A method of screening misidentified labels

Abstract

The application provides a method for screening misrecognized tags, comprising: obtaining a current tag of a robot and reading a global position of the current tag in a database; comparing whether a distance between the global position of the current tag of the robot and a current robot pose meets a standard; judging whether a historical correct tag exists, and if so, performing a check on the current tag of the robot; comparing whether a distance between the global position of the current tag of the robot and a position of the historical correct tag meets the standard; discarding the current tag of the robot that fails the check; using a trusted current tag of the robot to calculate a positioning pose, and updating the current tag of the robot to the historical correct tag. The application improves the reliability of robot positioning, and even if a tag is misrecognized, the robot will not use the misrecognized tag to calculate a pose, ensuring that a trusted tag is used for positioning each time, effectively ensuring the accuracy and stability of robot positioning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meal delivery robots, and in particular to a method for screening misrecognized labels. BACKGROUND

[0002] When a meal delivery robot performs autonomous indoor positioning, it needs to use the currently recognized label and the pre-established label database to perform pose calculation. Misrecognition of the label may occur due to external reasons such as light, thereby negatively affecting the accuracy of the robot positioning.

[0003] To solve the above technical problems, the prior art uses a dynamic exposure technique to prevent misrecognition of the label using an infrared camera. By using different exposure values to take pictures of the same label for recognition, the influence of light on recognition is reduced.

[0004] For example, the 202110038492.3 patent discloses an infrared image brightness evaluation, exposure control method, device and medium. The method includes: obtaining an infrared image as a to-be-evaluated image; in a rectangular region of N pixels in the center of the to-be-evaluated image, counting the number of pixel points less than or equal to a first pixel threshold and greater than or equal to a second pixel threshold; comparing the ratio of the number of pixel points in the above threshold to the total number of pixel points in the rectangular region with a certain preset proportion threshold to determine whether the rectangular region is high light or low light; calculating the average brightness value of each pixel point in the rectangular region when it is high light or low light, which is used as the light measurement value; based on the difference between the light measurement value and the target light measurement value, determining the exposure adjustment step, setting the exposure parameter

[0005] , and realizing exposure control.

[0006] However, the dynamic exposure technique ensures that the label in the picture is clear, facilitates picture processing and decoding of information, and reduces the error rate of label recognition. However, the decoded information is not checked, and there is a possibility that the picture is correct but the recognition is wrong. SUMMARY

[0007] In view of the deficiencies of the prior art, the present application provides a method for screening misrecognized labels. The present application jointly checks real-time labels with real-time pose and label library data, thereby excluding incorrect labels and avoiding the problem of affecting positioning accuracy.

[0008] The technical solution of the present application is: a method for screening misrecognized labels, comprising the following steps:

[0009] S1), obtain the current label of the robot, and determine whether the current label of the robot belongs to the label database. If it belongs to the label database, read the global position of the robot recorded in the database;

[0010] S2), compare the distance between the global position of the robot current label and the current robot pose, if it meets the requirements, it is proved that the robot current label is reliable, and jump to step S6), otherwise jump to step S3);

[0011] S3), judge whether the historical correct label exists, if the historical correct label exists, the verification of the robot current label is carried out, jump to step S4) processing; otherwise, jump to step S5);

[0012] S4), compare the distance between the global position of the robot current label and the historical correct label position, if it meets the requirements, it is proved that the robot current label is reliable, and jump to step S6), otherwise, jump to step S5);

[0013] S5), discard the robot current label which cannot pass the verification;

[0014] S6), use the reliable robot current label to calculate the positioning pose, and update the robot current label as the historical correct label.

[0015] As preferred, in step S2), if the distance between the global position of the robot current label and the current robot pose is less than 5m, it is proved that the robot current label is reliable, and jump to step S6); otherwise jump to step S3).

[0016] As preferred, in step S4), if the distance between the global position of the robot current label and the historical correct label position is less than 5m, it is proved that the robot current label is reliable, and jump to step S6); otherwise, log the label of this misrecognition, jump to step S5).

[0017] As preferred, in step S6), the reliable robot current label is used to calculate the robot positioning pose, which specifically includes the following steps:

[0018] S61), calculate the angle of the robot current label,

[0019] The graphic design of the robot current label is a sign with a starting point and an ending point, and the pixel coordinates of the starting point and the ending point in the picture are obtained And The angle is calculated by using the arctangent function :

[0020] ;

[0021] S62), calculate the angle of the world coordinate system of the camera,

[0022] The world coordinate system angle of the tag in the known database is Therefore, the angle obtained in step S1) is superimposed on , and the world coordinate system angle of the camera ;

[0023] ;

[0024] S63), the real distance between the camera and the identified tag is calculated,

[0025] The start and end coordinates of the tag are used to calculate the center point of the tag , and then the pixel distance between the center point of the tag and the center point of the picture is calculated, and the real distance is calculated using the pixel distance and :

[0026] ;

[0027] ;

[0028] In the formula, f is the focal length of the camera, h is the corresponding height of the tag, and both data are recorded in the database;

[0029] S64), the world coordinate system position of the camera

[0030] The world coordinate of the tag in the known database is , the world coordinate angle of the camera is

[0031] , the real distance between the camera and the tag is and , so the world coordinate of the camera can be calculated by coordinate system conversion:

[0032] ;

[0033] S65), the world coordinate of the robot

[0034] Since the camera and the robot center are oriented in the same direction, but the positions are not coincident, and the camera is located at the (0, s) position of the robot center coordinate system, so the world coordinate of the robot center is obtained:

[0035] ;

[0036] The beneficial effects of the present application are:

[0037] 1、The present application improves the reliability of robot positioning, even if the label appears misidentification problem, the robot will not use the misidentification label for pose calculation, ensures that each positioning uses the label that can be trusted, effectively ensures the accuracy and stability of robot positioning;

[0038] 2、The real-time pose, label database joint verification method of the present application judges whether the currently identified label conforms to the actual position, the label verified by joint verification will be regarded as a trusted label and used for positioning pose calculation, and the label not verified will be discarded and not used for pose calculation, reducing the possibility of positioning jump caused by misidentified labels. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is the flowchart of the present application;

[0040] Figure 2 is the flowchart of the robot positioning pose calculation of the present application; DETAILED DESCRIPTION

[0041] The specific embodiments of the present application will be further described below in combination with the drawings:

[0042] As Figure 1 shown, the present embodiment provides a method for screening misidentified labels, comprising the following steps:

[0043] S1), acquire the current label of the robot, judge whether the current label of the robot belongs to the label database, if it belongs to the label database, read the global position of the current label of the robot recorded in the database;

[0044] S2), compare the distance between the global position of the current label of the robot and the current robot pose; if the distance between the global position of the current label of the robot and the current robot pose is less than 5m, it proves that the current label of the robot is trusted, and jump to step S6); otherwise, jump to step S3);

[0045] S3), judge whether the historical correct label exists, if the historical correct label exists, the current label of the robot is verified, jump to step S4) for processing; otherwise, jump to step S5);

[0046] S4), compare the distance between the global position of the current label of the robot and the position of the historical correct label, if the distance between the global position of the current label of the robot and the position of the historical correct label is less than 5m, it proves that the current label of the robot is trusted, and jump to step S6); otherwise, log the misidentified label this time, jump to step S5);

[0047] S5), discard the current label of the robot that cannot pass the verification;

[0048] S6), using the trusted robot current label to calculate the pose of the robot, and updating the robot current label to the historical correct label.

[0049] As preferred in the embodiment, in step S6), the robot pose is calculated using the trusted robot current label, which specifically includes the following steps: Figure 2

[0050] S61), calculating the angle of the robot current label,

[0051] The graphic design of the robot current label is a mark with a start point and an end point, and the pixel coordinates of the start point and the end point in the picture are obtained and , and the angle is calculated using the arctangent function :

[0052] ;

[0053] S62), calculating the angle of the camera in the world coordinate system,

[0054] The angle of the label in the world coordinate system in the database is known , so the angle obtained in step S1) is superimposed on , and the angle of the camera in the world coordinate system is obtained;

[0055] ;

[0056] S63), calculating the real distance between the camera and the recognized label,

[0057] The center point of the label is calculated using the start point and end point coordinates of the label , and then the pixel distance between the center point of the label and the center point of the picture is calculated , and the real distance is calculated using the pixel distance and :

[0058] ;

[0059]

[0060] In the formula, f is the focal length of the camera, and h is the corresponding height of the label, both of which are recorded in the database;

[0061] S64), calculating the position of the camera in the world coordinate system

[0062] The world coordinates of the label in the database are known ​, the camera's world coordinate angle

[0063] , the actual distance between the camera and the tag and Therefore, the world coordinate of the camera can be calculated through coordinate system conversion:

[0064] ;

[0065] S65), the world coordinate of the computer robot

[0066] Since the camera and the robot center are oriented to coincide, but the positions do not coincide, and the camera is located at the (0, s) position of the robot center coordinate system, the world coordinate of the robot center is obtained as:

[0067]

[0068] The above embodiments and descriptions described in the specification are only to illustrate the principles and best embodiments of the present application, and various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application.

Claims

1. A method for filtering misidentified labels, characterized in that, Includes the following steps: S1) Obtain the robot's current tag, determine whether the robot's current tag belongs to the tag database, and if it belongs to the tag database, read the global position of the robot's current tag record in the database; S2) Compare the distance between the global position of the robot's current label and the current robot pose. If the requirements are met, it proves that the robot's current label is trustworthy and jumps to step S6); otherwise, jumps to step S3. S3) Determine if a historical correct label exists. If a historical correct label exists, then perform a verification on the robot's current label and proceed to step S4) for further processing. Otherwise, proceed to step S5); S4) Compare the distance between the robot's current label's global position and the historical correct label position. If the distance meets the requirements, it proves that the robot's current label is reliable, and jumps to step S6); otherwise, jumps to step S5. S5) Discard the current label of the robot that fails the verification; S6) Calculate the localization pose using the trusted robot's current label and update the robot's current label to the correct historical label.

2. The method for filtering misidentified tags according to claim 1, characterized in that: In step S2), if the distance between the global position of the robot's current tag and the current robot pose is less than 5m, it proves that the robot's current tag is trustworthy, and the process jumps to step S6); otherwise, the process jumps to step S3).

3. The method for filtering misidentified tags according to claim 1, characterized in that: In step S4), if the distance between the global position of the robot's current tag and the historical correct tag position is less than 5m, it proves that the robot's current tag is reliable, and the process jumps to step S6); otherwise, the misidentified tag is logged, and the process jumps to step S5.

4. The method for filtering misidentified tags according to claim 1, characterized in that: In step S6), the robot's localization pose is calculated using the trusted robot's current label, specifically including the following steps: S61) Calculate the angle of the robot's current label. The graphic design of the robot's current label is a marker with a start and end point. Obtain the pixel coordinates of the start and end points in the image. and Then, the angle is calculated using the arctangent function. : ； S62) Calculate the world coordinate system angle of the camera. The world coordinate system angle of the labels in the database is known to be... Therefore, the angle obtained in step S1) is superimposed onto... The camera's world coordinate system angle can be obtained from the above. ; ； S63) Calculate the true distance between the camera and the recognition tag, and use the starting and ending coordinates of the tag to calculate the tag center point. Next, calculate the center point of the label and the center point of the image. The pixel distance is used to calculate the true distance. and : ； ； In the formula, h is the camera's focal length, and h is the height corresponding to the label. Both data are recorded in the database. S64) Calculate the camera's world coordinate system position. The world coordinates of the labels in the database are known to be... The world coordinates of the camera The actual distance between the camera and the label and Therefore, the world coordinates of the camera can be calculated through coordinate system transformation: ； S65) Calculate the robot's world coordinates. Since the camera and the robot's center point are aligned, but their positions are not, and the camera is located at (0, s) in the robot's center coordinate system, the world coordinates of the robot's center are obtained as follows: 。

Images