Visual navigation method of binocular backpack AGV (Automatic Guided Vehicle)

Abstract

The invention discloses a visual navigation method of a binocular backpack AGV (Automatic Guided Vehicle), which comprises the following steps: a domain controller detects and matches X angular points in real time according to image information, and calculates image pixel coordinates and image physical coordinates of all current X angular points; the domain controller calculates the real-time length from the current AGV to the visual label and the real-time angle of the current AGV relative to the visual label according to the camera coordinate and the label coordinate of the current X corner point; the domain controller sends a corresponding driving signal to the driving motor in real time according to the calculation result of the length and the angle; the driving motor adjusts the position and the vehicle body angle of the AGV in real time according to the driving signal; the navigation effect of the AGV is remarkably improved by arranging the binocular stereo camera to shoot the visual label in real time and combining the domain controller to perform image analysis and processing in real time, so that the AGV runs more stably and safely, and the environmental suitability of the AGV is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of AGV car vision navigation, and in particular to a vision navigation method for a binocular back-mounted AGV car. Background Technique

[0002] The most prominent advantage of an AGV car is that it can work continuously for 24 hours, can reduce the labor intensity of workers, can also reduce labor costs, and more importantly, can reduce the probability of on-site accidents.

[0003] AGV cars have various navigation methods, such as laser navigation, magnetic stripe navigation, two-dimensional code navigation, vision navigation, etc. Among them, vision navigation is becoming an emerging navigation method because it is suitable for scenarios with relatively simple paths and convenient maintenance. Vision navigation is divided into monocular vision-based and binocular vision-based. Monocular vision has low accuracy and low reliability, so binocular stereo vision is the mainstream hotspot in the field of machine vision.

[0004] At present, AGV cars using binocular stereo vision have problems with poor environmental adaptability, especially low accuracy of image feature matching, low image parsing efficiency, and no body pose correction function. Summary of the Invention

[0005] In view of this, the present invention provides a vision navigation method for a binocular back-mounted AGV car, which can effectively improve the environmental adaptability of the AGV car and can correct the body pose of the AGV car.

[0006] To achieve the above object, the present invention provides the following technical solutions: A visual navigation method for a binocular back-mounted AGV vehicle, comprising: S1, designing and arranging a plurality of checkerboard-style visual tags at key positions at the implementation site; S2, using a binocular stereo camera to obtain image information of the visual tags in real time and transmitting the image information to the domain controller in real time; S3, the domain controller establishing an image pixel coordinate system and an image physical coordinate system according to the image information in real time; S4, the domain controller detecting and matching X corner points in real time according to the image information, and calculating the image pixel coordinates and image physical coordinates of all current X corner points; S5, the domain controller establishing a tag coordinate system in real time according to the X corner points with image pixel coordinates and image physical coordinates, and checking the angles and vector lengths of the coordinate axes of the tag coordinate system to realize the recognition of the visual tags. At the same time, calculate the position coordinates of the current X corner points in the tag coordinate system; S6, the domain controller establishing a camera coordinate system in real time according to the X corner points with tag coordinates and calculating the position coordinates of the current X corner points in the camera coordinate system; S7, the domain controller calculating the real-time length of the current AGV vehicle from the visual tag and the real-time angle of the current AGV vehicle relative to the visual tag according to the camera coordinates and tag coordinates of the current X corner points; S8, the domain controller sending corresponding drive signals to the drive motor in real time according to the calculation results of the length and angle; S9, the drive motor adjusting the position and body angle of the AGV vehicle in real time according to the drive signals; during the driving process of the AGV vehicle, whenever the AGV vehicle encounters a new visual tag, S2-S9 will be repeatedly executed to realize the visual navigation of the AGV vehicle.

[0007] Preferably, the establishment of the image pixel coordinate system includes: taking pixels as the basic unit, setting the upper left corner of the image as the coordinate origin, the horizontal direction as the U axis, and the vertical direction as the V axis to establish a coordinate system; the establishment of the image physical coordinate system includes: taking millimeters as the unit, setting the midpoint of the image as the coordinate origin, the horizontal direction as coordinates, and the vertical direction as coordinates to establish a coordinate system; the establishment of the camera coordinate system includes: taking millimeters as the unit, setting the optical center of the left camera as the coordinate origin, and the coordinate system axis coincides with the optical axis to establish a coordinate system; the establishment of the tag coordinate system includes: taking millimeters as the unit, setting a certain X corner point as the coordinate origin of the coordinate system to establish a coordinate system.

[0008] Preferably, the X corner point detection includes: quickly screening the area and symmetry of the X corner points by using a candidate X corner point quick screening template map to obtain multiple candidate X corner points; performing a segment test on the candidate X corner points to calculate the image pixel coordinates of the candidate X corner points; performing sub-pixel localization of the X corner points on the candidate X corner points to calculate the floating-point coordinates of the X corner points in the image pixel coordinate system, that is, the sub-pixel level image pixel coordinates.

[0009] Preferably, the area screening condition for the candidate X corner points is that the area of the tile forming the X corner point is less than half of the area of the camera window.

[0010] Preferably, the content of the segment test for the candidate X corner points includes: setting a candidate X corner point pixel as the test center, the test segment is composed of n pixels, where the i-th pixel is , the test segment is represented by the vector S: ; the pixel gradient between adjacent pixels inside the test segment is: ; traverse all elements within the pixel gradient to find the maximum value and the minimum value corresponding pixels and ; if the currently detected pixel is an X corner point, then in the test segment and the corresponding pixel located on the same diameter have a local maximum gradient value , similarly, find corresponding to among the adjacent test pixels centered on and , where and are greater than 0, while and are less than 0.

[0011] The content of the sub-pixel localization of the X corner points includes: connecting the pixels corresponding to the maximum value and the maximum value and , connecting the pixels corresponding to the minimum value and the minimum value and , the intersection of the two connecting lines is the position of the X corner point, and the sub-pixel level image pixel coordinates of the X corner point are calculated.

[0012] Preferably, the X-corner matching includes: the domain controller compares the X-corner coordinates at the same position on the same visual tag image captured by the left and right cameras of the binocular stereo camera, and filters out the X-corners that do not meet the left-right constraint and the epipolar line pair matching constraint; Left-right constraint: in the image pixel coordinate system, the U coordinate value of the X-corner on the tag image captured by the left camera is greater than the U coordinate value of the X-corner on the tag image captured by the right camera; Epipolar line pair matching constraint: in the image pixel coordinate system, the difference range of the V coordinate value of the X-corner on the tag image captured by the left camera and the V coordinate value of the X-corner on the tag image captured by the right camera is ±2 pixels.

[0013] Preferably, the verification of the angle and vector length of the coordinate axes of the tag coordinate system specifically includes: constructing two three-dimensional vectors according to the tag coordinates of the X-corner , , and being able to calculate the lengths of the two coordinate vectors and , as well as the angle between them: ; ; ; Set the included angle range to be between 85° and 95°, meeting the geometric feature that the interior angle of the rectangle is 90°. At the same time, if the vector lengths are equal and conform to the tag length, it is considered a square tag. If it is recognized that the lengths are one long and one short and conform to the tag length, it is considered a rectangular tag.

[0014] Preferably, the pose calculation of the AGV vehicle specifically includes: Calculation of the distance between the current AGV vehicle and the visual tag: The calculation formula for the vector length is: ; The modulus of the vector length is the size of the current visual tag. Among them, the modulus of the Z vector is the distance from the AGV vehicle to the current visual tag. When the modulus of the Z vector is less than the preset safety value, the AGV vehicle emits an alarm signal; The angle of the current AGV vehicle relative to the visual tag is the included angle between the X-axis of the camera coordinate system and the X-axis of the tag coordinate system , and the calculation formula for the included angle is:

[0015] The beneficial effects of the present invention are as follows: Compared with the prior art, a visual navigation method for a binocular backpack AGV vehicle disclosed in the present application significantly improves the navigation effect of the AGV vehicle by setting a binocular stereo camera to capture visual tags in real time and combining with a domain controller to perform image analysis and processing in real time, making the operation of the AGV vehicle more stable and safe. Among them, the domain controller completes the recognition of the visual tag by detecting and matching the X corner point information on the visual tag, and then quickly and accurately calculates the relative position of the current AGV vehicle with respect to the visual tag, and further controls the drive motor to adjust the attitude of the AGV vehicle in real time, so that the AGV vehicle can always drive smoothly on the planned route, effectively preventing the AGV vehicle from colliding, and significantly improving the environmental adaptability of the AGV vehicle.

[0016] Additional aspects and advantages of the present invention will be given in part in the following description, will become apparent in part from the following description, or will be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is the automatic control operation flowchart of the present invention; Figure 2 is the schematic flowchart of the segment test of the X corner point of the present invention; Figure 3 is the schematic flowchart of the sub-pixel positioning of the X corner point of the present invention; Figure 4 is the schematic flowchart of the recognition of the visual tag of the present invention; Figure 5 is the navigation coordinate system diagram of the present invention; Figure 6 is the schematic diagram of the rectangular visual tag of the present invention; Figure 7 is the schematic diagram of the square visual tag of the present invention; Figure 8 is the schematic diagram of the X corner point on the visual tag of the present invention; Figure 9 is the candidate X corner point quick screening template diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0019] In addition, the terms "first" and "second" are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined.

[0020] The following refers to Figures 1 to 9 the visual navigation method of the binocular back-mounted AGV vehicle in the embodiments of the present invention.

[0021] The embodiments of the present application disclose a visual navigation method for a binocular back-mounted AGV vehicle, including: S1. Design and arrange a plurality of checkerboard-style visual tags at key positions at the implementation site; S2. Use a binocular stereo camera to obtain image information of the visual tags in real time and transmit the image information to the domain controller in real time; S3. The domain controller establishes an image pixel coordinate system and an image physical coordinate system according to the image information in real time; S4. The domain controller detects and matches X corner points according to the image information in real time, and calculates the image pixel coordinates and image physical coordinates of all current X corner points; S5. The domain controller establishes a tag coordinate system according to the X corner points with image pixel coordinates and image physical coordinates in real time, and checks the angles and vector lengths of the coordinate axes of the tag coordinate system to realize the recognition of the visual tags. At the same time, the position coordinates of the current X corner points in the tag coordinate system are calculated; S6. The domain controller establishes a camera coordinate system according to the X corner points with tag coordinates in real time and calculates the position coordinates of the current X corner points in the camera coordinate system; S7. The domain controller calculates the real-time length of the current AGV vehicle from the visual tag and the real-time angle of the current AGV vehicle relative to the visual tag according to the camera coordinates and tag coordinates of the current X corner points; S8. The domain controller sends corresponding drive signals to the drive motor in real time according to the calculation results of the length and angle; S9. The drive motor adjusts the position and body angle of the AGV vehicle in real time according to the drive signal.

[0022] During the driving process of the AGV vehicle, whenever the AGV vehicle encounters a new visual tag, S2-S9 will be repeatedly executed, thereby realizing the visual navigation of the AGV vehicle.

[0023] By setting up a binocular stereo camera to capture visual tags in real time and combining with a domain controller to perform real-time image analysis and processing, the navigation effect of the AGV vehicle is significantly improved, making the operation of the AGV vehicle smoother and safer. Among them, the domain controller completes the recognition of the visual tag by detecting and matching the X corner point information on the visual tag, and then quickly and accurately calculates the relative position of the current AGV vehicle with respect to the visual tag, and further controls the drive motor to adjust the attitude of the AGV vehicle in real time, so that the AGV vehicle can always drive smoothly on the planned route, effectively preventing the AGV vehicle from colliding, and significantly improving the environmental adaptability of the AGV vehicle.

[0024] In some embodiments, the establishment of the image pixel coordinate system includes: taking pixels as the basic unit, setting the upper left corner of the image as the coordinate origin, the horizontal direction as the U axis, and the vertical direction as the V axis to establish a coordinate system; the establishment of the image physical coordinate system includes: taking millimeters as the unit, setting the midpoint of the image as the coordinate origin, the horizontal direction as coordinates, and the vertical direction as coordinates to establish a coordinate system; the establishment of the camera coordinate system includes: taking millimeters as the unit, setting the optical center of the left camera as the coordinate origin, the axis of the coordinate system coincides with the optical axis to establish a coordinate system; the establishment of the tag coordinate system includes: taking millimeters as the unit, setting a certain X corner point as the coordinate origin of the coordinate system to establish a coordinate system.

[0025] In this embodiment, the image pixel coordinate system, the image physical coordinate system, the camera coordinate system and the tag coordinate system together constitute a navigation coordinate system. The function of the navigation coordinate system is: enabling the domain controller to accurately identify the position coordinates of the X corner points on the visual tag according to the navigation coordinate system, so as to calculate the distance and angle of the current AGV vehicle body relative to the visual tag according to the conversion relationship between the coordinate systems, accurately controlling the body attitude of the current AGV vehicle, and then adjusting the body attitude of the AGV vehicle in real time through the drive motor to ensure that the AGV vehicle drives smoothly on the planned route, effectively preventing the AGV vehicle from deviating during driving, and significantly improving the environmental adaptability of the AGV vehicle.

[0026] In some embodiments, the X corner point detection includes: using a candidate X corner point quick screening template image to quickly screen the area and symmetry of the X corner point to obtain multiple candidate X corner points; performing a segment test on the candidate X corner points to calculate the image pixel coordinates of the candidate X corner points; performing sub-pixel positioning of the X corner points on the candidate X corner points to calculate the floating-point coordinates of the X corner points in the image pixel coordinate system, that is, the sub-pixel level image pixel coordinates.

[0027] In this embodiment, the image pixel coordinates of the X corner point can be accurately calculated by performing the X corner point test. Among them, screening the candidate X corner points can quickly eliminate the non-X corner points, realizing the preliminary positioning of the X corner point. The screening basis is as follows: Any pixel point P has the following three distribution situations in the X corner point diagram: 1. The pixel point P is located on the edge of the tile, in the bright tile, or in the dark tile; 2. There is a pixel symmetric to the pixel point P about the X corner point, and this pixel and the pixel point P are both in the bright tile or the dark tile; 3. The area of any tile forming the X corner point is less than half of the area of the camera window.

[0028] Further, after the candidate X corner points are preliminarily screened for the X corner points, the candidate X corner points are further screened and positioned through the segment test, and the image pixel coordinates of the candidate X corner points passing the segment test are calculated.

[0029] Furthermore, sub-pixel positioning is performed on the candidate X corner points passing the segment test to finally determine the accurate position of the X corner point and calculate the floating-point coordinates of the X corner point, that is, the image pixel coordinates at the sub-pixel level.

[0030] In this embodiment, the segment test content of the candidate X corner points includes: Set a candidate X corner point pixel as the test center, and the test segment consists of n pixels, where the i-th pixel is , and the test segment is represented by the vector S: ; The gradient of adjacent pixels inside the test segment is: ; Traverse all elements within the pixel gradient to find the maximum value and the minimum value corresponding pixels and ; If the currently detected pixel is an X corner point, then in the test segment and the corresponding pixel on the same diameter have the local maximum gradient value . Similarly, find corresponding in the adjacent test pixels centered on as well as , where and are greater than 0, while and are less than 0.

[0031] Further, the content of the sub-pixel positioning of the X corner point includes: the maximum value and the maximum value corresponding pixels and Connect the lines to the minimum value and the minimum value to the corresponding pixels and Connect the lines. The intersection point of the two line segments is the position of the X corner point, and the sub-pixel image pixel coordinates of the X corner point are calculated.

[0032] In some embodiments, the X corner point matching includes: the domain controller compares the X corner point coordinates at the same position on the same visual tag image captured by the left and right cameras of the binocular stereo camera, and filters out the X corner points that do not meet the left-right constraint and the epipolar line pair matching constraint; Left-right constraint: In the image pixel coordinate system, the U coordinate value of the X corner point on the tag image captured by the left camera is greater than the U coordinate value of the X corner point on the tag image captured by the right camera; Epipolar line pair matching constraint: In the image pixel coordinate system, the difference range between the V coordinate value of the X corner point on the tag image captured by the left camera and the V coordinate value of the X corner point on the tag image captured by the right camera is ±2 pixels.

[0033] In this embodiment, by performing X corner point matching, the X corner points with specific coordinates that have been calculated can be further screened, and the X corner points that do not meet the left-right constraint and the epipolar line pair matching constraint are filtered out. The role of this step is: to compare the X corner point coordinates at the same position on the same visual tag image captured by the left and right cameras of the binocular stereo camera, so as to screen out the X corner points that can correspond to each other on both pictures.

[0034] In some embodiments, the verification of the angle and vector length of the coordinate axes of the tag coordinate system specifically includes: According to the tag coordinates of the X corner point, construct two three-dimensional vectors , , and the lengths of the two coordinate vectors can be calculated and , as well as the angle between them: ; ; ; Set the included angle range to be between 85° and 95°, which satisfies the geometric feature that the interior angle of a rectangle is 90°. At the same time, if the vector lengths are equal and meet the tag length, it is considered a square tag. If it is recognized that one length is long and the other is short, and it meets the tag length, it is considered a rectangular tag. By verifying the angle of the coordinate axis and the vector length, the recognition of the visual tag is completed.

[0035] In some embodiments, the pose calculation of the AGV vehicle specifically includes: Calculation of the current distance of the AGV vehicle from the visual tag: Vector length The calculation formula is as follows: ; Vector length The modulus of the vector is the size of the current visual tag. Among them, the modulus of the Z vector is the length from the AGV vehicle to the current visual tag. When the modulus of the Z vector is less than the preset safety value, the AGV vehicle sends an alarm signal; The angle of the current AGV vehicle relative to the visual tag is the included angle between the X-axis of the camera coordinate system and the X-axis of the tag coordinate system , included angle The calculation formula is as follows: .

[0036] In this embodiment, by calculating the length of the current AGV vehicle from the visual tag and the angle of the current AGV vehicle relative to the visual tag, the body pose of the current AGV vehicle can be determined, that is, the position of the body of the current AGV vehicle relative to the visual tag. In this way, the domain controller can send corresponding control signals to the drive motor according to the calculated data to control the drive motor to adjust the body pose of the AGV vehicle in real time, so as to ensure that the AGV vehicle can drive smoothly on the planned route.

[0037] The following uses a specific embodiment to illustrate the visual navigation method of the present application: Visual tags are arranged at key positions at the implementation site, such as left turns, right turns, straight-ahead sections, etc. For example, when the trolley travels straight for a certain distance and reaches a turning point, during the turning process of the AGV trolley, the binocular stereo camera will continuously capture the visual tags and transmit the image signals to the domain controller. When the visual tags are recognized, the AGV trolley will continuously capture the tags. At the same time, the domain controller calculates the distance and angle between the AGV trolley body and the visual tags and sends the drive signal to the integrated drive motor. The motor adjusts the position and body angle of the trolley. When it is confirmed that the distance and angle are within the safe range, the AGV trolley will go straight again. The binocular stereo camera will re-recognize the instructions on the tag and make actions again according to the instructions, and so on until the carried goods are delivered to the target point. During this period, if the RGB light strip operates normally, it will flash green, and the music speaker will not make any sound. If the distance and angle between the AGV trolley and the visual tags are non-safe values, the RGB light strip will flash red, and the music speaker will emit an alarm sound, and the alarm information will be sent to the server via the wireless terminal. If the AGV trolley does not emit an alarm signal during driving and touches an obstacle, the edge switch will be triggered and the signal will be transmitted to the domain controller. The domain controller will send a signal to immediately stop the vehicle, causing the integrated drive motor to immediately pause. If you need to resume, you need the operator to troubleshoot on-site and then operate on the touch display screen to make the vehicle resume its working state. In addition, if the on-site staff finds that the vehicle is operating abnormally, they can also press the emergency stop switch to stop the vehicle.

[0038] Other components and operations of the visual navigation method of the binocular back-mounted AGV trolley according to the embodiments of the present invention are known to those of ordinary skill in the art and will not be described in detail here.

[0039] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0040] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and purposes of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. A visual navigation method for a binocular backpack AGV vehicle, characterized in that, Including: S1. Design and arrange multiple checkerboard - style visual tags at key positions at the implementation site; S2. Use a binocular stereo camera to obtain the image information of the visual tags in real - time and transmit the image information to the domain controller in real - time; S3. The domain controller establishes an image pixel coordinate system and an image physical coordinate system in real - time according to the image information; S4. The domain controller detects and matches the X - corners in real - time according to the image information, and calculates the image pixel coordinates and image physical coordinates of all current X - corners; S5. The domain controller establishes a tag coordinate system in real - time according to the X - corners with image pixel coordinates and image physical coordinates, and checks the angles and vector lengths of the coordinate axes of the tag coordinate system to realize the recognition of the visual tags. At the same time, it calculates the position coordinates of the current X - corners in the tag coordinate system; S6. The domain controller establishes a camera coordinate system in real - time according to the X - corners with tag coordinates and calculates the position coordinates of the current X - corners in the camera coordinate system; S7. The domain controller calculates the real - time distance between the current AGV cart and the visual tag and the real - time angle of the current AGV cart relative to the visual tag according to the camera coordinates and tag coordinates of the current X - corners; S8. The domain controller sends corresponding drive signals to the drive motor in real - time according to the calculation results of the length and angle; S9. The drive motor adjusts the position and body angle of the AGV cart in real - time according to the drive signals; During the driving process of the AGV cart, whenever the AGV cart encounters a new visual tag, it will repeat the operations of S2 - S9, thereby realizing the visual navigation of the AGV cart.

2. The visual navigation method of the binocular back - mounted AGV cart according to claim 1, characterized in that The establishment of the image pixel coordinate system includes: taking pixels as the basic unit, setting the upper left corner of the image as the coordinate origin, the horizontal direction as the U axis, and the vertical direction as the V axis to establish the coordinate system; The establishment of the physical coordinate system of the image includes: with the unit of millimeters, setting the midpoint of the image as the coordinate origin, the horizontal direction as coordinates, and the vertical direction as coordinates, to establish the coordinate system; The establishment of the camera coordinate system includes: setting the optical center of the left camera with the unit of millimeters as the coordinate origin, and the coordinate system axis coincides with the optical axis to establish the coordinate system; The establishment of the label coordinate system includes: with the unit of millimeters, setting a certain X corner point as the origin of the coordinate system, and establishing the coordinate system.

3. The visual navigation method of the binocular back-mounted AGV vehicle according to claim 2, characterized in that, The X - corner detection includes: Use a candidate X - corner fast - screening template image to quickly screen the area and symmetry of the X - corners to obtain multiple candidate X - corners; Conduct a segment test on the candidate X - corners to calculate the image pixel coordinates of the candidate X - corners; Conduct sub - pixel localization of the X - corners on the candidate X - corners to calculate the floating - point coordinates of the X - corners in the image pixel coordinate system, that is, the sub - pixel - level image pixel coordinates.

4. The visual navigation method of the binocular back - mounted AGV cart according to claim 3, characterized in that The area screening condition for candidate X - corners is: the area of the tile forming the X - corner is less than half of the area of the camera window.

5. The visual navigation method of the binocular back-mounted AGV cart according to claim 3, characterized in that The content of the segment test for candidate X - corners includes: Set a candidate X corner pixel as the test center, the test segment consists of n pixels, where the i-th pixel is , the test segment is represented by the vector S: ; The adjacent pixel gradients inside the test segment are as follows: ; Traverse all elements within the pixel gradient to find the maximum value among them and the minimum value corresponding pixels and ; If the currently detected pixel is an X corner point, then in the test segment and the corresponding pixels located on the same diameter have a local maximum gradient value Similarly, among the adjacent test pixels centered on find the corresponding as well as where and are greater than 0, while and are less than 0.

6. The visual navigation method of the binocular back-mounted AGV vehicle according to claim 3, characterized in that The content of the sub - pixel localization of the X - corners includes: The maximum value and the maximum value corresponding pixels and are connected by a line. Connect the minimum value and the minimum value corresponding pixels and by a line. The intersection point of the two lines is the position of the X corner point, and the sub-pixel image pixel coordinates of the X corner point are calculated.

7. The visual navigation method of the binocular back-mounted AGV vehicle according to claim 3, characterized in that, The X - corner matching includes: The domain controller compares the X - corner coordinates at the same position on the images of the same visual tag taken by the left and right cameras of the binocular stereo camera, and screens out the X - corners that do not meet the left - right constraint and the epipolar line pair matching constraint; Left - right constraint: In the image pixel coordinate system, the U - coordinate value of the X - corner on the tag image taken by the left camera is greater than the U - coordinate value of the X - corner on the tag image taken by the right camera; Epipolar line pair matching constraint: In the image pixel coordinate system, the difference range of the V - coordinate values of the X - corners on the tag images taken by the left and right cameras is ±2 pixels.

8. The visual navigation method of the binocular back-mounted AGV vehicle according to claim 1, wherein The verification of the angles and vector lengths of the coordinate axes of the label coordinate system specifically includes: Construct two three-dimensional vectors according to the labeled coordinates of the X corner points , , and be able to calculate the lengths of the two coordinate vectors and , as well as the angle between them :[[]]END]] ； ； ； Set the included angle range to be between 85° and 95°, which satisfies the geometric feature that the interior angle of a rectangle is 90°. At the same time, if the vector lengths are equal and conform to the label length, it is considered a square label. If it is recognized that the lengths are one long and one short and conform to the label length, it is considered a rectangular label.

9. The visual navigation method of the binocular backpack AGV vehicle according to claim 1, characterized in that, The pose calculation of the AGV vehicle specifically includes: Calculation of the length of the current AGV vehicle from the vision label: Vector length The calculation formula is as follows: ; Vector length The modulus of the vector is the size of the current visual tag. Among them, the modulus of the Z vector is the distance from the AGV vehicle to the current visual tag. When the modulus of the Z vector is less than the preset safety value, the AGV vehicle emits an alarm signal; The angle of the current AGV vehicle relative to the vision tag is the included angle between the X-axis of the camera coordinate system and the X-axis of the tag coordinate system , the included angle The calculation formula is as follows: .

Images