Elevator leveling precision detection device based on binocular vision

Abstract

According to the binocular vision-based elevator leveling precision detection device disclosed by the invention, a detection task is automatically received through a robot body, manual triggering is not needed, automatic starting of the detection task is realized, time is greatly saved, and the overall detection efficiency is improved; the sensing module can collect environment and state data of the robot in real time and quickly judge detection conditions, the binocular vision detection module is started when the conditions are met, time delay of manual judgment is avoided, and detection is more compact and efficient; the binocular vision detection module obtains image information through double cameras, more accurately obtains three-dimensional information and reduces limitation and operation errors caused by manual use of a single measuring tool, meanwhile, data of the sensing module provides a comprehensive information basis for detection, judgment and analysis are carried out by comprehensively considering multiple factors, the influence of a single factor is avoided, and the detection accuracy is improved. And the detection result is more accurate and reliable.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of elevator leveling precision detection, and particularly relates to an elevator leveling precision detection device based on binocular vision. BACKGROUND

[0002] As an indispensable vertical transportation equipment in modern buildings, the safety and comfort of the elevator are of great importance. Among them, the elevator leveling precision is one of the key indicators to measure the performance of the elevator, which directly affects the convenience and safety of passengers getting on and off the elevator. If the elevator leveling precision is poor, it may cause passengers to stumble, the elevator door to fail to normally close or open, and even may cause safety accidents.

[0003] The traditional elevator leveling precision detection method mainly relies on manual operation. The detection personnel usually uses professional measuring tools such as laser range finder, level, etc. to manually measure when the elevator stops at each floor. This method has many drawbacks: on the one hand, manual detection is low in efficiency, and a large amount of time and manpower is needed, especially in high-rise buildings, when multiple floors need to be detected, the workload is huge; on the other hand, the accuracy of manual detection is easily affected by the skill level of the detection personnel, operation specification and subjective judgment, which may cause errors in the detection results. SUMMARY

[0004] Therefore, the present application provides an elevator leveling precision detection device based on binocular vision, which can effectively solve the defects of low detection efficiency and errors in the detection results in the prior art.

[0005] The technical scheme of the present application is as follows:

[0006] An elevator leveling precision detection device based on binocular vision, comprising:

[0007] A robot body for receiving an elevator leveling precision detection task;

[0008] A perception module for collecting environmental data and robot body state data in real time, and determining whether the detection conditions are met based on the environmental data and robot body state data;

[0009] A binocular vision detection module for detecting the precision of elevator leveling.

[0010] As a further optional scheme of the elevator leveling precision detection device based on binocular vision, the perception module comprises:

[0011] An environmental sensor for collecting environmental data;

[0012] A body sensor for collecting robot body state data;

[0013] a positioning unit configured to realize global positioning by using GPS or visual SLAM;

[0014] a construction unit configured to construct an environment map according to environment data, robot body state data and global positioning information;

[0015] a judgment unit configured to judge whether a detection condition is met according to the environment map.

[0016] As a further optional solution of the elevator leveling precision detection device based on binocular vision, the binocular vision detection module comprises:

[0017] a positioning floor unit configured to confirm a floor where the elevator is located by using the collected actual floor information of the elevator;

[0018] a camera alignment sill unit configured to adjust a camera angle in real time by using a visual algorithm to ensure that the sill is centered in the image;

[0019] an image acquisition unit configured to acquire an elevator shaft image;

[0020] an image preprocessing unit configured to preprocess the elevator shaft image to obtain a preprocessed elevator shaft image;

[0021] a feature extraction unit configured to locate key points of a car and a sill edge according to the preprocessed elevator shaft image;

[0022] a stereo optimization unit configured to generate a disparity map according to the key points of the car and the sill edge;

[0023] a disparity optimization unit configured to optimize the disparity map by using a post-processing algorithm;

[0024] an output leveling precision unit configured to process the optimized disparity map based on a binocular stereo vision three-dimensional positioning principle to calculate the leveling precision of a car sill and a landing sill.

[0025] As a further optional solution of the elevator leveling precision detection device based on binocular vision, the camera alignment sill unit comprises:

[0026] a sill line detection subunit configured to detect a sill line in real time based on a YOLOv5 model;

[0027] a PID closed-loop control subunit configured to adjust the camera angle according to the detected sill line to center the sill line.

[0028] As a further optional solution of the elevator leveling precision detection device based on binocular vision, the image preprocessing unit comprises:

[0029] The denoising processing subunit is configured to perform denoising processing on the elevator shaft image by using a median filter and an inter-frame difference denoising method.

[0030] The distortion correction subunit is configured to eliminate the distortion of the elevator shaft image.

[0031] As a further optional solution of the elevator flatness precision detection device based on binocular vision, the feature extraction unit comprises:

[0032] The edge detection subunit is configured to locate the continuous edge of the car sill based on a Canny operator detection method and a Sobel operator detection method.

[0033] The geometric feature extraction subunit is configured to identify the straight line segment of the sill by using a Hough transform method and extract the color region of the car and the landing door sill by using a color segmentation method.

[0034] As a further optional solution of the elevator flatness precision detection device based on binocular vision, the stereo optimization unit comprises:

[0035] The cost calculation subunit is configured to calculate the Census transform cost between the key points of the car and the sill edge by using an SGBM algorithm and generate an initial disparity map by using a pre-trained PSMNet network.

[0036] The cost aggregation subunit is configured to dynamically adjust the aggregation path length according to the gradient intensity of the sill edge, suppress the noise in the non-edge region of the initial disparity map, and obtain a final disparity map.

[0037] An elevator flatness precision detection method based on binocular vision, which is based on any one of the elevator flatness precision detection devices based on binocular vision, and specifically comprises:

[0038] The robot body receives the elevator flatness precision detection task.

[0039] The perception module collects environmental data and robot body state data in real time, and determines whether the detection condition is met based on the collected environmental data and robot body state data. If the detection condition is met, the binocular vision detection module is used to detect the precision of the elevator flatness. Otherwise, the elevator flatness precision detection operation is not performed.

[0040] A computing device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the elevator flatness precision detection method based on binocular vision when executing the computer program.

[0041] A computer readable storage medium, the storage medium has a computer program stored thereon, the computer program is executed by a processor to implement the steps of the above-mentioned elevator layering precision detection method based on binocular vision.

[0042] The robot body can automatically receive the elevator layering precision detection task without manual triggering and intervention in the detection process. The perception module can collect environmental data and robot body state data in real time and quickly judge whether the detection conditions are met based on these data. The binocular vision detection module detects the elevator layering precision using the binocular vision principle. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description.

[0044] Fig. 1 The figure is a composition schematic diagram of the elevator layering precision detection device based on binocular vision.

[0045] Fig. 2 The figure is a composition schematic diagram of the binocular vision detection module.

[0046] Fig. 3 The figure is a flowchart of the elevator layering precision detection method based on binocular vision.

[0047] Fig. 4 The figure is a composition schematic diagram of the computing device. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the protection scope of the present application.

[0049] Reference Figs. 1 to 4 A double-vision-based elevator leveling precision detection device, comprising a robot body, a perception module and a double-vision detection module, wherein:

[0050] The robot body is used to receive the elevator leveling precision detection task. In some embodiments, the robot body comprises:

[0051] The driving system is used to provide power for the robot to drive the joint or wheel group to move.

[0052] The control system is used to coordinate the work of each module, realize motion planning, balance control and real-time decision-making.

[0053] The mechanical structure system provides physical support and motion basis for the robot, and is suitable for different terrains.

[0054] The energy system provides continuous power support for the robot.

[0055] The communication and interaction module realizes human-computer interaction and remote control, including voice interaction, wireless communication and tactile feedback.

[0056] The safety and fault-tolerant module ensures the stable operation of the robot, and contains fault diagnosis and emergency braking functions.

[0057] The task planning module generates an optimal path and adapts to new tasks.

[0058] The data storage and processing module realizes the storage, calculation and processing of data.

[0059] Specifically, the driving system provides power for the robot to drive the joint or wheel group to move, and the mechanical structure system provides physical support and motion basis for the robot. The two systems work together to enable the robot to adapt to different terrains and complex environments in the elevator shaft. Whether the shaft is flat or has a certain slope and obstacles, the robot can move flexibly to ensure that it can reach each floor of the elevator for leveling precision detection, thereby expanding the applicable range of detection. The energy system provides continuous power support for the robot, ensuring that the robot will not interrupt work due to insufficient power during long-time and multi-floor detection tasks, so that the detection process can be continuously and stably carried out, improving the reliability of the completion of the detection task.

[0060] The control system coordinates the work of each module, realizes motion planning, balance control and real-time decision-making, and in the detection process, it can plan the robot's motion path according to the actual situation to ensure the robot moves safely and efficiently; at the same time, it performs balance control to ensure the stability of the robot in various postures; it can also make real-time decisions based on perception information, such as adjusting the detection strategy, making the detection process more intelligent; the task planning module can generate the optimal path and adapt to new tasks. When facing detection tasks on multiple floors, it can plan the most time-saving and resource-saving path to improve detection efficiency; and when the detection task changes, it can quickly adapt and re-plan, enhancing the robot's autonomy and ability to handle complex tasks.

[0061] The communication and interaction module realizes human-computer interaction and remote control, including voice interaction, wireless communication and tactile feedback; the operator can conveniently control the robot through voice instructions without complex operation procedures; the wireless communication function enables the operator to remotely monitor the working state of the robot and receive detection data, so that they can grasp the situation in time even if they are not at the detection site; tactile feedback enables the operator to more intuitively feel the working condition of the robot, improving the experience and efficiency of human-computer interaction.

[0062] The safety and fault-tolerant module contains fault diagnosis and emergency braking functions, which can monitor the running state of the robot in real time, and once a fault or abnormal condition is found, timely diagnosis and emergency braking measures are taken to avoid collisions, damage and other safety problems, ensuring the safety of the detection process and the stability of the robot.

[0063] The data storage and processing module realizes data storage, calculation and processing. During the detection process, the robot will collect a large amount of data, and this module can effectively store these data and extract valuable information through calculation and processing to support subsequent data analysis and report generation, which helps to more accurately evaluate the elevator leveling precision.

[0064] The perception module is used to collect environmental data and robot body state data in real time, and determine whether the detection condition is met based on the environmental data and robot body state data; in some embodiments, the perception module comprises:

[0065] The environmental sensor is used to collect environmental data, wherein the environmental sensor comprises a laser radar, a camera, an ultrasonic sensor, an inclination sensor and a voiceprint sensor;

[0066] The body sensor is used to collect robot body state data, wherein the body sensor comprises an inertial measurement unit, a force / torque sensor and a tactile sensor;

[0067] The positioning unit is used to realize global positioning by using GPS or visual SLAM.

[0068] a construction unit configured to construct an environment map according to the environment data, the robot body state data and the global positioning information;

[0069] a judgment unit configured to judge whether the detection condition is met according to the environment map.

[0070] Specifically, the perception module integrates multiple environmental sensors such as laser radar, camera, ultrasonic sensor, tilt sensor and voiceprint sensor; the laser radar can accurately measure the distance and construct a three-dimensional environment model; the camera can obtain rich visual information for identifying key elements such as elevator doors and floor signs; the ultrasonic sensor has advantages in close-range detection and obstacle detection; the tilt sensor can sense the inclination state of the robot and judge the flatness of the environment; the voiceprint sensor can capture the sound of the elevator operation to assist in judging the elevator state; the fusion use of multiple sensors realizes the all-around and multi-angle perception of the elevator shaft environment, providing comprehensive and accurate environmental data for subsequent detection;

[0071] The body sensor includes an inertial measurement unit, a force / torque sensor and a tactile sensor; the inertial measurement unit can monitor the motion posture and acceleration of the robot in real time to ensure the stability of the robot during movement; the force / torque sensor can sense the interaction force between the robot and the environment to avoid damage to the equipment due to excessive force or collision; the tactile sensor enables the robot to have a perception ability similar to human touch, which can sense the surface characteristics of the contacted object and other information; through accurate monitoring of the state of the robot itself, the motion and behavior of the robot can be adjusted in time to ensure the smooth progress of the detection process;

[0072] The positioning unit uses GPS or visual SLAM to realize global positioning; GPS can provide relatively accurate geographic position information in outdoor or open areas; visual SLAM constructs an environment map in real time and determines the position of the robot in the map through image information collected by the camera, which has better applicability in complex environments such as indoor elevator shafts; accurate global positioning provides a basis for navigation and path planning of the robot, enabling the robot to accurately reach the designated detection floor and position;

[0073] The construction unit constructs an environment map according to the environment data, the robot body state data and the global positioning information, providing the robot with overall cognition of the detection environment; the judgment unit judges whether the detection condition is met based on the constructed environment map, such as whether the elevator is in a stationary state, whether the shaft environment is safe, etc.; this intelligent judgment mechanism can avoid invalid detection in cases where the conditions are not met, improving the efficiency and accuracy of detection, and also ensuring the safety of the detection process.

[0074] a binocular vision detection module for detecting the precision of elevator leveling; in some embodiments, the binocular vision detection module comprises:

[0075] a positioning floor unit for confirming the floor where the elevator is located through the collected actual floor information of the elevator;

[0076] a camera alignment sill unit for adjusting the camera angle in real time through a visual algorithm to ensure that the sill is centered in the image;

[0077] an image acquisition unit for acquiring elevator shaft images;

[0078] an image preprocessing unit for preprocessing the elevator shaft images to obtain preprocessed elevator shaft images;

[0079] a feature extraction unit for locating the key points of the car and the sill edge based on the preprocessed elevator shaft images;

[0080] a stereo optimization unit for generating a disparity map based on the key points of the car and the sill edge;

[0081] a disparity optimization unit for optimizing the disparity map through a post-processing algorithm;

[0082] an output leveling precision unit for processing the optimized disparity map based on the binocular stereo vision three-dimensional positioning principle to calculate the leveling precision of the car sill and the landing sill.

[0083] Specifically, the camera alignment sill unit adjusts the camera angle in real time through a visual algorithm to ensure that the sill is always centered in the image. This helps to obtain clear, complete and key detection information containing elevator shaft images, reduces the problem of missing or blurred sill features caused by poor image acquisition angle, and improves the usability of the image;

[0084] The image acquisition unit is responsible for acquiring elevator shaft images, and the image preprocessing unit preprocesses the acquired images, such as denoising, contrast enhancement and other operations; the preprocessed image presents the details of the elevator shaft more clearly, especially the key parts such as the car and the sill edge, providing a high-quality data basis for subsequent feature extraction;

[0085] The feature extraction unit accurately locates the key points of the car and the sill edge based on the preprocessed elevator shaft images. These key points are an important basis for calculating the leveling precision. Accurate feature extraction can ensure the accuracy of subsequent disparity map generation and leveling precision calculation, making the detection result closer to the actual situation;

[0086] The stereo optimization unit generates a disparity map according to the key points of the car and the edge of the sill. The disparity map can intuitively reflect the depth information of the object in binocular vision. Through the processing of the stereo optimization unit, the disparity map can more accurately present the spatial relationship between the car and the sill, providing key data for the calculation of the leveling precision.

[0087] The disparity optimization unit optimizes the disparity map through a post-processing algorithm, further reducing the noise and errors in the disparity map, improving the accuracy and reliability of the disparity map. The optimized disparity map can more accurately reflect the actual depth information, thereby improving the precision of the leveling precision detection.

[0088] It should be noted that the output leveling precision unit processes the optimized disparity map based on the binocular stereo vision three-dimensional positioning principle to calculate the leveling precision of the car sill and the door sill. Specifically, it includes:

[0089] Obtain the three-dimensional coordinates of the door sill feature point A and the car sill feature point B;

[0090] The baseline distance of the two cameras is k, and the focal length is f;

[0091] The two cameras view the same feature point A of the door sill at the same time ( , , ), and obtain the image of point A on the left camera (L) and the right camera (R), respectively. Their image coordinates are = ( , ), =( , );

[0092] Now the two cameras are on the same plane of the image, so the Y coordinate of the image coordinates of feature point A is the same, that is = =Y;

[0093] From the triangular geometric relationship, we get: ;

[0094] Then the disparity (disparity d: refers to the horizontal pixel offset of the same spatial point in the left and right images) is: - ;

[0095] Thus, the three-dimensional coordinates of feature point A in the camera coordinate system are calculated as: ;

[0096] Similarly, the three-dimensional coordinates of point B ( , , ) can be obtained.

[0097] The horizontal precision of the car door sill and the landing door sill is h = h1 - h2. .

[0098] In some embodiments, the camera alignment sill unit comprises:

[0099] A sill line detection subunit for real-time detection of the sill line based on a YOLOv5 model;

[0100] A PID closed-loop control subunit for adjusting the camera angle based on the detected sill line to center the sill line.

[0101] Specifically, the sill line detection subunit uses the YOLOv5 model to detect the sill line in real time. The YOLOv5 model can quickly analyze the images captured by the camera in a short time and accurately identify the position and contour of the sill line. Compared with traditional image processing methods, it greatly improves the efficiency and accuracy of sill line detection, reduces the situation of missed detection and false detection, and provides a reliable foundation for subsequent camera angle adjustment.

[0102] The PID closed-loop control subunit adjusts the camera angle based on the detected sill line to center the sill line. The PID control algorithm has good stability, rapidity and accuracy. It can adjust the angle of the camera in real time and accurately according to the deviation between the actual position of the sill line and the target position (center position). Through continuous feedback and adjustment, it ensures that the camera is always aligned with the sill line, ensuring the stability and consistency of image acquisition, thereby providing high-quality image data for subsequent elevator horizontal precision detection.

[0103] In some embodiments, the image preprocessing unit comprises:

[0104] A denoising processing subunit for denoising the elevator shaft image using median filtering and inter-frame difference denoising methods;

[0105] A distortion correction subunit for eliminating the distortion of the elevator shaft image.

[0106] ​Specifically, the denoising processing subunit adopts a median filter and an inter-frame difference denoising method to perform denoising processing on the elevator shaft image; the median filter can effectively remove impulse noise such as salt and pepper noise in the image, and it replaces the original pixel value by sorting the pixel values in the neighborhood of the pixel point and taking the median value, which can better preserve the edge information of the image while removing noise; the inter-frame difference denoising method uses the difference between consecutive frames of images to eliminate random noise, and for an environment such as an elevator shaft which is relatively static but may have minor dynamic interference (such as light flickering), the image can be further purified; the image after denoising processing is clearer, reducing the interference of noise on subsequent feature extraction and precision calculation, and improving the reliability of image data.

[0107] The distortion correction subunit is used to eliminate the distortion of the elevator shaft image; during image acquisition, due to factors such as the optical properties of the camera lens, the image may produce radial distortion and tangential distortion, etc.; these distortions will cause the shape and position of objects in the image to be distorted, and if the elevator leveling precision is directly detected based on the distorted image, the feature extraction will be inaccurate, which will affect the final detection result; by correcting the image through the distortion correction subunit, the objects in the image can restore the true shape and position relationship, ensuring the accuracy of subsequent image-based measurement and calculation (such as sill line detection, disparity map generation, etc.), and providing protection for accurate calculation of elevator leveling precision.

[0108] In some embodiments, the feature extraction unit includes:

[0109] The edge detection subunit is used to locate the continuous edge of the car sill based on the Canny operator detection method and the Sobel operator detection method.

[0110] The geometric feature extraction subunit is used to identify the straight line segment of the sill using the Hough transform method, and extract the color region of the car and the landing door sill using the color segmentation method.

[0111] Specifically, the edge detection subunit locates the continuous edge of the car sill based on the Canny operator and Sobel operator detection methods; the Canny operator is a multi-stage optimization algorithm with low error rate, high positioning accuracy, and single edge response, which can effectively suppress noise and accurately detect the edges in the image; the Sobel operator responds well to step edges in digital images, and the calculation is relatively simple, which can quickly detect the gray scale changes of the image and thus locate the edges; the combination of the two can more comprehensively and accurately detect the continuous edge of the car sill, providing accurate edge information for subsequent leveling precision calculation.

[0112] The geometric feature extraction subunit adopts a Hough transform method to identify the straight line segment of the sill; the Hough transform is a classical method for detecting straight lines, circles and other geometric shapes in an image, and it has good robustness to noise and partial occlusion in the image. Through the Hough transform, the straight line segment feature of the sill can be accurately identified, and even if the sill image has certain noise or local blur, the straight line information can be effectively extracted, thereby providing a reliable basis for determining the position and shape of the sill; the color segmentation method is used to extract the color region of the car and the landing door sill. In the elevator shaft image, the car and the landing door sill usually have specific color characteristics, and through color segmentation, they can be distinguished from other background regions, which helps to further clarify the detection target, reduces background interference, makes the subsequent calculation and analysis more focused on the relevant features of the car and the landing door sill, and improves the accuracy and pertinence of detection.

[0113] In some embodiments, the stereo optimization unit comprises:

[0114] a cost calculation subunit configured to calculate Census transform costs between feature points using an SGBM algorithm and aggregate matching costs along eight directions, and extract multi-scale context features to generate an initial disparity map in combination with a pre-trained PSMNet network;

[0115] a cost aggregation subunit configured to dynamically adjust the aggregation path length according to the gradient intensity of the sill edge, suppress the noise in the non-edge region of the initial disparity map, and obtain a final disparity map.

[0116] Specifically, the cost calculation subunit calculates Census transform costs between feature points using an SGBM (Semi-Global Block Matching) algorithm and aggregates matching costs along eight directions. The SGBM algorithm can more comprehensively capture the disparity information in the image by considering local information and aggregating costs in multiple directions, effectively reducing the influence of local noise and mismatching. In combination with a pre-trained PSMNet (Pyramid Stereo Matching Network) network to extract multi-scale context features, PSMNet can extract feature information of different scales using a pyramid structure, better handle various details and overall structures in the image, and thus generate more accurate and rich feature representations. The combination of the two can generate a high-quality initial disparity map.

[0117] The cost aggregation small unit dynamically adjusts the aggregation path length according to the gradient strength of the sill edge, and suppresses the noise of the non-edge area of the initial disparity map; the gradient strength of the sill edge reflects the degree of feature change of the sill area in the image, by dynamically adjusting the aggregation path length according to the gradient strength of the sill edge, the cost aggregation can be more suitable for the characteristics of the sill area, further improve the accuracy of the disparity calculation of the sill edge and other key parts, at the same time, the noise of the non-edge area can reduce the interference of irrelevant information on the disparity map, make the disparity map more clear and accurate, more truly reflect the space relationship between the car and the sill;

[0118] The stereo optimization unit generates an initial disparity map and effectively optimizes the quality of the disparity map, providing more reliable data for calculating the leveling accuracy of the car sill and the landing sill based on the principle of binocular stereo vision three-dimensional positioning. The accurate disparity map can ensure the accuracy of three-dimensional positioning, so that the calculated leveling accuracy is closer to the actual situation, improving the accuracy and reliability of elevator leveling accuracy detection.

[0119] In some embodiments, the disparity optimization unit is used to optimize the disparity map through a post-processing algorithm, specifically, a bilateral filter is used to apply a spatial domain and disparity domain joint weight filter on the disparity map to retain edge sharpness, a median filter is used to take the median value of the disparity values in a 3x3 window to eliminate isolated noise points, a left-right consistency check is used to eliminate mismatched points with a disparity difference of >1 pixel, and finally an optimized disparity map with noise reduction of >90% and effective disparity coverage of >95% is output.

[0120] An elevator leveling accuracy detection method based on binocular vision, the method is based on any one of the above elevator leveling accuracy detection devices based on binocular vision, specifically comprising:

[0121] The robot body receives the elevator leveling accuracy detection task;

[0122] The perception module collects environmental data and robot body state data in real time, and determines whether the detection conditions are met based on the collected environmental data and robot body state data. If the detection conditions are met, the binocular vision detection module is used to detect the accuracy of the elevator leveling, otherwise, the elevator leveling accuracy detection operation is not performed.

[0123] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executes the computer program to implement the steps of the above elevator leveling accuracy detection method based on binocular vision.

[0124] A computer readable storage medium, the storage medium stores a computer program, the computer program is executed by a processor to implement the steps of the above elevator leveling accuracy detection method based on binocular vision.

[0125] The above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An elevator leveling accuracy detection device based on binocular vision, characterized in that: include: The robot body is used to receive elevator leveling accuracy detection tasks; The perception module is used to collect environmental data and the robot's own status data in real time, and judge whether the detection conditions are met based on the environmental data and the robot's own status data; Binocular vision detection module, used to detect the accuracy of elevator leveling.

2. The binocular vision-based elevator leveling accuracy detection device according to claim 1, characterized in that: The perception module includes: Environmental sensors, used to collect environmental data; Body sensor, used to collect the robot's own status data; Positioning unit, used to achieve global positioning using GPS or visual SLAM; A construction unit, used to construct an environment map based on environmental data, the robot's own state data and global positioning information; The judgment unit is used to judge whether the detection conditions are met according to the environment map.

3. The binocular vision-based elevator leveling accuracy detection device according to claim 2, characterized in that: The binocular vision detection module includes: The positioning floor unit is used to confirm the elevator's location on the floor by collecting the actual elevator floor information; The camera is aimed at the sill unit, which is used to adjust the camera angle in real time through the visual algorithm to ensure that the sill is centered in the image; An image acquisition unit, used for acquiring images of the elevator shaft; An image preprocessing unit, configured to preprocess the elevator shaft image to obtain a preprocessed elevator shaft image; A feature extraction unit is used to locate key points of the elevator car and the sill edge based on the pre-processed elevator shaft image; A stereo optimization unit, used to generate a disparity map based on key points between the car and the sill edge; A disparity optimization unit, used to optimize the disparity map through a post-processing algorithm; The output leveling accuracy unit is used to process the optimized disparity map based on the binocular stereo vision 3D positioning principle and calculate the leveling accuracy of the car door sill and the landing door sill.

4. The binocular vision-based elevator leveling accuracy detection device according to claim 3, characterized in that: The camera is aimed at the sill unit and includes: The sill line detection unit is used to detect sill lines in real time based on the YOLOv5 model; The PID closed-loop control unit is used to adjust the camera angle based on the detected ground sill line to center the ground sill line.

5. The binocular vision-based elevator leveling accuracy detection device according to claim 4, characterized in that: The image preprocessing unit includes: A denoising processing unit is used to denoise the elevator shaft image using a median filter and an inter-frame difference noise reduction method; Distortion correction unit, used to eliminate elevator shaft image distortion.

6. The binocular vision-based elevator leveling accuracy detection device according to claim 5, characterized in that: The feature extraction unit includes: The edge detection unit is used to locate the continuous edges of the car sill based on the Canny operator detection method and the Sobel operator detection method; The geometric feature extraction unit is used to identify the straight line segments of the sill using the Hough transform method and to extract the color areas of the car and floor door sills using the color segmentation method.

7. The binocular vision-based elevator leveling accuracy detection device according to claim 6, characterized in that: The three-dimensional optimization unit includes: A cost calculation unit is used to calculate the Census transformation cost between the key points of the car and the edge of the sill using the SGBM algorithm, and to generate an initial disparity map using a pre-trained PSMNet network; The cost aggregation unit is used to dynamically adjust the aggregation path length according to the gradient strength of the edge of the sill, suppress the noise in the non-edge area of ​​the initial disparity map, and obtain the final disparity map.

8. A binocular vision-based elevator leveling accuracy detection method, characterized in that: The method is based on the binocular vision-based elevator leveling accuracy detection device according to any one of claims 1 to 7, and specifically comprises: Use the robot body to receive the elevator leveling accuracy detection task; The perception module collects environmental data and the robot's own status data in real time, and judges whether the detection conditions are met based on the collected environmental data and the robot's own status data. If so, the binocular vision detection module is used to detect the accuracy of the elevator leveling. Otherwise, the elevator leveling accuracy detection operation is not performed.

9. A computing device, characterized in that It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the elevator leveling accuracy detection method based on binocular vision according to claim 8 are implemented.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the steps of the binocular vision-based elevator leveling accuracy detection method according to claim 8.

Images