A vision-based method for detecting axial and circumferential deviations of UUVs

Abstract

The present invention provides a vision-based UUV axial and circumferential deviation detection method, which performs preprocessing after image acquisition to obtain target concentric circle information. If the number of detected circle centers is 0, it means that the visual system has detected that there is no target in the field of view, and then sends relevant information to control the charging platform to make axial movement to find the target. If at least one concentric circle center is successfully identified, the circumferential alignment process is entered. First, the deflection direction is determined, and then the angle θ between the current position and the alignment state is calculated according to the corresponding formula, and the rotation direction and rotation angle are sent to the charging platform. If the angle falls within the allowable error range, the circumferential alignment process is completed and the axial alignment process is entered. During axial alignment, the distance between the current position and the alignment position is calculated, and the translation direction and distance are sent to the charging platform. Finally, if the distance falls within the allowable error range, the entire alignment process is completed.

Description

Technical Field

[0001] The present invention relates to a UUV charging alignment method, and in particular to a vision-based UUV axial and circumferential deviation detection method. Background Art

[0002] Current charging alignment methods [A Vision-Based Robot Automatic Charging Alignment System and Method 201810359873.X] [A Vision-Based Automatic Charging Alignment Method 202011014009.X] use the position and size of a target pattern in the field of view to determine the directional difference between the current position and the alignment position, and then determine the next movement direction. However, these methods only provide the direction of the next movement, without specifying the specific distance or rotation angle, making alignment inefficient. Furthermore, if the target is partially obscured, alignment cannot be continued. Summary of the Invention

[0003] The purpose of the present invention is to solve the problem of autonomous charging alignment of UUVs and to provide a vision-based method for axial and circumferential alignment of structures. The method can first return the rotation direction and rotation angle required for circumferential alignment, and after the circumferential alignment is completed, return the translation direction and translation distance required for axial alignment. It can continue to be used even when the target is partially blocked, and has a certain degree of robustness. The UUV is fixed when charging and is moved to a specified distance from the camera. During charging, the charging platform moves to align with the charging port on the UUV. During the process, the camera is mounted on the charging platform and moves with the charging platform.

[0004] The purpose of the present invention is achieved in this way: it includes three processes: an image preprocessing process, a circumferential alignment process, and an axial alignment process.

[0005] After image acquisition, preprocessing is performed to obtain the target concentric circle information. If the number of detected circle centers is 0, it means that the visual system has detected that there is no target in the field of view, and then sends relevant information to control the charging platform to perform axial movement to find the target. If at least one concentric circle center is successfully identified, the circumferential alignment process begins. First, determine the deflection direction, and then calculate the angle θ between the current position and the alignment state according to the corresponding formula, and send the rotation direction and rotation angle to the charging platform. If the angle falls within the allowable error range, the circumferential alignment process is completed and the axial alignment process begins. During axial alignment, calculate the distance between the current position and the alignment position, and send the translation direction and distance to the charging platform. Finally, if the distance falls within the allowable error range, the entire alignment process is completed.

[0006] 1. Preprocessing

[0007] The pretreatment process flow chart is as follows Figure 8As shown in the figure, the preprocessing stage processes image information from the bottom layer, ultimately returning the coordinates of the target feature points, the length of the major axis of the concentric ellipses, and the occlusion status of each feature point. The occlusion mode consists of four bits, each representing whether one of the four concentric circles is recognized. If recognized, the bit is set to "1"; otherwise, it is set to "0". This code reflects the current occlusion mode and serves as the basis for determining the available chord endpoints in the next stage.

[0008] 2. Circumferential alignment

[0009] The flow chart of the circumferential alignment procedure is as follows: Figure 9 As shown in the figure, the chord endpoints involved in the calculation are determined through the occlusion pattern. The rotation direction is then determined based on the relationship between π-β-α and γ. The rotation angle is then calculated using the relevant formula and sent to the host computer. After stopping the rotation, θ is calculated again. If θ falls within the allowable error range, the axial alignment phase begins. Otherwise, rotation commands are sent to adjust the camera position.

[0010] 3. Axial alignment

[0011] After circumferential alignment, the target surface can be approximately considered perpendicular to the optical axis. In this case, the actual size represented by a pixel can be determined using the known actual size of the target (the actual size of the target is known). The required axial movement distance can be determined by calculating the difference between the coordinates of the target center point and the image center in the current image coordinate system. This difference is then converted from the pixel coordinate system to the camera coordinate system using the camera calibration parameters.

[0012] Compared with existing technologies, the present invention offers the following advantages: it can directly return clear movement instructions, including the direction and distance of movement, improving alignment efficiency. Furthermore, alignment can be completed by simply identifying a single concentric circle. It can also be used in the presence of partial occlusion, demonstrating a certain degree of robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 (a)-(b) is a schematic diagram of the working principle of the alignment system using the left end point of the target string for circumferential alignment. Figure 1a Schematic diagram of the camera's left-biased state, 1 is the circumferential alignment state, and 2 is the left-biased state; Figure 1b The diagram below shows the camera's right-biased state. 1 is the circumferential alignment state, and 2 is the right-biased state. The letters are defined as follows: A represents the rotation center of the charging platform; B represents the right endpoint of the target chord; C represents the left endpoint of the target chord; D represents the midpoint of the target chord; O1 represents the camera's optical center in the circumferential alignment state; O2 represents the camera's optical center in the circumferential offset state; C represents the camera's optical center in the circumferential alignment state; and D represents the camera's optical center in the circumferential offset state. o is the coordinate of the optical center in the pixel coordinate system; x cis the coordinate of the endpoint C of the target chord in the pixel coordinate system; K represents the distance from the camera rotation center to one endpoint of the target chord, in millimeters (mm); L is the target chord length, in millimeters (mm); R represents the rotation radius, in millimeters (mm); f represents the focal length of the camera, that is, the distance from the optical center of the camera to the imaging plane, in pixels (pixel); θ is the angle to be determined, representing the angle between the circumferential alignment state and the current misalignment state, in degrees (°); the remaining angles are intermediate variables, in degrees (°);

[0014] Figure 2 Schematic diagram of the range beyond circumferential rotation, where 1 is the circumferential alignment state and 2 is the over-deflection state;

[0015] Figure 3 (a)-(b) is a schematic diagram of the working principle of the alignment system using the right end point of the target chord for circumferential alignment, where Figure 3a Schematic diagram of the camera's left-biased state, 1 is the circumferential alignment state, and 2 is the left-biased state; Figure 3b Schematic diagram of the camera's right-biased state, 1 represents the circumferential alignment state, and 2 represents the right-biased state. The definitions of the letters are the same as above.

[0016] Figure 4 The imaging principle diagram of a monocular camera is shown in Figure 1, where 1 is the optical center of the camera, 2 is the camera imaging plane, and 3 is the actual object. a and b are the object sizes in the camera imaging plane (pixel). A and B are the actual object sizes (mm).

[0017] Figure 5 Schematic diagram of the conversion from camera coordinate system to pixel coordinate system.

[0018] Figure 6 To align the overall process flow chart.

[0019] Figure 7 Target style.

[0020] Figure 8 This is the flow chart of the preprocessing process.

[0021] Figure 9 Flowchart of the circumferential alignment procedure. DETAILED DESCRIPTION

[0022] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] The whole process of the present invention first pre-processes the image, extracts the target key point information, then performs circumferential alignment, and then performs axial alignment.

[0024] The pretreatment process flow chart is as follows Figure 8As shown in the figure, the preprocessing stage processes the image information from the bottom layer, ultimately returning the coordinates of the target feature points, the length of the major axis of the concentric ellipses, and the occlusion status of each feature point. The occlusion mode consists of four bits, each representing whether one of the four concentric circles is recognized. If recognized, the bit is set to "1", otherwise it is set to "0".

[0025] The flow chart of the circumferential alignment procedure is as follows: Figure 9 As shown in the figure, the chord endpoints involved in the calculation are determined through the occlusion mode, the rotation direction is determined, the rotation angle is calculated, and then sent to the host computer. After stopping the rotation, the circumferential alignment is checked again. If it is aligned, the axial alignment phase is entered. Otherwise, the rotation command is continued to adjust the camera position.

[0026] Axial alignment: Calculate the difference between the coordinates of the target center point in the current image coordinate system and the image center. Then, use the camera calibration parameters to convert it from the pixel coordinate system to the camera coordinate system to determine the required axial movement distance.

[0027] The specific process is as follows:

[0028] 1. Circumferential alignment

[0029] Circumferential alignment calculates the angle between the current state and the circumferential alignment state based on the geometric relationship between the UUV, charging platform, and camera. Since the target is in the form of four concentric circles with different radius ratios, the line connecting the centers of two horizontal concentric circles is called the target chord, and the center of each concentric circle is called the target chord endpoint. The result can be obtained using only one target chord endpoint.

[0030] (1) Circumferential alignment using the left end point of the target chord

[0031] The schematic diagram of the working principle of circumferential alignment using the target chord endpoint C is shown in Figure 1(a) and (b). The figure describes the state of the camera being deflected to the right and left relative to the circumferential alignment position. The range of circumferential rotation is ±15° relative to the circumferential alignment position, ignoring the case where θ>γ, such as Figure 2 shown.

[0032] Where K represents the distance from the camera's rotation center to one end point of the target's chord, in millimeters (mm); R represents the rotation radius, in millimeters (mm); f represents the camera's focal length, that is, the distance from the camera's optical center to the imaging plane, in pixels (pixels); c o x is the coordinate of the optical center in the pixel coordinate system, the unit is pixel; c is the coordinate of the endpoint C of the chord of the target in the pixel coordinate system, in pixels; θ is the angle to be determined, representing the angle between the circumferentially aligned state and the current misaligned state, in degrees; the remaining angles are intermediate variables, in degrees.

[0033] Regardless of whether the camera is tilted to the right or left, as shown in Figure 1, the following applies:

[0034] In the right triangle O2x c In c:

[0035]

[0036] In triangle ABO2:

[0037]

[0038] In right triangle ABD:

[0039]

[0040] When the camera is in the right-biased state, as shown in Figure 1(a), in triangle ACO2:

[0041] π-β-α=θ+γ>γ

[0042] therefore:

[0043] θ=π-β-α-γ

[0044] When the camera is in the left-biased state, as shown in Figure 1(b), in triangle ACO2:

[0045] π-β-α=γ-θ<γ

[0046] therefore:

[0047] θ=γ-(π-β-α)

[0048] In summary, the relationship between π-β-α and γ can be used to determine whether the camera is deflected to the right or left. Substituting this into the corresponding formula, we can obtain the angle θ between the circumferentially aligned state and the current misaligned state.

[0049] (2) Circumferential alignment using the right end point of the target chord

[0050] As shown in FIG3 , θ can be obtained using the coordinates of the other end point B of the target chord in the pixel coordinate system.

[0051] Regardless of whether the camera is tilted to the right or to the left, as shown in Figure 3, the following applies:

[0052] In the right triangle O2x b c o middle:

[0053]

[0054] In triangle ABO2:

[0055]

[0056] In a right triangle ADC, we have:

[0057]

[0058] When the camera is in the right-biased state, as shown in Figure 3(a), in triangle ABO2:

[0059] π-ω-ξ=γ-θ<γ

[0060] therefore:

[0061] θ=γ-(π-ξ-ω)

[0062] When the camera is in the left-biased state, as shown in Figure 3(b), in triangle ABO2:

[0063] π-ω-ξ=γ+θ>γ

[0064] therefore:

[0065] θ=π-ω-ζ-γ

[0066] In summary, the relationship between π-ω-ζ and γ can be used to determine whether the camera is deflected to the right or left. Substituting this into the corresponding formula, we can obtain the angle θ between the circumferentially aligned state and the current misaligned state.

[0067] 2. Axial alignment

[0068] After circumferential alignment, the target surface can be approximately considered perpendicular to the optical axis. In this case, the actual size represented by one pixel can be calculated using the known actual size (the actual target size is known). This allows the number of pixels between the current target center and the image center to be calculated, giving the actual distance from the zero position.

[0069] In the camera coordinate system, from similar triangles, we have:

[0070]

[0071] If a, b, and B are known, then A can be expressed as:

[0072]

[0073] During axial alignment, the transformation relationship from a point on the imaging plane to the pixel coordinate system is as follows.

[0074]

[0075] Where (u, v) is the coordinate of the point in the pixel coordinate system, and (x, y) is the coordinate of the point in the camera coordinate system.c , α y is the magnification, (u0, v0) is the coordinate of the camera optical center in the pixel coordinate system, which can be obtained by camera calibration.

[0076] It can be seen that the transformation relationship between the points on the imaging plane and the pixel coordinate system is linear, so the ratio of the two distances in the pixel coordinate system is the same as the ratio of the two points in the camera coordinate system. In the equation, a can be replaced by the difference between the current position and the zero position in the camera coordinate system, B can be replaced by a known dimension on the target (e.g., the distance between the centers of two concentric circles or the diameters of concentric circles), and b can be replaced by the corresponding dimension of B in the pixel coordinate system. This yields the actual value A of the difference between the current position and the zero position.

Claims

1. A vision-based UUV axial and circumferential deviation detection method, characterized by: It includes image preprocessing process, circumferential alignment process and axial alignment process; After image acquisition, preprocessing is performed to obtain target concentric circle information. If the number of circle centers detected is 0, it means that the visual system has detected no target in the field of view, and then relevant information is sent to control the charging platform to perform axial movement to search for the target. If at least one concentric circle center is successfully identified, the circumferential alignment process begins. First, the deflection direction is determined, then the angle θ between the current position and the alignment state is calculated, and the rotation direction and angle are sent to the charging platform. If the angle falls within the allowable error range, the circumferential alignment process is completed and the axial alignment process begins. During axial alignment, the distance between the current position and the alignment position is calculated, and the translation direction and distance are sent to the charging platform. Finally, if the distance falls within the allowable error range, the entire alignment process is completed. The preprocessing stage processes image information from the bottom layer, ultimately returning the coordinates of the target feature points, the length of the major axis of the concentric ellipses, and the occlusion status of each feature point. The occlusion mode consists of four bits, each representing whether one of the four concentric circles is recognized. If recognized, the bit is set to "1", otherwise it is set to "0". The current occlusion mode is the basis for determining the available chord endpoints in the next stage.

2. The vision-based UUV axial and circumferential deviation detection method according to claim 1, characterized in that: The endpoints of the chord involved in the calculation are determined through the occlusion pattern. Then, the rotation direction is determined based on the size relationship between π-β-α and γ. The rotation angle is calculated and sent to the host computer. After stopping the rotation, θ is calculated again. If θ falls within the allowable error range, the axial alignment phase is entered. Otherwise, the rotation command is continued to adjust the camera position.

3. The vision-based UUV axial and circumferential deviation detection method according to claim 1, characterized in that: After circumferential alignment, the target surface can be approximately regarded as perpendicular to the optical axis. In this case, the actual size represented by one pixel is calculated through the actual size; the difference between the coordinates of the target center point and the image center in the current image coordinate system is calculated, and then it is converted from the pixel coordinate system to the camera coordinate system through the camera calibration parameters, that is, the distance required for axial movement is obtained.

Images