A method for accurate identification and positioning of different models of furnace rear pistons

Abstract

The present application relates to a kind of accurate identification and positioning method suitable for different models of furnace rear piston, the method uses 2D industrial camera to collect the image information of piston on chain plate, then the contour of piston top surface circle is extracted, and then the pixel coordinates of each pixel point on the contour are corrected;Then the pixel coordinates of each point are converted to world coordinates, and the contour point column is fitted by least square circle for the first time, to obtain the center coordinates and radius of the fitted circle, by comparing the distance of each point in the contour point column to the center of the fitted circle with the difference value of the fitted radius, the noise points in the contour point column are removed;Second least square circle fitting is carried out, to obtain the center position and radius size of the piston top surface circle, by comparing the detection value of the radius of the piston top surface circle with the radius value of the top surface circle in the piston model information, the validity of the detection result of the piston top surface circle is verified, and the detection result information is output.The method ensures the validity and reliability of the detection result.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of accurate identification and positioning method suitable for different models of furnace rear piston, especially to accurate positioning method of mechanical arm grabbing position in the automatic sorting process of curing furnace rear piston, belong to the technical field of automobile engine piston detection positioning. BACKGROUND

[0002] Piston is the "heart" of automobile engine, due to the variety, large batch, the automatic transfer problem of piston between different process equipment is an important problem to realize the automation of piston production. High temperature curing is an important means of piston surface treatment, and is also the last process of piston production. After high temperature curing treatment in the curing furnace, the piston is conveyed to the curing furnace outlet by the conveying chain plate, and is transferred and stored by manual or automatic sorting system. Influenced by many factors such as production cost and technical condition, the piston after curing treatment (referred to as furnace rear piston) is mainly sorted manually at present. This method not only has low efficiency and high labor intensity, but also is easy to cause personnel scalding.

[0003] With the rapid development and wide application of machine vision and industrial robot technology, automatic sorting technology has been applied and popularized in many fields such as warehousing, logistics and manufacturing, but it has not been effectively applied in furnace rear piston sorting. In addition to the factors of cost and benefit, there are also certain technical barriers. First, the furnace rear pistons are placed in rows on the conveying chain plate, and the distance between the pistons is very small. Influenced by many factors such as placement error and vibration, the position of the piston on the chain plate is uncertain. Therefore, the most commonly used teaching method of mechanical hand cannot be used for sorting, and the piston needs to be accurately identified and positioned to avoid collision, grabbing failure and other abnormal phenomena. Second, the piston skirt is not a regular cylinder, and the suitable part for grabbing is the piston head at a certain height from the placement surface. The piston top surface circle should be used as the reference for piston grabbing. Since there are many types of pistons, the height of the top surface of different types of pistons is quite different, and the calibration plane of visual detection is the chain plate plane on which the piston is placed. How to ensure the detection and positioning accuracy of different types of pistons under the condition of 2D visual detection is an important problem.

[0004] Although 3D vision technology has developed rapidly in recent years, its problems such as poor environmental light interference resistance, limited detection area and high cost have not been solved, and it is not suitable for furnace rear piston sorting at present. In addition, after long-term work of the equipment, the placement marks similar to the shape of the piston will be left on the conveying chain, which will interfere with and affect the effective identification of the piston. SUMMARY

[0005] In view of the shortcomings of the prior art, the present application provides an accurate identification and positioning method for different types of furnace rear piston sorting and grabbing, especially for different height furnace rear piston sorting and grabbing.

[0006] The technical scheme of the present application is as follows:

[0007] A precise identification and positioning method suitable for different types of furnace rear pistons, comprising the following steps:

[0008] (1) Install a 2D industrial camera above the curing furnace rear conveying piston chain plate, with the camera optical axis direction perpendicular to the chain plate, and then complete camera calibration;

[0009] (2) Collect image information of the piston on the chain plate using the 2D industrial camera, and pre-process the image information to eliminate information irrelevant to the piston;

[0010] (3) Detect the search area of the processed piston image information to exclude interference factors;

[0011] (4) Extract the piston top surface circle contour in the piston detection area determined after the search;

[0012] (5) According to the known piston height in the piston model information, use the projection transformation method to correct the pixel coordinates of each pixel point on the extracted piston top surface circle contour;

[0013] (6) Convert the pixel coordinates of each point on the piston top surface circle contour to world coordinates to obtain the contour point column of the piston top surface circle in the world coordinate system, and perform first least squares circle fitting on the contour point column to obtain the center coordinates and radius of the fitted circle. By comparing the difference between the distance of each point in the contour point column to the center of the fitted circle and the fitted radius, the noise points in the contour point column are removed;

[0014] (7) Perform second least squares circle fitting on the contour point column of the denoising processed piston top surface circle to obtain the center position and radius size of the piston top surface circle as the detection value of the center position and radius of the piston top surface circle;

[0015] (8) Verify the effectiveness of the piston top surface circle detection result by comparing the detection value of the piston top surface circle radius with the top surface circle radius value in the piston model information, and output the detection result information.

[0016] Preferably, the camera calibration in step (1) is completed by Zhang Dingyou calibration method, and the calibration plane is the placement plane of the conveying piston chain plate.

[0017] Preferably, in step (2), median filtering is used for denoising processing of the image information, and a 5x5 kernel is used for image erosion and expansion processing.

[0018] Preferably, in step (3), the image template of the same type of piston is set in advance, and the image template matching technology is used to complete the search and positioning of the piston detection area.

[0019] Preferably, the interference factors excluded in step (3) include the piston residual trace and the information of the piston top surface circle similar shape.

[0020] Preferably, the piston top surface circle profile extraction in step (4) refers to the profile point column of the piston top surface circle detected and extracted by using the Canny edge detection operator.

[0021] Preferably, the piston model information in step (5) includes the model, batch, height and top surface circle radius of the piston.

[0022] Preferably, the pixel coordinate correction process in step (5) is as follows:

[0023] The piston placement plane, which is also the camera calibration plane, is set as plane B, the plane where the piston top surface circle is located is set as plane S, plane S is parallel to plane B, the distance between the two planes is the height h of the piston; O c -X c Y c Z c is the camera coordinate system, the Z c axis of which is perpendicular to the calibration plane B, the origin O c is at a distance of the installation height H of the camera from plane B; o2-xy is the pixel coordinate system, and the origin o2 of the pixel coordinate system is on the Z c axis of the camera coordinate system, the x axis and the y axis of which are parallel to the X c axis and the Y c axis respectively; P s is a point on the piston top surface circle, P b is the projection point of P s on the calibration plane, P s ′ and P b ′ are the corresponding points of P s and P b in the pixel coordinate system after imaging by the camera, A s and A b are the projection points of P s and P b on the Z c axis respectively, then A s A b = h, and the following formula can be obtained

[0024]

[0025]

[0026] Since A b P b =A s P s , O cA b = H, combined with formula (1), (2) can get the point P on the top surface circle s Real pixel point P in pixel coordinate system s ′ satisfies the following formula:

[0027]

[0028] Let P b ′ pixel coordinates of the point (x b ,y b ), then the real pixel coordinates of the piston top circle P s point P s ′(x s ,y s ) coordinates are:

[0029]

[0030]

[0031] According to the above correction process, the pixel coordinates of the extracted piston top circle contour C b are corrected respectively, and the corrected real contour C s of the piston top circle is obtained in the pixel coordinate system.

[0032] Preferably, the pixel coordinate conversion to the world coordinate in step (6) is as follows:

[0033] Using the imaging principle, the pixel coordinates of each point P i (i=0, 1, 2, …, N-1, N is the number of pixel points on C s , N is a natural number) on the corrected piston top circle contour C s are converted into world coordinates to obtain a point sequence Q i (i=0, 1, 2, …, N-1) in the world coordinate system.

[0034] The least square circle fitting is performed on the point sequence Q i (i=0, 1, 2, …, N-1) to obtain the center point A0 of the fitting circle and the radius r0.

[0035] Preferably, the method for removing noise points in the contour point sequence in step (6) is as follows:

[0036] The distance d i (i=0, 1, 2, …, N-1) of each point in Q i (i=0, 1, 2, …, N-1) to the center A0 is calculated, and then the absolute value ε i of the difference between d i and the fitting radius r0 is calculated according to formula (6).

[0037] ε i = |d i -r i | (6)

[0038] If the ε i of a point in the point series is greater than a given error threshold ε0(0.005r0≤ε0≤0.01r0), it indicates that the point is a noise point and should be removed from the piston top surface circle contour point series; noise points are identified and removed from the piston top surface circle contour point series one by one to obtain a new contour point series W j (j = 0, 1, 2, …, M-1, M is the number of contour points after removing noise points, M is a natural number).

[0039] Preferably, step (7) uses a least squares circle fitting method to fit the point series W j , and obtains the center coordinates (x wA , y wA ) and radius r w of the fitting circle in the world coordinate system; the center coordinates and radius of the fitting circle are the detection values of the piston top surface circle in the 2D vision detection system.

[0040] Preferably, step (8) is based on the known piston model information of the piston top surface circle radius r, and if the difference between r and the detection value r w is less than a pre-set threshold ε r (0.5mm≤ε r ≤1mm), i.e. |r-r w |<ε r , it indicates that the detection result is valid, and the piston top surface circle center coordinates (x wA , y wA ) obtained by vision detection can be used as the positioning information of the piston and output for precise positioning when a mechanical hand is sorting and grabbing.

[0041] A server comprising:

[0042] one or more processors;

[0043] a storage device having stored thereon one or more programs,

[0044] when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned accurate identification and positioning method suitable for different models of back pistons of furnaces.

[0045] A computer readable medium having stored thereon a computer program, wherein the computer program is executed by a processor to implement the above-mentioned accurate identification and positioning method suitable for different models of back pistons of furnaces.

[0046] Technical features and benefits of the present application:

[0047] 1. The template matching technique is used to search the piston detection area in the image to complete the coarse positioning of the piston, effectively eliminating the interference of environmental factors, especially the influence of the piston traces left on the conveying chain plate, effectively ensuring the accuracy and efficiency of piston recognition.

[0048] 2. In view of the problem that 2D visual detection is difficult to intuitively reflect the height information of the object, based on the imaging principle, the projection transformation method is used to correct the image deviation caused by the fact that the piston detection plane and the calibration plane are not in the same plane, and only one calibration is needed to effectively detect pistons of different heights, providing reliable guarantee for accurate recognition and precise positioning of pistons of different heights.

[0049] 3. In view of the problem that the identification and positioning accuracy of the piston top surface circle are greatly affected by the contour noise, the twice least square circle fitting method is used to ensure the identification accuracy of the piston top surface circle. The distance fitting circle distance of the top surface circle obtained by the first fitting is used to eliminate the noise points with too large distance, and then the top surface circle point column after eliminating the noise points is re-performed least square circle fitting to obtain the accurate position information of the piston top surface circle, laying a foundation for subsequent sorting and grabbing. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 It is a schematic diagram of the installation of the 2D industrial camera of the present application;

[0051] Figure 2 It is a principle diagram of the position correction of the points on the top surface circle of the post-furnace piston of the present application;

[0052] Figure 3 It is a principle diagram of the contour position correction of the top surface circle of the post-furnace piston of the present application;

[0053] Figure 4 It is a flow chart of the accurate identification and positioning method of the post-furnace piston of different heights of the present application;

[0054] In the figure: 1-2D industrial camera, 2-camera mounting frame, 3-light source, 4-piston, 5-curing furnace. DETAILED DESCRIPTION

[0055] The present application will be further described below by examples and in conjunction with the drawings, but is not limited thereto.

[0056] Example 1:

[0057] As shown in the figure, the present embodiment provides an accurate identification and positioning method suitable for different types of post-furnace pistons, including the following steps: Figure 4

[0058] ​(1) Install a 2D industrial camera above the curing furnace conveying piston chain plate, and the camera optical axis direction is perpendicular to the chain plate, then complete the camera calibration;

[0059] The curing furnace is an existing curing furnace for high-temperature curing of piston surface. The camera installation method is designed and installed on the basis of the curing furnace equipment structure. The installation effect is shown in Figure 1 After installation, Zhang Dingyou calibration method is used to complete camera calibration. The calibration plane is the piston placement plane of the curing furnace conveying chain plate.

[0060] (2) Use a 2D industrial camera to collect image information of the piston conveyed from the curing furnace to the chain plate, and filter noise and filter the image information to eliminate information unrelated to the piston;

[0061] In order to obtain clear image information of the piston, first, use median filtering to denoise the image. In order to obtain a smoother target surface, use a 5x5 kernel to respectively perform erosion and expansion processing on the image.

[0062] (3) Detect the processed piston image information to search for the detection area and exclude interference factors;

[0063] In order to search for the effective piston detection area, use template matching to confirm. Use the pre-set piston template image of this model to traverse the detection area of each piston placement area in the image to obtain the similarity of the template and the target image in each detection area. Based on the size of the similarity, quickly confirm whether there is a piston missing in each detection area, exclude the interference of similar circular information (previous piston trace and piston top surface circular similar shape information), and effectively ensure the accuracy of piston detection.

[0064] (4) In the piston detection area determined after searching, use Canny edge detection operator to detect and extract the contour edge of the piston top surface circle.

[0065] (5) According to the known piston height in the piston model information, use projection transformation method to correct the pixel coordinates of each pixel point on the extracted piston top surface circle contour;

[0066] Since the detection target piston top surface circle and the calibration plane are not in the same plane, and the heights of different models of pistons are also different, it is necessary to correct the pixel coordinate values of the piston top surface circle contour edge extracted in step (4). The principle is shown in Figure 2 , and the specific method is as follows:

[0067] Set the piston placement plane, which is also the camera calibration plane, as plane B, and the plane where the piston top surface circle is located as plane S. Plane S is parallel to plane B, and the distance between the two planes is the height h of the piston. O c -Xc Y c Z c is the camera coordinate system, the Z c axis is perpendicular to the calibration plane B, the origin O c is the distance from the plane B, the installation height H of the camera; o 2- xy is the pixel coordinate system, and the origin o2 of the pixel coordinate system is on the Z c axis of the camera coordinate system, and the x axis and the y axis are parallel to the X c axis and the Y c axis respectively; P s is a point on the piston top surface circle, P b is the projection point of P s on the calibration plane, P s ' and P b ' are the corresponding points of P s and P b in the pixel coordinate system after imaging by the camera, A s and A b are the projection points of P s and P b on the Z c axis, and A s A b = h, so the following formula is obtained

[0068]

[0069]

[0070] Since A b P b = A s P s , O c A b = H, combined with formulas (1) and (2), the real pixel point P s ' of the point P s on the top surface circle in the pixel coordinate system satisfies the following formula:

[0071]

[0072] Let the pixel coordinates of P b ' be (x b , y b ), and the real pixel coordinates P s ' (x s , y s ) of the point P s on the top surface circle are as follows:

[0073]

[0074]

[0075] According to the above correction process, the pixel coordinates of the extracted piston top surface circle contour C b are respectively corrected to obtain the corrected piston top surface circle true contour C s in the pixel coordinate system.

[0076] (6) The pixel coordinates of each point on the piston top surface circle contour are converted to world coordinates to obtain a contour point list of the piston top surface circle in the world coordinate system, and the contour point list is subjected to first least square circle fitting to obtain the center coordinates and radius of the fitted circle. By comparing the difference between the distance of each point in the contour point list to the center of the fitted circle and the fitted radius, the noise points in the contour point list are removed.

[0077] The process of converting the pixel coordinates to world coordinates is as follows:

[0078] The pixel coordinates of each point P i (i = 0, 1, 2, …, N-1, N is the number of pixel points on C s , N is a natural number) on the corrected piston top surface circle contour C s are converted to world coordinates by using the imaging principle to obtain a point list Q i (i = 0, 1, 2, …, N-1) in the world coordinate system.

[0079] The point list Q i (i = 0, 1, 2, …, N-1) is subjected to least square circle fitting to obtain the center point A0 and the radius r0 of the fitted circle.

[0080] The method for removing noise points in the contour point list is as follows:

[0081] The distance d i (i = 0, 1, 2, …, N-1) of each point in Q i (i = 0, 1, 2, …, N-1) to the center A0 is calculated, and then the absolute value ε i of the difference between d i and the fitted radius r0 is calculated according to formula (6).

[0082] ε i = |d i -r i | (6)

[0083] If the ε i corresponding to a point in the point list is greater than a given error threshold ε0 (0.005r0≤ε0≤0.01r0), it indicates that the point is a noise point and should be removed from the piston top surface circle contour point list. Noise identification is performed on each point in the piston top surface circle contour point list in turn, and all noise points are removed to obtain a new contour point list W.j (j = 0, 1, 2, …, M-1, M is the number of contour points after removing noise points, M is a natural number).

[0084] (7) The second least square circle fitting is performed on the denoised piston top surface circle contour point series to obtain the center position and radius size of the piston top surface circle, which are taken as the detection values of the center position and radius of the piston top surface circle;

[0085] That is, the point series W j is fitted by using the least square circle fitting method to obtain the center coordinates (x wA , y wA ) and radius r w of the fitting circle in the world coordinate system, and the center coordinates and radius of the fitting circle are the detection values of the piston top surface circle in the 2D vision detection system.

[0086] (8) The effectiveness of the piston top surface circle detection result is verified by comparing the detection value of the piston top surface circle radius with the top surface circle radius value in the piston model information, and the detection result information is output.

[0087] Specifically, according to the piston top surface circle radius r in the known piston model information, if the difference between r and the detection value r w is less than a pre-set threshold value ε r (0.5mm≤ε r ≤1mm), i.e., |r-r w |<ε r , it indicates that the detection result is effective, and the piston top surface circle center coordinates (x wA , y wA ) obtained by vision detection can be taken as the positioning information of the piston and output for precise positioning when the mechanical hand is sorting and grabbing.

[0088] Since the output top surface circle center coordinates correct the image distortion caused by the inconsistency between the piston top surface circle and the camera calibration surface, the output piston position information is completely consistent with the actual position, and precise positioning of pistons of different models, especially different heights, can be achieved. If the difference between the detected piston top surface circle radius r w and the radius r in the piston model information is greater than or equal to the threshold value ε r , i.e., |r-r w |≥ε r , it indicates that the detection result is abnormal, and the detection abnormal information is output for decision-making processing by the control system.

[0089] The present application has reasonable concept, can standardize and adaptively accurately recognize and accurately position different models, especially different height pistons, can realize accurate extraction of the center coordinates and radius of the top surface circle of different height pistons once calibration is completed, and can check the detection result according to known piston information, thereby ensuring effectiveness and reliability of the detection result.

[0090] The above merely describes a specific implementation of the present application, and the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which shall be covered within the protection scope of the present application.

Claims

1. A method for accurate identification and positioning of different models of furnace rear piston, characterized in that, The method comprises the following steps: (1) installing a 2D industrial camera above the curing furnace, with the camera optical axis perpendicular to the chain plate, and then completing camera calibration; (2) collecting image information of the piston on the chain plate by using the 2D industrial camera, and pre-processing the image information to eliminate information irrelevant to the piston; (3) searching for a detection area of the piston after the pre-processed piston image information, and excluding interference factors; (4) extracting a piston top surface circle contour in the piston detection area determined after the search; (5) correcting pixel coordinates of each pixel point on the extracted piston top surface circle contour according to the piston height in the known piston model information by using a projection transformation method; the pixel coordinate correction process is as follows: The setting piston placement plane, also the camera calibration plane is plane B, the piston top surface circle plane is plane S, plane S is parallel to plane B, the distance between the two planes is the height h of the piston; O c -X c Y c Z c is the camera coordinate system, the Z c axis is perpendicular to the calibration plane B, the origin O c is the distance from the plane B, the installation height of the camera is H; o2-xy is the pixel coordinate system, and the origin o2 of the pixel coordinate system is on the Z c axis of the camera coordinate system, the x axis and the y axis are respectively parallel to the X c axis and the Y c axis; P s is a point on the piston top surface circle, P b is the projection point of P s on the calibration plane, P s ' and P b ' are respectively P s and P b through the camera imaging in the pixel coordinate system, A s and A b are respectively P s and P b in the Z c axis projection point, then A s A b =h, the following formula can be obtained Since A b P b = A s P s , O c A b = H, combined with formula (1), (2) can be obtained on the top surface of the circle point P s In the pixel coordinate system, the real pixel point P s ' satisfies the following formula: Let P b be a point on the piston top surface, and P b ′ be the pixel coordinate of P b , then the real pixel coordinate P s ′(x s ,y s ) of P s on the piston top surface circle is: According to the above-mentioned modified process, the pixel coordinates of the extracted piston top surface circle contour C b are respectively modified to obtain the modified piston top surface circle true contour C s In the pixel coordinate system; (6) converting the pixel coordinates of each point on the piston top surface circle contour to world coordinates to obtain a contour point column of the piston top surface circle in a world coordinate system, performing first least square circle fitting on the contour point column to obtain a center coordinate and a radius of the fitted circle, and removing noise points in the contour point column by comparing a distance of each point in the contour point column to the center of the fitted circle with a difference value of the fitted radius; the pixel coordinate conversion to the world coordinate is as follows: The modified piston top surface circle contour C s The pixel coordinates of the points P i are converted into world coordinates to obtain a point series Q in the world coordinate system i , where i=0, 1, 2, …, N-1, N is the number of pixel points C s , and N is a natural number; To the point column Q i A least square circle fitting is performed, where i = 0, 1, 2, …, N-1, to obtain a center point A0of the fitted circle and a radius r0. the method for removing noise points in the contour point column is as follows: Q is calculated for each point i the distance d of each point in the set to the center A0 of the circle i where i = 0, 1, 2,..., N - 1, and then d is calculated according to equation (6) i the absolute value of the difference between the distance d and the fitted radius r0, ε i : e i = |d i -r i | (6) If a certain point in the point series corresponds to ε i greater than a given error threshold ε0, where 0.005r0≤ε0≤0.01r0, it indicates that the point is a noise point and should be removed from the piston top surface circle contour point series; noise identification is performed on each point in the piston top surface circle contour point series in turn, and all noise points are removed to obtain a new contour point series W j where j=0,1,2,…,M-1, M is the number of contour points after removing noise points, and M is a natural number. (7) performing second least square circle fitting on the contour point column of the piston top surface circle after noise reduction processing to obtain a center position and a radius size of the piston top surface circle, as detection values of the center position and the radius of the piston top surface circle; The least square circle fitting method is used to fit the point column W j , and the center coordinate (x wA , y wA ) and the radius r w of the fitting circle in the world coordinate system are obtained, wherein the center coordinate and the radius of the fitting circle are the detection values of the piston top surface circle in the 2D vision detection system. (8) verifying validity of the detection result of the piston top surface circle by comparing the detection value of the radius of the piston top surface circle with a radius value of the piston top surface circle in the piston model information, and outputting detection result information.

2. The method for accurate identification and positioning of different models of oven rear pistons as claimed in claim 1, wherein, The camera calibration in the step (1) is completed by using the Zhang Dingyou calibration method, and the calibration plane is a placement plane of the piston chain plate.

3. The method for accurate identification and positioning of different models of oven rear pistons as claimed in claim 1, wherein, In the step (2), median filtering is used for denoising processing of the image information, and a 5x5 kernel is used for image erosion and expansion processing.

4. The method for accurate identification and positioning of different models of oven rear pistons as claimed in claim 1, wherein, In the step (3), the image template of the same model piston is set in advance for the detection area search, and the image template matching technology is used to complete the search and positioning of the piston detection area.

5. The method for accurate identification and positioning of different models of oven rear pistons as claimed in claim 1, wherein, In the step (4), the piston top surface circle contour extraction refers to detecting and extracting a contour point column of the piston top surface circle by using a Canny edge detection operator.

6. The method for accurate identification and positioning of different models of oven rear pistons as claimed in claim 1, wherein, The step (8) determines whether the difference between the piston top surface circle radius r in the known piston model information and the detected value r w is less than a preset threshold value ε r , where 0.5mm≤ε r ≤1mm, i.e., |r-r w |<ε r , indicating that the detection result is valid. In this case, the piston top surface circle center coordinates (x wA ,y wA ) obtained through visual detection are output as the positioning information of the piston, for accurate positioning when the mechanical hand sorts and grabs.

Images