debugging stage, the template acquisition program is executed: the lifting device is controlled to run to the center area of ​​the common field of view of the left and right industrial cameras. Under the condition that the crane is stationary and the lighting is uniform, the template image containing the feature markers of the lifting device is acquired synchronously by the left and right industrial cameras. The central fluorescent area of ​​the feature markers of the lifting device in the images of the left and right industrial cameras is extracted as the matching template, and the ORB feature library corresponding to the industrial camera and the right industrial camera is established synchronously for subsequent fast matching.

[0030] Furthermore, step 2 specifically includes:

[0031] Image acquisition: When the crane is stationary, the industrial computer triggers the left and right industrial cameras to synchronously acquire a third image containing the fixed calibration checkerboard pattern on the opposite side in real time;

[0032] Calculate real-time extrinsic parameters: Based on the world coordinates of the corner points of the fixed calibration chessboard grid pre-stored in step 1, and combined with the pixel coordinates extracted from the third image, solve the current real-time absolute extrinsic parameters of the left and right industrial cameras respectively, including the rotation matrix R and the translation vector T.

[0033] Calibration result storage: Store the current real-time absolute extrinsic parameters of the left and right industrial cameras calculated in step 2, the intrinsic parameter matrix calibrated in step 1, and the radial distortion coefficients k1 and k2 into the industrial control computer.

[0034] Furthermore, step 3 specifically includes:

[0035] The industrial control computer drives the trolley to move to the left through the PLC controller. The trolley moves the lifting device synchronously via the wire rope. When the trolley reaches the leftmost position allowed by the machinery, the PLC controller immediately issues a deceleration and stop command. The trolley stops moving. After the PLC controller detects that the trolley has stopped moving and is not moving, it continues to wait for the lifting device to stop swinging completely. The stationary position of the lifting device at this time is the left limit position of the trolley's travel, which is also the left limit point of the lifting device measured in this case. After that, the industrial control computer triggers the left industrial camera and the right industrial camera to synchronously acquire a complete fourth image.

[0036] The industrial control computer drives the trolley to move to the right through the PLC controller. The trolley moves the lifting device synchronously via the wire rope. When the trolley reaches the rightmost position allowed by the machinery, the PLC controller immediately issues a deceleration and stop command. The trolley stops moving. After the PLC controller detects that the trolley has stopped moving and is not moving, it continues to wait for the lifting device to stop swinging completely. The stationary position of the lifting device at this time is the right limit position of the trolley's travel, which is also the right limit point of the lifting device measured in this measurement. After that, the industrial control computer triggers the left and right industrial cameras to synchronously acquire a complete fifth image.

[0037] The complete fourth and fifth images acquired must simultaneously include: 1. Feature markers of the lifting device: serving as the measurement benchmark for calculating the spatial position of the lifting device in subsequent steps; 2. Fixed calibration checkerboard on the opposite side: serving as a reference benchmark for verifying the validity of the online calibration parameters obtained in step 2;

[0038] Extract the pixel coordinates of all corner points of the fixed calibration checkerboard in the fourth and fifth images respectively. Using the camera absolute extrinsic parameters obtained in step 2, reproject the world coordinates of the corner points stored in step 1 onto the fourth and fifth images respectively. Calculate the root mean square error of reprojection, i.e., the RMS error.

[0039] If the RMS errors of the fourth and fifth images do not exceed the preset threshold, the camera absolute extrinsic parameters obtained from the online dynamic calibration in step 2 are deemed valid, and step 4 is executed. If the RMS error of at least one of the fourth and fifth images exceeds the threshold, the camera absolute extrinsic parameters are deemed to have drifted, the corresponding image with the RMS error exceeding the threshold is discarded, and step 2 is returned to be re-performed for online dynamic calibration. Then step 3 is executed again until the RMS errors of the fourth and fifth images do not exceed the preset threshold.

[0040] Furthermore, step 4 specifically includes:

[0041] Distortion Correction: Using the radial distortion coefficients k1 and k2 calibrated in step 1, distortion correction is performed on each pixel in the fourth and fifth images, as shown in the following formula:

[0042]

[0043]

[0044] Where (x, y) are the original image pixel coordinates of the fourth and fifth images, (x... 修正 ,y 修正 ) represents the pixel coordinates after distortion correction, (c x ,c y ) represents the coordinates of the principal point obtained in step 1, r 2 =(xc x ) 2 +(yc) y ) 2 r 4 =(r 2 ) 2 ;

[0045] The fourth image is corrected to obtain the sixth image, and the fifth image is corrected to obtain the seventh image;

[0046] ROI detection area delineation: In the sixth and seventh images, ROI detection areas containing only the characteristic markers of the lifting equipment are delineated based on the running trajectory of the lifting equipment. These ROI detection areas spatially avoid the area where the fixed calibration checkerboard is located on the opposite side, thus eliminating the interference of checkerboard corner points on marker recognition.

[0047] Adaptive Binarization: Taking into account the characteristic that the center of the lifting device feature is a high reflectivity fluorescent circular area, the system uses the Otsu algorithm to segment the image within the ROI detection area, analyzes the grayscale histogram, finds the optimal threshold to separate the foreground and background, sets pixels above the threshold to white and pixels below the threshold to black, generates a binary image, and highlights the bright features of the lifting device feature.

[0048] SSDA coarse matching: Based on the binarized image, the Sequential Similarity Detection (SSDA) algorithm is used for fast localization. A pre-stored template image of the lifting device's feature markers is called, and matching is performed block by block within the ROI detection area using a sliding window approach. For each window position, the cumulative absolute error of pixel grayscale between the template image and the image block covered by the current window is calculated. Once the cumulative absolute error exceeds a preset threshold, the calculation for the current window is immediately terminated, and the process moves to the starting position of the next sliding window to continue matching. After traversing all windows, the top K candidate regions with the smallest cumulative absolute error are selected.

[0049] ORB fine matching: For the candidate regions selected by SSDA, the ORB (Oriented FAST and Rotated BRIEF) algorithm is used for precise verification, extracting the feature point descriptors of the candidate regions, and calculating the Hamming distance with the template features in the ORB feature library pre-stored in step 1.

[0050] The candidate region with the smallest Hamming distance is selected as the final matching result. If the smallest Hamming distance is less than the preset threshold, the matching is considered successful; otherwise, the matching is considered to have failed. The current frame data is discarded and the process returns to step 3 to re-acquire the image of the extreme position on that side. If the number of consecutive failures reaches the preset threshold, the measurement is paused, a marker recognition failure is reported, and a prompt is made to check the light source or the visibility of the marker.

[0051] Centroid coordinate calculation: After determining the successfully matched marker region, the precise pixel coordinates of the center point are calculated using the grayscale centroid method.

[0052] Left industrial camera:

[0053]

[0054] Right industrial camera:

[0055]

[0056] ∑X 左 : Represents the sum of the x-coordinates of all white pixels in the sixth image;

[0057] ∑Y 左 : Represents the sum of the y-coordinates of all white pixels in the sixth image;

[0058] M1: Represents the total number of white pixels in the sixth image;

[0059] ∑X 右 : Represents the sum of the x-coordinates of all white pixels in the seventh image;

[0060] ∑Y 右: Represents the sum of the y-coordinates of all white pixels in the seventh image;

[0061] M2: Represents the total number of white pixels in the seventh image;

[0062] Output: The calculated (x1,y1) and (x2,y2) are the precise pixel coordinates of the center point of the lifting device feature marker in the photos of the left and right industrial cameras. These two coordinates will directly correspond to the pixel positions of the lifting device at the left and right extreme positions, and will be used for the 3D reconstruction in the subsequent step 6.

[0063] Furthermore, the number of candidate regions K in step 4 is 3.

[0064] Furthermore, step 5 specifically includes:

[0065] 3D Reconstruction Based on Triangulation: Using the pixel coordinates (x1, y1) and (x2, y2) of the center point of the lifting device feature marker in the left and right images output in step 4, the pixel coordinates are converted into physical coordinates (X0, Y0, Z0) in the world coordinate system. Let P be the measured point on the lifting device feature marker, and its pixel coordinates on the imaging planes of the left and right industrial cameras are (x1, y1) and (x2, y2) respectively. Using the camera intrinsic parameter matrix calibrated in step 1, the 3D coordinates are calculated according to the following formula:

[0066]

[0067] Parameter description: x1-x2: disparity, which is the horizontal pixel offset of the same feature point in the left and right images. Disparity is inversely proportional to depth. The larger the disparity, the closer to the camera.

[0068] b: Baseline distance, which is the physical distance between the optical centers of the left and right industrial cameras, is obtained by calibration in step 1;

[0069] Multiple sampling and extreme value determination at extreme positions: Due to mechanical vibration and wire rope swing when the trolley stops at the extreme position, the X0 value of a single measurement may fluctuate. To ensure the effectiveness of the safety boundary, multiple independent measurements are performed on the left and right extreme positions respectively. During continuous measurements, the three-dimensional coordinate calculations in steps 4 and 5 are fully executed for each measurement. Then, extreme value screening is performed.

[0070] Left Limit Determination: When the trolley reaches the left limit position, perform N measurements and take the minimum value of all X0 values, denoted as X. 0min Simultaneously, take the average of all Y0 and Z0 values, denoted as... 0、 0, the world coordinates of the left extreme position are obtained as (X 0min , 0, 0);

[0071] Right-side limit determination: When the trolley reaches the right-side limit position, perform N measurements, and take the maximum value of all X0 values, denoted as X. 0max， Simultaneously, take the average of all Y0 and Z0 values, denoted as... 0、 0, the world coordinates of the extreme position on the right are (X 0max , 0, 0);

[0072] Finally, output the world coordinates (X) of the left extreme position. 0min , 0, 0) and the world coordinates (X) of the right extreme position. 0max , 0, 0).

[0073] The beneficial effects of this invention are:

[0074] (1) The robustness and real-time performance of the algorithm are significantly improved.

[0075] This invention employs a hybrid feature recognition algorithm combining SSDA coarse matching and ORB fine matching. The SSDA algorithm rapidly eliminates background interference regions by accumulating an error threshold, significantly reducing unnecessary computation. The ORB algorithm, relying on a multi-scale pyramid and rotation-invariant BRIEF descriptors, achieves stable matching for rotation, scale changes, and local occlusion. Combined with high-precision pixel coordinates after distortion correction and a virtual ROI dynamic detection area, this hybrid algorithm achieves a higher target recognition rate compared to single matching algorithms under complex conditions such as strong light interference, mechanical vibration, and slight twisting of markers. The single-frame processing time meets real-time detection requirements, effectively solving the problems of target loss, mismatch, and false detection inherent in traditional visual algorithms under harsh environments.

[0076] 2. It has online dynamic calibration capability, effectively suppressing accuracy drift.

[0077] This invention constructs a dual calibration mechanism consisting of a fixed calibration checkerboard grid on the lower crossbeam and pre-calibration of the relative extrinsic parameters of the binocular camera. The fixed calibration checkerboard grid is rigidly connected to the lower crossbeam of the crane, ensuring a constant relative pose between the camera and the checkerboard grid over a long period. The auxiliary checkerboard grid surrounding the characteristic markers of the lifting device is used only for matching and does not participate in the calibration. Simultaneously, the initial relative extrinsic parameters and baseline distance pre-calibration of the left and right cameras are completed. Before each measurement, the system uses the PnP algorithm to calculate the current absolute extrinsic parameters of the camera in real time based on the fixed checkerboard grid on the same side, automatically compensating for pose changes caused by vibration and temperature drift. Compared to traditional one-time offline calibration methods, this invention significantly suppresses the long-term decay of measurement accuracy and greatly improves the long-term stability and consistency of lifting device limit position measurements.

[0078] 3. High system integration, enabling non-destructive retrofitting of cranes.

[0079] This invention highly integrates binocular vision measurement, PLC closed-loop control, and anti-interference image processing modules. All peripherals, including the camera and industrial computer, are externally wall-mounted, eliminating the need for drilling, welding, or modification of the existing electrical control circuits on the crane's main beam and lower crossbeam, achieving completely non-destructive installation. The system seamlessly integrates with the crane's existing PLC control system, automatically completing non-contact measurement of extreme positions, real-time data monitoring, and automated over-limit early warning processes, effectively reducing on-site deployment costs and equipment downtime for maintenance.

[0080] 4. Strong environmental adaptability, completely eliminating safety hazards of manual high-altitude measurement.

[0081] This system is suitable for complex on-site conditions such as ports and hydropower stations with high dust, high humidity, and large temperature differences. Its core testing equipment meets high protection requirements. Combined with histogram equalization, distortion correction, and ROI dynamic area detection technologies, it can operate stably and reliably across a wide range of lighting and temperatures. The entire measurement process eliminates the need for personnel to climb to high altitudes or enter hazardous areas of tracks for manual measurement, completely avoiding safety risks such as falls from heights and being struck by objects. The system's testing efficiency and data reliability are far superior to traditional manual measurement methods, making it highly valuable for engineering application.

[0082] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0083] Figure 1 Diagram of a non-contact measurement system for the limit positions of a crane spreader under dynamic calibration;

[0084] Figure 2 This is an enlarged schematic diagram of the combined structure of the lifting device's characteristic markers and the calibration target.

[0085] Explanation of reference numerals in the attached figures:

[0086] 1-Lower crossbeam (1-1 Lower left crossbeam, 1-2 Lower right crossbeam); 2-Left industrial camera; 3-Right industrial camera; 4-Trolley; 5-Wire rope; 6-Lifting device; 7-Lifting device feature marker; 8-ROI detection area; 9-Fixed calibration checkerboard; 10-PLC controller; 11-Industrial computer; 12-Left limit switch; 13-Right limit switch. Detailed Implementation

[0087] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0088] Combination Figure 1A non-contact measurement system for the limit positions of crane spreaders with dynamic calibration is used in gantry cranes. The system includes a binocular vision unit, a calibration unit, and a control unit.

[0089] The gantry crane includes a lower crossbeam 1, a trolley 4, a wire rope 5, and a lifting device 6. The trolley 4 is located at the upper part and can move between the left and right extreme positions. The lower part of the trolley 4 is connected to the lifting device 6 through the wire rope 5. The lower crossbeam 1 includes a left lower crossbeam 1-1 and a right lower crossbeam 1-2.

[0090] The calibration unit includes a lifting device feature marker 7 and two sets of fixed calibration checkerboard grids 9. The lifting device feature marker 7 is fixed in the unobstructed center area of ​​the lifting device 6 and moves synchronously with the lifting device 6; combined with Figure 2 The lifting device feature marker 7 adopts a composite structure of a central fluorescent circular area and an outer auxiliary checkerboard grid. The central fluorescent circular area is a high-precision measurement target surface, and the outer auxiliary checkerboard grid is used to provide auxiliary corner point features and does not participate in camera calibration. Two sets of fixed calibration checkerboard grids 9 are fixedly installed on the opposite inner sides of the lower left crossbeam 1-1 and the lower right crossbeam 1-2, respectively, and are specifically used for online dynamic calibration of the camera.

[0091] The binocular vision unit includes a left industrial camera 2 and a right industrial camera 3. The two industrial cameras are symmetrically arranged, with their optical axes tilted upwards and pointing towards the working area of ​​the lifting device 6, for synchronously acquiring images of the lifting device feature markers 7 and the fixed calibration checkerboard 9.

[0092] The control unit is used to realize real-time image data processing, three-dimensional coordinate calculation, and control the operation of the crane.

[0093] Specifically, the control unit includes an industrial computer 11 and a PLC controller 10. The industrial computer 11 is used for real-time image data processing and three-dimensional coordinate calculation. The PLC controller 10 is used to receive instructions from the industrial computer 11 and control the operation of the crane. The industrial computer 11 and the PLC controller 10 establish a high-speed data link through an industrial communication protocol. The PLC controller 10 is installed in the electrical box of the lower crossbeam 1 and is used to receive limit switch signals and control the operation and braking of the trolley 4. The industrial computer 11 is installed in the internal electrical cabinet of the trolley 4.

[0094] The trolley 4 is equipped with a left limit switch 12 and a right limit switch 13 at both ends of the track. These switches are used to trigger the trolley 4 to reach the left and right limit positions. The signals are collected by the PLC controller 10 and used to control the trolley 4 to decelerate and stop.

[0095] Specifically, the left industrial camera 2 is rigidly mounted on the lower left crossbeam 1-1 of the crane, and the right industrial camera 3 is rigidly mounted on the lower right crossbeam 1-2 of the crane.

[0096] A measurement method for a dynamically calibrated non-contact measurement system for the limit positions of crane spreaders includes the following steps:

[0097] Step 1: System initialization and parameter configuration;

[0098] Step 2: Online dynamic calibration;

[0099] Step 3: Image acquisition and validity verification at extreme positions;

[0100] Step 4: Mixed matching and center point pixel coordinate extraction;

[0101] Step 5: 3D reconstruction and limit position determination.

[0102] Specifically, step 1 includes:

[0103] Camera Intrinsic Calibration: Before system installation, an independent standard checkerboard (not the fixed calibration checkerboard 9 in this system) is used to calibrate the intrinsic parameters of the left industrial camera 2 and the right industrial camera 3. During calibration, the independent standard checkerboard is placed in front of the left industrial camera 2 and the right industrial camera 3 within a clear field of view, and their spatial orientation is changed (e.g., translation, rotation, tilt). Multiple first images with different orientations are acquired, and their respective intrinsic parameter matrices and radial distortion coefficients k1 and k2 are solved. The intrinsic parameter matrices include the pixel focal length f. x ,f y Principal point coordinates (c x ,c y This intrinsic parameter matrix and distortion coefficients remain constant in all subsequent measurements;

[0104] Left and right camera relative extrinsic parameter calibration: The fixed spatial transformation relationship between the two fixed calibration checkerboard grids 9 on the left and right is determined by a high-precision measuring instrument (such as a total station). Combined with the triggering of the industrial computer 11 to trigger the left industrial camera 2 and the right industrial camera 3 to synchronously acquire a second image containing the fixed calibration checkerboard grid 9 on the opposite side, the fixed relative extrinsic parameters between the left industrial camera 2 and the right industrial camera 3 are calculated by the PnP (Perspective-n-Point) algorithm. The fixed relative extrinsic parameters include the baseline distance b.

[0105] World coordinate system definition: The world coordinate system is defined with the geometric center of the lower beam 1 of the gantry crane as the origin: the X-axis is along the track direction of the trolley 4, i.e., the left and right movement direction of the spreader 6, the Y-axis is perpendicular to the track direction of the trolley 4 (the movement direction of the main trolley), and the Z-axis is vertically upward. The three-dimensional coordinates of the corner points of the fixed calibration checkerboard 9 in the world coordinate system are pre-determined and stored. Template image pre-storage and feature library construction: During the system installation and debugging phase, the template acquisition program is executed: the spreader 6 is controlled to move to the center area of ​​the common field of view of the left industrial camera 2 and the right industrial camera 3. Under the condition that the crane is stationary and the lighting is uniform, the left industrial camera 2 and the right industrial camera 3 simultaneously acquire template images containing the spreader feature marker 7. The central fluorescent area of ​​the spreader feature marker 7 in the images of the left industrial camera 2 and the right industrial camera 3 is extracted as the matching template, and the ORB feature library corresponding to the industrial camera 2 and the right industrial camera 3 is simultaneously established for subsequent rapid matching.

[0106] This step aims to establish the camera's internal property model and the system's initial spatial geometry. It is usually performed after the system's initial installation or major overhaul, and the calibration results are stored in a configuration file for later use.

[0107] Specifically, step 2 includes:

[0108] Image acquisition: When the crane is stationary, the industrial computer 11 triggers the left industrial camera 2 and the right industrial camera 3 to synchronously acquire a third image containing the fixed calibration checkerboard 9 on the opposite side in real time;

[0109] Calculate real-time extrinsic parameters: Based on the world coordinates of the corner points of the fixed calibration chessboard 9 pre-stored in step 1, combined with the pixel coordinates extracted from the third image, the current real-time absolute extrinsic parameters of the left industrial camera 2 and the right industrial camera 3 are solved by the PnP (Perspective-n-Point) algorithm, including the rotation matrix R and the translation vector T.

[0110] Calibration result storage: Store the current real-time absolute extrinsic parameters of the left and right industrial cameras calculated in step 2, the intrinsic parameter matrix calibrated in step 1, and the radial distortion coefficients k1 and k2 into the industrial control computer 11.

[0111] Step 1 calibrates the fixed relative extrinsic parameters between the left and right cameras, which do not change over time. Step 2, on the other hand, calibrates the absolute extrinsic parameters between the left industrial camera 2 and the right industrial camera 3, which change due to vibration and temperature drift. The two have different purposes: the fixed relative extrinsic parameters are used to unify the absolute extrinsic parameters to the world coordinate system, while the absolute extrinsic parameters are used for distortion correction and stereo correction before this measurement.

[0112] This step is used to eliminate camera pose drift caused by vibration and temperature changes, ensuring measurement accuracy.

[0113] Specifically, step 3 includes:

[0114] The industrial computer 11 drives the trolley 4 to move to the left through the PLC controller 10. The trolley 4 drives the lifting device 6 to move synchronously via the wire rope 5. When the trolley 4 reaches the leftmost position allowed by the machinery (when the left limit switch 12 installed by the original crane is triggered), the PLC controller 10 immediately issues a deceleration and stop command. The trolley 4 stops running. After the PLC controller 10 detects that the trolley 4 has stopped running and is not moving, it continues to wait for the lifting device 6 to stop swinging completely. The stationary position of the lifting device 6 at this time is the left limit position of the trolley 4's travel, which is also the left limit point of the lifting device measured in this measurement. After that, the industrial computer 11 triggers the left industrial camera 2 and the right industrial camera 3 to synchronously acquire a complete fourth image.

[0115] The industrial computer 11 drives the trolley 4 to move to the right through the PLC controller 10. The trolley 4 drives the lifting device 6 to move synchronously via the wire rope 5. When the trolley 4 reaches the rightmost position allowed by the machinery (when the original right limit switch 13 of the crane is triggered), the PLC controller 10 immediately issues a deceleration and stop command. The trolley 4 stops running. After the PLC controller 10 detects that the trolley 4 has stopped running and is not moving, it continues to wait for the lifting device 6 to stop swinging completely. The stationary position of the lifting device 6 at this time is the right limit position of the trolley 4's stroke, which is also the right limit point of the lifting device measured in this measurement. After that, the industrial computer 11 triggers the left industrial camera 2 and the right industrial camera 3 to synchronously acquire a complete fifth image.

[0116] The complete fourth and fifth images acquired must simultaneously include: 1. Lifting device feature marker 7: serving as the measurement benchmark for calculating the spatial position of the lifting device in subsequent steps; 2. Fixed calibration checkerboard 9 on the opposite side: serving as a reference benchmark for verifying the validity of the online calibration parameters obtained in step 2;

[0117] Extract the pixel coordinates of all corner points of the fixed calibration checkerboard 9 in the fourth and fifth images respectively. Using the camera absolute extrinsic parameters obtained in step 2, reproject the corner point world coordinates pre-stored in step 1 onto the fourth and fifth images respectively. Calculate the root mean square error of reprojection, i.e., the RMS error.

[0118] If the RMS errors of the fourth and fifth images do not exceed the preset threshold, the camera absolute extrinsic parameters obtained from the online dynamic calibration in step 2 are deemed valid, and step 4 is executed. If the RMS error of at least one of the fourth and fifth images exceeds the threshold, the camera absolute extrinsic parameters are deemed to have drifted, the corresponding image with the RMS error exceeding the threshold is discarded, and step 2 is returned to be re-performed for online dynamic calibration. Then step 3 is executed again until the RMS errors of the fourth and fifth images do not exceed the preset threshold.

[0119] Specifically, step 4 includes:

[0120] Distortion Correction: Using the radial distortion coefficients k1 and k2 calibrated in step 1, distortion correction is performed on each pixel in the fourth and fifth images, as shown in the following formula:

[0121]

[0122]

[0123] Where (x, y) are the original image pixel coordinates of the fourth and fifth images, (x... 修正 ,y 修正 ) represents the pixel coordinates after distortion correction, (c x ,c y The coordinates of the principal point obtained in step 1.1 are r. 2 =(xc x ) 2 +(yc) y ) 2 r 4 =(r 2 ) 2 ;

[0124] The fourth image is corrected to obtain the sixth image, and the fifth image is corrected to obtain the seventh image;

[0125] ROI Detection Region Delineation: In the sixth and seventh images, based on the trajectory of the hoist 6, a ROI detection region 8 is delineated that contains only the hoist feature marker 7. This ROI detection region 8 spatially avoids the area where the fixed calibration checkerboard 9 is located on the opposite side, eliminating interference from the checkerboard corners on marker recognition. The ROI detection region 8 is a virtual search range delineated during image processing, not a physical structure. Figure 1 The dashed box in the middle is used to define the feature recognition area and suppress background interference;

[0126] Adaptive Binarization: Considering the characteristic that the center of the lifting device feature marker 7 is a high-reflectivity fluorescent circular area, the system uses the Otsu algorithm (maximum inter-class variance method) to segment the image within the ROI detection area 8, analyzes the grayscale histogram, finds the optimal threshold to separate the foreground (marker) from the background, sets pixels above the threshold to white and pixels below the threshold to black, generates a binarized image, and highlights the bright features of the lifting device feature marker 7;

[0127] SSDA coarse matching: Based on the binarized image, the Sequential Similarity Detection (SSDA) algorithm is used for rapid localization. The template image of the pre-stored lifting device feature marker 7 is called, and the matching is performed block by block in the ROI detection area 8 in a sliding window manner. For each window position, the cumulative absolute error of pixel grayscale between the template image and the image block covered by the current window is calculated. Once the cumulative absolute error exceeds the preset threshold, the calculation of the current window is immediately terminated, and the process moves to the starting position of the next sliding window to continue matching. After traversing all windows, the top K candidate regions with the smallest cumulative absolute error are selected (K is determined by the system debugging stage based on the balance between matching success rate and computational efficiency. In this embodiment, K=3 is taken). This step can quickly select a small number of high-confidence candidate regions from a large number of background regions, which greatly reduces the amount of subsequent calculations.

[0128] ORB fine matching: For the candidate regions selected by SSDA, the ORB (Oriented FAST and Rotated BRIEF) algorithm is used for precise verification, extracting the feature point descriptors of the candidate regions, and calculating the Hamming distance with the template features in the ORB feature library pre-stored in step 1.

[0129] The candidate region with the smallest Hamming distance is selected as the final matching result. If the smallest Hamming distance is less than the preset threshold, the matching is considered successful; otherwise, the matching is considered to have failed. The current frame data is discarded and the process returns to step 3 to re-acquire the image of the extreme position on that side. If the number of consecutive failures reaches the preset threshold, the measurement is paused, a marker recognition failure is reported, and a prompt is made to check the light source or the visibility of the marker.

[0130] Centroid coordinate calculation: After determining the successfully matched marker region, the precise pixel coordinates of the center point are calculated using the grayscale centroid method.

[0131] Left industrial camera 2:

[0132]

[0133] Right industrial camera 3:

[0134]

[0135] ∑X 左 : Represents the sum of the x-coordinates of all white pixels in the sixth image;

[0136] ∑Y 左 : Represents the sum of the y-coordinates of all white pixels in the sixth image;

[0137] M1: Represents the total number of white pixels in the sixth image;

[0138] ∑X 右: Represents the sum of the x-coordinates of all white pixels in the seventh image;

[0139] ∑Y 右 : Represents the sum of the y-coordinates of all white pixels in the seventh image;

[0140] M2: Represents the total number of white pixels in the seventh image;

[0141] Output: The calculated (x1,y1) and (x2,y2) are the precise pixel coordinates of the center point of the lifting device feature marker 7 in the photos of the left industrial camera 2 and the right industrial camera 3. These two coordinates will directly correspond to the pixel positions of the lifting device 6 at the left and right extreme positions, and will be used for the three-dimensional reconstruction in the subsequent step 6.

[0142] Specifically, step 5 includes:

[0143] 3D Reconstruction Based on Triangulation: Using the pixel coordinates (x1, y1) and (x2, y2) of the center point of the lifting device feature marker 7 in the left and right images output in step 4 (assuming y1=y2), the pixel coordinates are converted to physical coordinates (X0, Y0, Z0) in the world coordinate system. Let P be the measured point on the lifting device feature marker 7, and its pixel coordinates on the imaging planes of the left industrial camera 2 and the right industrial camera 3 are (x1, y1) and (x2, y2) respectively. Using the camera intrinsic parameter matrix calibrated in step 1, the 3D coordinates are calculated according to the following formula:

[0144]

[0145] Parameter description:

[0146] x1-x2: Parallax, which is the horizontal pixel offset of the same feature point in the left and right images. Parallax is inversely proportional to depth. The larger the parallax, the closer to the camera.

[0147] b: Baseline distance, which is the physical distance between the optical centers of the left industrial camera 2 and the right industrial camera 3, is obtained by calibration in step 1;

[0148] Multiple sampling and extreme value determination at extreme positions: Due to mechanical vibration and wire rope swing when the trolley 4 stops at the extreme position, the X0 value of a single measurement may fluctuate. To ensure the effectiveness of the safety boundary, multiple independent measurements are performed on the left and right extreme positions respectively. During continuous measurements, the three-dimensional coordinate calculations of steps 4 and 5 are fully executed for each measurement. Then, extreme value screening is performed.

[0149] Left Limit Determination: When the trolley 4 reaches the left limit position, perform N measurements and take the minimum value of all X0 values, denoted as X. 0min Simultaneously, take the average of all Y0 and Z0 values, denoted as... 0、 0, the world coordinates of the left extreme position are obtained as (X 0min , 0, 0);

[0150] Right-side limit determination: When the trolley 4 reaches the right-side limit position, perform N measurements, and take the maximum value of all X0 values, denoted as X. 0max， Simultaneously, take the average of all Y0 and Z0 values, denoted as... 0、 0, the world coordinates of the extreme position on the right are (X 0max , 0, 0);

[0151] Finally, output the world coordinates (X) of the left extreme position. 0min , 0, 0) and the world coordinates (X) of the right extreme position. 0max , 0, 0). Where X 0min and X 0max These are the horizontal coordinates of the left and right limits of the lifting device 6 along the track direction, which serve as the direct basis for subsequent safety control. The industrial control computer 11 transmits these coordinates to the PLC controller 10 via an industrial communication protocol to achieve a safety interlock control that automatically decelerates when the lifting device approaches its limits and automatically stops when it reaches its limits.

[0152] This invention does not require modification to the original structure of the crane and can achieve high-precision, real-time, non-contact measurement of the limit position of the lifting device. It solves the problems of low accuracy, poor safety, and low efficiency of traditional manual measurement. At the same time, it has anti-strong light interference and anti-vibration performance, can adapt to complex on-site working conditions, significantly improves the efficiency and data reliability of special equipment inspection and testing, and provides technical support for the safe operation of gantry cranes.

[0153] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A non-contact measurement system for the limit positions of a crane spreader, characterized in that, Includes a binocular vision unit, a calibration unit, and a control unit; The crane includes a lower crossbeam (1), a trolley (4), a wire rope (5), and a lifting device (6). The trolley (4) is located at the top and can move between the left and right extreme positions. The lower part of the trolley (4) is connected to the lifting device (6) by the wire rope (5). The lower crossbeam (1) includes a left lower crossbeam (1-1) and a right lower crossbeam (1-2). The calibration unit includes a lifting device feature marker (7) and two sets of fixed calibration checkerboard grids (9). The lifting device feature marker (7) is fixed on the lifting device (6). The lifting device feature marker (7) adopts a composite structure of a central fluorescent circular area and an outer auxiliary checkerboard grid. The central fluorescent circular area is a high-precision measurement target surface, and the outer auxiliary checkerboard grid is used to provide auxiliary corner point features. The two sets of fixed calibration checkerboard grids (9) are respectively fixedly installed on the opposite inner sides of the lower left crossbeam (1-1) and the lower right crossbeam (1-2). The binocular vision unit includes a left industrial camera (2) and a right industrial camera (3). The two industrial cameras are arranged symmetrically, with the optical axis tilted upwards and pointing towards the working area of ​​the lifting device (6) to simultaneously acquire images of the lifting device's characteristic markers (7) and the fixed calibration checkerboard (9). The control unit is used to realize real-time image data processing, three-dimensional coordinate calculation, and control the operation of the crane.

2. The non-contact measurement system for the limit positions of crane lifting devices with dynamic calibration according to claim 1, characterized in that, The control unit includes an industrial computer (11) and a PLC controller (10). The industrial computer (11) is used for real-time image data processing and three-dimensional coordinate calculation. The PLC controller (10) is used to receive instructions from the industrial computer (11) and control the operation of the crane. The industrial computer (11) and the PLC controller (10) establish a high-speed data link through an industrial communication protocol.

3. The non-contact measurement system for the limit position of a crane lifting device with dynamic calibration according to claim 2, characterized in that, The left industrial camera (2) is rigidly mounted on the lower left crossbeam (1-1) of the crane, and the right industrial camera (3) is rigidly mounted on the lower right crossbeam (1-2) of the crane.

4. A measurement method for a non-contact measurement system for the limit position of a crane lifting device with dynamic calibration according to claim 2 or 3, characterized in that, Includes the following steps: Step 1: System initialization and parameter configuration; Step 2: Online dynamic calibration; Step 3: Image acquisition and validity verification at extreme positions; Step 4: Mixed matching and center point pixel coordinate extraction; Step 5: 3D reconstruction and limit position determination.

5. The measurement method according to claim 4, characterized in that, Step 1 specifically includes: Camera Intrinsic Calibration: An independent standard checkerboard grid was used to calibrate the intrinsic parameters of the left industrial camera (2) and the right industrial camera (3) respectively. During calibration, the independent standard checkerboard grid was placed in front of the left industrial camera (2) and the right industrial camera (3) within the field of view where clear imaging was possible. The spatial orientation of the checkerboard grid was changed, and multiple first images with different orientations were acquired. The intrinsic parameter matrices and radial distortion coefficients k1 and k2 were solved for each camera. The intrinsic parameter matrices include the pixel focal length f. x ,f y Principal point coordinates (c x ,c y ); Left and right camera relative extrinsic calibration: Determine the fixed spatial transformation relationship between the two fixed calibration checkerboard grids (9) on the left and right sides, and combine the industrial control computer (11) to trigger the left industrial camera (2) and the right industrial camera (3) to synchronously acquire the second image containing the fixed calibration checkerboard grid (9) on the opposite side, and calculate the fixed relative extrinsic parameters between the left industrial camera (2) and the right industrial camera (3), wherein the fixed relative extrinsic parameters include the baseline distance b; World coordinate system definition: The world coordinate system is defined with the geometric center of the following beam (1) as the origin: the X-axis is along the track direction of the trolley (4), that is, the left and right movement direction of the lifting device (6), the Y-axis is perpendicular to the track direction of the trolley (4), and the Z-axis is vertically upward. The three-dimensional coordinates of the corner points of the fixed calibration checkerboard (9) in the world coordinate system are pre-determined and stored. Template image pre-storage and feature library construction: During the system installation and debugging stage, the template acquisition program is executed: the lifting device (6) is controlled to run to the common field of view center area of ​​the left industrial camera (2) and the right industrial camera (3). Under the condition that the crane is stationary and the lighting is uniform, the template image containing the lifting device feature marker (7) is simultaneously acquired by the left industrial camera (2) and the right industrial camera (3). The central fluorescent area of ​​the lifting device feature marker (7) in the images of the left industrial camera (2) and the right industrial camera (3) is extracted as the matching template. The ORB feature library corresponding to the industrial camera (2) and the right industrial camera (3) is simultaneously established for subsequent fast matching.

6. The measurement method according to claim 5, characterized in that, Step 2 specifically includes: Image acquisition: When the crane is stationary, the industrial computer (11) triggers the left industrial camera (2) and the right industrial camera (3) to synchronously acquire a third image containing the fixed calibration checkerboard (9) on the opposite side in real time; Calculate real-time extrinsic parameters: Based on the world coordinates of the corner points of the fixed calibration chessboard (9) pre-stored in step 1, and combined with the pixel coordinates extracted from the third image, solve the current real-time absolute extrinsic parameters of the left industrial camera (2) and the right industrial camera (3), including the rotation matrix R and the translation vector T. Calibration result storage: Store the current real-time absolute external parameters of the left and right industrial cameras calculated in step 2, the internal parameter matrix calibrated in step 1, and the radial distortion coefficients k1 and k2 into the industrial control computer (11).

7. The measurement method according to claim 6, characterized in that, Step 3 specifically includes: The industrial computer (11) drives the trolley (4) to move to the left through the PLC controller (10). The trolley (4) drives the lifting device (6) to move synchronously through the wire rope (5). When the trolley (4) has reached the leftmost position allowed by the machinery, the PLC controller (10) immediately issues a deceleration and stop command. The trolley (4) stops running. After the PLC controller (10) detects that the trolley (4) has stopped running and is not moving, it continues to wait for the lifting device (6) to stop swinging completely. At this time, the stationary position of the lifting device (6) is the left limit position of the trolley (4)'s travel, which is also the left limit point of the lifting device measured in this measurement. After that, the industrial computer (11) triggers the left industrial camera (2) and the right industrial camera (3) to synchronously acquire a complete fourth image. The industrial computer (11) drives the trolley (4) to move to the right through the PLC controller (10). The trolley (4) drives the lifting device (6) to move synchronously through the wire rope (5). When the trolley (4) has reached the rightmost position allowed by the machinery, the PLC controller (10) immediately issues a deceleration and stop command. The trolley (4) stops running. After the PLC controller (10) detects that the trolley (4) has stopped running and is not moving, it continues to wait for the lifting device (6) to stop swinging completely. At this time, the stationary position of the lifting device (6) is the right limit position of the trolley (4)'s travel, which is also the right limit point of the lifting device measured in this measurement. After that, the industrial computer (11) triggers the left industrial camera (2) and the right industrial camera (3) to synchronously acquire a complete fifth image. The complete fourth and fifth images acquired must simultaneously include:

1. a lifting device feature marker (7): serving as a measurement benchmark for calculating the spatial position of the lifting device in subsequent steps; 2. a fixed calibration checkerboard grid on the opposite side (9): serving as a reference benchmark for verifying the validity of the online calibration parameters obtained in step 2; Extract the pixel coordinates of all corner points of the fixed calibration checkerboard (9) in the fourth and fifth images respectively. Using the absolute extrinsic parameters of the camera obtained in step 2, reproject the world coordinates of the corner points stored in step 1 onto the fourth and fifth images respectively. Calculate the root mean square error of reprojection, i.e., the RMS error. If the RMS errors of the fourth and fifth images do not exceed the preset threshold, the camera absolute extrinsic parameters obtained from the online dynamic calibration in step 2 are deemed valid, and step 4 is executed. If the RMS error of at least one of the fourth and fifth images exceeds the threshold, the camera absolute extrinsic parameters are deemed to have drifted, the corresponding image with the RMS error exceeding the threshold is discarded, and step 2 is returned to be re-performed for online dynamic calibration. Then step 3 is executed again until the RMS errors of the fourth and fifth images do not exceed the preset threshold.

8. The measurement method according to claim 7, characterized in that, Step 4 specifically includes: Distortion Correction: Using the radial distortion coefficients k1 and k2 calibrated in step 1, distortion correction is performed on each pixel in the fourth and fifth images, as shown in the following formula: Where (x, y) are the original image pixel coordinates of the fourth and fifth images, (x... 修正 ,y 修正 ) represents the pixel coordinates after distortion correction, (c x ,c y The coordinates of the principal point obtained in step 1.1 are r. 2 =(xc x ) 2 +(yc) y ) 2 r 4 =(r 2 ) 2 ; The fourth image is corrected to obtain the sixth image, and the fifth image is corrected to obtain the seventh image; ROI detection area delineation: In the sixth and seventh images, an ROI detection area (8) containing only the feature markers (7) of the hoist (6) is delineated based on the running trajectory of the hoist (6). This ROI detection area (8) avoids the area where the fixed calibration checkerboard (9) is located on the opposite side in a spatial manner, thus eliminating the interference of checkerboard corner points on marker recognition. Adaptive binarization: In view of the characteristic that the center of the lifting device feature marker (7) is a high reflectivity fluorescent circular area, the system uses the Otsu algorithm to segment the image in the ROI detection area (8), analyzes the gray histogram, finds the best threshold to separate the foreground and background, sets the pixels above the threshold to white and the pixels below the threshold to black, generates a binarized image, and highlights the bright features of the lifting device feature marker (7). SSDA coarse matching: Based on the binarized image, the sequential similarity detection algorithm (SSDA) is used for fast localization. The template image of the pre-stored lifting device feature marker (7) is called, and the matching is performed block by block in the ROI detection area (8) in a sliding window manner. For each window position, the cumulative absolute error of pixel gray level between the template image and the image block covered by the current window is calculated. Once the cumulative absolute error exceeds the preset threshold, the calculation of the current window is immediately terminated, and the matching is moved to the starting position of the next sliding window to continue. After traversing all windows, the top K candidate regions with the smallest cumulative absolute error are selected. ORB fine matching: For the candidate regions selected by SSDA, the ORB (Oriented FAST and Rotated BRIEF) algorithm is used for precise verification, extracting the feature point descriptors of the candidate regions, and calculating the Hamming distance with the template features in the ORB feature library pre-stored in step 1. The candidate region with the smallest Hamming distance is selected as the final matching result. If the smallest Hamming distance is less than the preset threshold, the matching is considered successful; otherwise, the matching is considered to have failed. The current frame data is discarded and the process returns to step 3 to re-acquire the image of the extreme position on that side. If the number of consecutive failures reaches the preset threshold, the measurement is paused, a marker recognition failure is reported, and a prompt is made to check the light source or the visibility of the marker. Centroid coordinate calculation: After determining the successfully matched marker region, the precise pixel coordinates of the center point are calculated using the grayscale centroid method. Left industrial camera (2): Right industrial camera (3): ∑X 左 : Represents the sum of the x-coordinates of all white pixels in the sixth image; ∑Y 左 : Represents the sum of the y-coordinates of all white pixels in the sixth image; M1: Represents the total number of white pixels in the sixth image; ∑X 右 : Represents the sum of the x-coordinates of all white pixels in the seventh image; ∑Y 右 : Represents the sum of the y-coordinates of all white pixels in the seventh image; M2: Represents the total number of white pixels in the seventh image; Output: The calculated (x1,y1) and (x2,y2) are the precise pixel coordinates of the center point of the lifting device feature marker (7) in the photos of the left industrial camera (2) and the right industrial camera (3). These two coordinates will directly correspond to the pixel positions of the lifting device (6) at the left and right extreme positions, and will be used for the three-dimensional reconstruction in the subsequent step 6.

9. The measurement method according to claim 8, characterized in that, The number of candidate regions K in step 4 is 3.

10. The measurement method according to claim 8 or 9, characterized in that, Step 5 specifically includes: 3D Reconstruction Based on Triangulation: Using the pixel coordinates (x1, y1) and (x2, y2) of the center point of the lifting feature marker (7) in the left and right images output in step 4, the pixel coordinates are converted into physical coordinates (X0, Y0, Z0) in the world coordinate system. Let P be the measured point on the lifting feature marker (7), and its pixel coordinates on the imaging planes of the left industrial camera (2) and the right industrial camera (3) are (x1, y1) and (x2, y2) respectively. Using the camera intrinsic parameter matrix calibrated in step 1, the 3D coordinates are calculated according to the following formula: Parameter description: x1-x2: disparity, which is the horizontal pixel offset of the same feature point in the left and right images. Disparity is inversely proportional to depth. The larger the disparity, the closer to the camera. b: Baseline distance, which is the physical distance between the optical centers of the left industrial camera (2) and the right industrial camera (3), is obtained by calibration in step 1; Multiple sampling and extreme value determination at extreme positions: Due to mechanical vibration and wire rope swing when the trolley (4) stops at the extreme position, the X0 value of a single measurement may fluctuate. To ensure the effectiveness of the safety boundary, multiple independent measurements are performed on the left and right extreme positions respectively. During continuous measurement, the three-dimensional coordinate calculations of steps 4 and 5 are performed completely for each measurement. Then, extreme value screening is performed. Left limit determination: When the trolley (4) reaches the left limit position, perform N measurements, take the minimum value of all X0 values, and record it as X. 0min Simultaneously, take the average of all Y0 and Z0 values, denoted as... 0、 0, the world coordinates of the left extreme position are obtained as (X 0min , 0, 0); Right limit determination: When the trolley (4) reaches the right limit position, perform N measurements and take the maximum value of all X0 values, which is recorded as X. 0max， Simultaneously, take the average of all Y0 and Z0 values, denoted as... 0、 0, the world coordinates of the extreme position on the right are (X 0max , 0, 0); Finally, output the world coordinates (X) of the left extreme position. 0min , 0, 0) and the world coordinates (X) of the right extreme position. 0max , 0, 0).

Images