A method for compensating errors by infinitely expanding the vision field of a mounter camera

Abstract

The present invention discloses a method for compensating errors by infinitely expanding the vision field of a mounter camera, which relates to the technical field of chip mounting. The method for compensating errors by infinitely expanding the vision field of a mounter camera includes: S1, establishing a coordinate system and extracting feature points; S2, calculating a homogeneous coordinate matrix and performing coordinate transformation; S3, expanding the vision field; S4, analyzing and compensating errors; S5, performing a mounting operation; by infinitely expanding the virtual vision field of the camera, the traditional cumbersome switching process that is prone to introducing errors is avoided. First, a dual coordinate system is defined and feature points are extracted, a homogeneous coordinate matrix is established to complete coordinate transformation and distortion correction, and then new feature points are continuously selected to achieve infinite expansion of the vision field to cope with a large-range chip substrate. Then, the deviation of the mounting position is analyzed and the nature of the error is judged, and the error is compensated by respectively recalibrating the parameters of the H matrix or the method of taking the average of multiple measurements. Finally, the compensated coordinates are output to the moving module of the mounter to complete the mounting.

Description

Technical Field

[0001] The present invention relates to the technical field of chip mounting, and specifically to a method for compensating errors by infinitely expanding the camera field of view of a mounter. Background Art

[0002] In the current field of precision electronics manufacturing, as a core process link, chip mounting has almost stringent requirements for mounting accuracy. In order to achieve high-precision chip mounting positioning, the existing chip mounters usually design the camera field of view to be relatively narrow. This is because in a small field of view scenario, the camera can more focusedly image the features of the chip and the substrate within a specific area. By means of high-resolution imaging and refined image processing algorithms, the positioning deviation caused by imaging blur, edge distortion, etc. can be effectively reduced, so as to achieve extremely high mounting accuracy indicators within a limited space range and meet the manufacturing requirements of many current high-integration, micro-miniaturized electronic products with extremely low tolerance for chip mounting position errors.

[0003] However, with the increasing complexity of electronic product functions and the continuous expansion of scale, the size of the chip substrate and the range of mounting layouts have also increased significantly. When the position of the chip to be mounted exceeds the current limited field of view of the camera, the obvious limitations of the conventional calculation method are exposed. In actual mounting operations, since the chip substrate coordinate system, the camera coordinate system, and the mechanical module coordinate system are established based on different physical benchmarks and logical rules respectively, they each have independent origins, axial definitions, and unit scales. In the conventional method, in order to accurately transfer the chip position information outside the field of view to the mechanical module to complete the mounting action, it is necessary to frequently switch back and forth between these three coordinate systems for calculation.

[0004] In the process of converting from the chip substrate coordinate system to the camera coordinate system, first, it is necessary to perform a preliminary projection transformation on the coordinates of the target chip on the chip substrate according to the pre-calibrated spatial relationship model. This process involves complex matrix operations and has extremely high requirements for the accuracy of the calibration parameters. Any small calibration error will be amplified in the calculation. Then, after the camera completes the acquisition of chip information within a partial field of view, the chip position information processed in the camera coordinate system needs to be converted to the mechanical module coordinate system so that the actuator such as the robotic arm can accurately position and grasp the chip. During this period, not only geometric transformations such as translation and rotation between coordinate systems need to be considered, but also problems such as scale differences and unit conversions that may exist in different coordinate systems need to be addressed.

[0005] This calculation method of switching back and forth between multiple coordinate systems may introduce new error sources in each conversion step. Calibration errors, rounding errors of coordinate transformation algorithms, dynamic errors during the movement of mechanical modules, etc. accumulate and superimpose during multiple coordinate system switches, ultimately resulting in a large deviation between the actual chip mounting position and the theoretical design position, seriously affecting the quality of chip mounting and the product yield, increasing the rework cost and time cost in the production process, and restricting the further development of the chip mounting process towards larger scale and higher precision. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a method for compensating errors by infinitely expanding the field of view of a pick-and-place machine camera, which solves the problems.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for compensating errors by infinitely expanding the field of view of a pick-and-place machine camera, comprising the following steps:

[0008] S1. Define the chip substrate coordinate system and the camera coordinate system, and use the SIFT algorithm to automatically extract the feature points with scale and rotation invariance in the chip substrate image, and obtain the coordinates of the feature points in the two coordinate systems;

[0009] S2. Establish a homogeneous coordinate matrix H to describe the transformation relationship from the chip substrate coordinate system to the camera coordinate system, transform the points in the chip substrate coordinate system to the camera coordinate system through the matrix H, and consider the camera optical distortion, and use a polynomial distortion model to correct the transformed coordinates to eliminate the distortion effect; [[ID=1�]]

[0010] S3. Continuously select new feature points of the chip substrate coordinate system, repeat the coordinate transformation process, and realize the infinite expansion of the camera field of view to process the chip substrate beyond the actual field of view of the camera;

[0011] S4. Analyze the deviation between the actual position and the theoretical position of the chip after mounting, and judge the nature of the error; for systematic errors, re-calibrate the parameters of the H matrix; for random errors, use the method of taking the average of multiple measurements; apply the error compensation strategy to the coordinate transformation process and correct the transformed coordinates;

[0012] S5. Output the camera coordinates after error compensation to the pick-and-place machine moving module for chip mounting work.

[0013] Preferably, the S1 specifically includes coordinate system establishment and feature point selection:

[0014] S1.1 Establish a coordinate system:

[0015] Define the chip substrate coordinate system , the origin O b is selected at one of the corner points of the chip substrate, Xb Axis and Y b The axes are respectively along two sides of the chip substrate;

[0016] Define the camera coordinate system , with the origin O c being the optical center of the camera, and the X c axis and Y c axis are respectively parallel to the horizontal and vertical directions of the camera imaging plane;

[0017] Feature point selection:

[0018] Use the SIFT feature point detection algorithm to process the chip substrate image and automatically extract feature points with scale invariance and rotation invariance;

[0019] Suppose n feature points are detected from the chip substrate image, and their coordinates in the chip substrate coordinate system are represented as , and their corresponding coordinates in the camera coordinate system are represented as .

[0020] Preferably, the homogeneous coordinate matrix H in S2 is:

[0021] ;

[0022] where λ is the scaling factor, representing the scale transformation ratio from the chip substrate coordinate system to the camera coordinate system; θ is the rotation angle, representing the rotation relationship between the two coordinate systems; and are respectively the translation amounts of the feature point in the X direction and Y direction of the two coordinate systems.

[0023] Preferably, S3 specifically includes coordinate transformation and field of view expansion:

[0024] S3.1 Coordinate transformation:

[0025] For any point in the chip substrate coordinate system, its homogeneous coordinate representation is , and its coordinates transformed to the camera coordinate system through the homogeneous coordinate matrix H are:

[0026] ;

[0027] Considering the optical distortion of the camera, a polynomial distortion model is used for correction; suppose the distortion model is:

[0028] ;

[0029] ;

[0030] where, , k1, k2, k3... are distortion coefficients, which are determined through calibration experiments;

[0031] S3.2 Field of view expansion: Continuously select new feature points in the chip substrate coordinate system and repeat the above coordinate transformation process to achieve infinite expansion of the camera's field of view.

[0032] Preferably, the error analysis step in S4 includes:

[0033] S4.1 Analyze the error source by actually measuring the deviation between the actual position and the theoretical position of the mounted chip; Let the theoretical position be , and the actual measured position be , then the error , ;

[0034] S4.2 Analyze the nature of the error and determine whether it is a systematic error or a random error; Systematic errors are regular, and random errors are random.

[0035] Preferably, the error compensation strategy in S4 includes:

[0036] For systematic errors, if a fixed deviation related to the H matrix is found, correct it by recalibrating the parameters in the H matrix; Let the deviation of the translation amount be ΔT x and ΔT y , then the corrected homogeneous coordinate matrix is:

[0037] ;

[0038] For random errors, reduce the impact by taking the average of multiple measurements and data smoothing processing;

[0039] Error compensation implementation: Apply the formulated error compensation strategy to the coordinate transformation process to correct the transformed coordinates.

[0040] Preferably, S5 specifically includes:

[0041] S5.1 Coordinate output: Output the coordinates (x c ′, y c ′) in the camera coordinate system after error compensation and distortion processing to the moving module of the mounter;

[0042] S5.2 Moving and mounting: The mounter controls the moving module to accurately mount the chip to the specified position on the substrate according to the received coordinate information.

[0043] Preferably, in the error nature analysis, the analysis of error regularity includes:

[0044] S4.2.1 Data collection:

[0045] Multiple measurements: Under the same experimental conditions, perform m measurements on the same mounting position, and record the actual position coordinates of each measurement. ;

[0046] S4.2.2 Analyze the regularity of errors:

[0047] Calculate the mean and variance of the errors:

[0048] Mean:

[0049] ;

[0050] ;

[0051] Variance:

[0052] ;

[0053] ;

[0054] Draw an error distribution graph: Use the measurement number i as the horizontal axis and the error and as the vertical axis to draw a graph of the change of error with the measurement number, or draw a histogram or scatter plot of the error to observe the distribution of the error;

[0055] Calculate the autocorrelation coefficient: The autocorrelation coefficient is used to reflect the correlation of errors between different time points or measurement numbers:

[0056] ;

[0057] Among them, represents the covariance of Δx i and , represents the standard deviation, and a represents the lag order.

[0058] Preferably, the systematic error judgment includes: If the error mean or is significantly non-zero, and the error distribution graph shows an obvious trend or periodic change in the error, or the autocorrelation coefficient is significantly non-zero at a certain lag order a, it indicates the existence of systematic errors.

[0059] Preferably, the random error judgment includes: If the error mean or is close to zero, and the error distribution graph shows a random distribution of the error without an obvious trend or periodic change, and at the same time the autocorrelation coefficient Approaching zero at all lag orders a indicates the presence of random errors.

[0060] The present invention provides a method for compensating errors by infinitely expanding the vision of a mounter camera. Compared with the prior art, it has the following beneficial effects:

[0061] 1. For the method of compensating errors by infinitely expanding the vision of a mounter camera, in the conventional method, due to the small vision of the camera, when processing the chip mounting positions beyond the vision, it is necessary to switch back and forth between the camera coordinate system, the chip coordinate system and the mechanical module for calculation, introducing large errors. However, in this solution, by infinitely expanding the virtual vision of the camera, this cumbersome and error-prone switching process is avoided. First, a dual coordinate system is defined and feature points are extracted, and a homogeneous coordinate matrix is established to complete coordinate transformation and distortion correction. Then, new feature points are continuously selected to achieve infinite expansion of the vision to cope with large-range chip substrates. Next, the mounting position deviation is analyzed and the nature of the error is judged, and the error compensation is carried out by re-calibrating the parameters of the H matrix or the method of taking the average of multiple measurements respectively. Finally, the compensated coordinates are output to the mounter moving module to complete the mounting. This solution improves the mounting efficiency and accuracy and reduces the errors.

[0062] 2. For the method of compensating errors by infinitely expanding the vision of a mounter camera, in the establishment of the coordinate system and the selection of feature points, the chip substrate and the camera coordinate system are accurately defined, and the SIFT algorithm is used to extract feature points with scale and rotation invariance, providing a reliable basis for subsequent coordinate transformation and ensuring the accurate matching of feature points under different perspectives. The establishment of the homogeneous coordinate matrix H clearly describes the transformation relationship between coordinate systems. Combined with the polynomial distortion model correction, the influence of camera optical distortion on coordinate transformation is effectively eliminated, improving the accuracy of coordinate transformation. By continuously selecting new feature points to achieve infinite expansion of the camera vision, the problem that the small camera vision is difficult to cover a large-range chip substrate is solved, and there is no need to switch coordinate systems back and forth, improving the processing efficiency. The error analysis can accurately locate the error source and judge the nature, and correspondingly adopt strategies such as re-calibrating the parameters of the H matrix or taking the average of multiple measurements and data smoothing processing for compensation, reducing the mounting error. Finally, the compensated coordinates are output to the mounter moving module, realizing high-precision mounting of the chip at the specified position on the substrate, improving the mounting quality and production efficiency, and reducing the defective rate.

[0063] 3. The method for compensating errors in the camera field of view of an infinitely expandable mounter can accurately judge the nature of errors based on error analysis. If there are systematic errors, their sources can be identified, such as mechanical installation, camera calibration, or environmental factors, etc., and then corresponding measures can be taken to correct them. If they are random errors, the causes of their generation can be understood, such as measurement noise, electromagnetic interference, etc., and methods such as taking the average of multiple measurements can be used to reduce the influence. This accurate judgment and effective processing of the nature of errors significantly improve the mounting accuracy, reduce the mounting deviation caused by errors, ensure the chip mounting quality, reduce the defective rate, and improve the production efficiency and product reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is a schematic flow chart of the steps of the present invention;

[0065] Figure 2 is a top view schematic diagram of two coordinate systems of the present invention;

[0066] Figure 3 is a three-dimensional schematic diagram of two coordinate systems of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0068] Refer to Figures 1 to 3 , the present invention provides the following three technical solutions:

[0069] The first embodiment: A method for compensating errors in the camera field of view of an infinitely expandable mounter, including the following steps:

[0070] S1. Define the chip substrate coordinate system and the camera coordinate system, and use the SIFT algorithm to automatically extract the feature points with scale and rotation invariance in the chip substrate image, and obtain the coordinates of the feature points in the two coordinate systems;

[0071] S2. Establish a homogeneous coordinate matrix H to describe the transformation relationship from the chip substrate coordinate system to the camera coordinate system, transform the points in the chip substrate coordinate system to the camera coordinate system through the matrix H, and consider the camera optical distortion, and use the polynomial distortion model to correct the transformed coordinates to eliminate the distortion influence;

[0072] S3. Continuously select new feature points of the chip substrate coordinate system, and repeat the coordinate transformation process to achieve infinite expansion of the camera field of view to process the chip substrate beyond the actual camera field of view;

[0073] S4. Analyze the deviation between the actual position and the theoretical position of the chip after placement, and determine the nature of the error; for systematic errors, recalibrate the H matrix parameters; for random errors, use the method of taking the average of multiple measurements; apply the error compensation strategy to the coordinate transformation process and correct the transformed coordinates.

[0074] S5. Output the camera coordinates after error compensation to the placement machine moving module for chip placement work.

[0075] Due to the small camera field of view in the conventional method, when dealing with the positions of chips to be placed outside the field of view, it is necessary to switch back and forth between the camera coordinate system, the chip coordinate system and the mechanical module for calculation, introducing large errors. However, this solution avoids this cumbersome and error-prone switching process by infinitely expanding the virtual field of view of the camera. First, define a dual coordinate system and extract feature points, establish a homogeneous coordinate matrix to complete coordinate transformation and distortion correction, and then continuously select new feature points to achieve infinite expansion of the field of view to cope with large-scale chip substrates. Then analyze the placement position deviation and determine the nature of the error, and use the method of recalibrating the H matrix parameters or taking the average of multiple measurements for error compensation respectively. Finally, output the compensated coordinates to the placement machine moving module to complete the placement. This solution improves the placement efficiency and accuracy and reduces errors.

[0076] The second implementation mode is mainly different from the first implementation mode in that: S1 specifically includes coordinate system establishment and feature point selection:

[0077] S1.1 Establish a coordinate system:

[0078] Define the chip substrate coordinate system , with the origin O b selected at one of the corner points of the chip substrate, and the X b axis and the Y b axis are respectively along two sides of the chip substrate;

[0079] Define the camera coordinate system , with the origin O c being the camera optical center, and the X c axis and the Y c axis are respectively parallel to the horizontal and vertical directions of the camera imaging plane;

[0080] Feature point selection:

[0081] Use the SIFT feature point detection algorithm to process the chip substrate image and automatically extract feature points with scale invariance and rotation invariance;

[0082] Suppose n feature points are detected from the chip substrate image, and their coordinates in the chip substrate coordinate system are represented as , and their corresponding coordinates in the camera coordinate system are represented as .

[0083] The homogeneous coordinate matrix H in S2 is as follows:

[0084] ;

[0085] where λ is the scaling factor, representing the scale transformation ratio from the chip substrate coordinate system to the camera coordinate system; θ is the rotation angle, representing the rotation relationship between the two coordinate systems; and are the translation amounts of the feature point in the X - direction and Y - direction of the two coordinate systems respectively.

[0086] In the establishment of the coordinate system and the selection of feature points, the chip substrate and camera coordinate systems are accurately defined, and the SIFT algorithm is used to extract feature points with scale and rotation invariance, providing a reliable basis for subsequent coordinate transformation and ensuring the accurate matching of feature points from different perspectives.

[0087] S3 specifically includes coordinate transformation and field - of - view expansion:

[0088] S3.1 Coordinate transformation:

[0089] For any point in the chip substrate coordinate system, its homogeneous coordinate representation is , and it is transformed to the coordinate in the camera coordinate system through the homogeneous coordinate matrix H as follows:

[0090] ;

[0091] Considering the optical distortion of the camera, a polynomial distortion model is used for correction; the distortion model is set as:

[0092] ;

[0093] ;

[0094] where , and k1, k2, k3... are distortion coefficients, which are determined through calibration experiments;

[0095] S3.2 Field - of - view expansion: Continuously select new feature points in the chip substrate coordinate system and repeat the above coordinate transformation process to achieve infinite expansion of the camera's field of view.

[0096] The establishment of the homogeneous coordinate matrix H clearly describes the transformation relationship between coordinate systems. Combined with the correction of the polynomial distortion model, it effectively eliminates the influence of camera optical distortion on coordinate transformation and improves the accuracy of coordinate transformation. By continuously selecting new feature points to achieve infinite expansion of the camera's field of view, the problem that the camera's field of view is small and difficult to cover a large - range chip substrate is solved. There is no need to switch coordinate systems back and forth, improving the processing efficiency.

[0097] The error analysis steps in S4 include:

[0098] S4.1 Analyze the error source by measuring the deviation between the actual position and the theoretical position of the mounted chip; let the theoretical position be , and the actual measured position be , then the error , ;

[0099] S4.2 Analyze the nature of the error to determine whether it is a systematic error or a random error; systematic errors are regular, such as those caused by mechanical installation errors, camera calibration errors, etc.; random errors are random, such as measurement noise, etc.

[0100] The error compensation strategy in S4 includes:

[0101] For systematic errors, if a fixed deviation related to the H matrix is found (for example, there is a deviation in the translation amount T x , T y ), correct it by recalibrating the parameters in the H matrix; let the deviation in the translation amount be ΔT x and ΔT y , and the angle between the two coordinate systems is θ (θ is caused by the torsional error of the camera's own installation or the deflection transmitted from the chip substrate), then the corrected homogeneous coordinate matrix is:

[0102] ;

[0103] For random errors, use the method of taking the average of multiple measurements and data smoothing to reduce the impact;

[0104] Error compensation implementation: Apply the formulated error compensation strategy to the coordinate transformation process to correct the transformed coordinates.

[0105] Error analysis can accurately locate the error source and determine the nature, and specifically adopt strategies such as recalibrating the H matrix parameters or taking the average of multiple measurements and data smoothing for compensation, reducing the mounting error.

[0106] S5 specifically includes:

[0107] S5.1 Coordinate output: Output the coordinates (x c ′, y c ′) in the camera coordinate system after error compensation and distortion processing to the movement module of the mounter;

[0108] S5.2 Moving and mounting: The mounter controls the movement module to accurately mount the chip to the specified position on the substrate according to the received coordinate information.

[0109] Finally, the compensated coordinates are output to the moving module of the mounter, achieving high-precision mounting of the chip at the specified position on the substrate, improving the mounting quality and production efficiency, and reducing the defective rate.

[0110] The third implementation mode is mainly different from the first implementation mode in that in the error property analysis, the error regularity analysis includes:

[0111] S4.2.1 Data collection:

[0112] Multiple measurements: Under the same experimental conditions, measure the same mounting position m times, and record the actual position coordinates of each measurement ;

[0113] S4.2.2 Analyze the error regularity:

[0114] Calculate the mean and variance of the error:

[0115] Mean:

[0116] ;

[0117] ;

[0118] Variance:

[0119] ;

[0120] ;

[0121] Draw an error distribution diagram: Take the number of measurements i as the horizontal axis and the error and as the vertical axis to draw a graph of the change of the error with the number of measurements, or draw a histogram or scatter plot of the error to observe the distribution of the error;

[0122] Calculate the autocorrelation coefficient: The autocorrelation coefficient is used to reflect the correlation between errors at different time points (or number of measurements):

[0123] ;

[0124] where represents the covariance of Δx i and , represents the standard deviation of , and a represents the lag order.

[0125] The systematic error judgment includes: If the error mean or is significantly non-zero, and the error distribution diagram shows an obvious trend or periodic change in the errors, or the autocorrelation coefficient is significantly non-zero at a certain lag order a, indicating the existence of systematic errors; systematic errors may be caused by mechanical installation errors, camera calibration errors, environmental factors (such as temperature and humidity changes), etc.;

[0126] The judgment of random errors includes: If the error mean or is close to zero, and the error distribution diagram shows that the errors are randomly distributed without obvious trends or periodic changes, and at the same time the autocorrelation coefficient is close to zero at all lag orders a, indicating the existence of random errors; random errors may be caused by measurement noise, electromagnetic interference, operator differences, etc.

[0127] By measuring the same mounting position multiple times and recording the actual position coordinates, it provides a sufficient and reliable data basis for error analysis. Calculating the error mean and variance can intuitively understand the overall level and dispersion degree of the errors, and initially judge the error distribution characteristics. Drawing the error distribution diagram presents the change or distribution of the errors with the number of measurements in an intuitive graphical way, which is convenient for discovering whether there are abnormalities such as trends and periodicity in the errors. Calculating the autocorrelation coefficient can further explore the correlation between the errors at different time points (or measurement times) and judge whether the errors have internal associations.

[0128] Based on these analyses, the nature of the errors can be accurately judged. If there are systematic errors, their sources can be identified, such as mechanical installation, camera calibration or environmental factors, etc., and then corresponding measures can be taken to correct them; if they are random errors, the causes can be understood, such as measurement noise, electromagnetic interference, etc., and methods such as taking the average of multiple measurements can be used to reduce the influence. This accurate judgment and effective handling of the nature of the errors significantly improve the mounting accuracy, reduce the mounting deviation caused by errors, ensure the chip mounting quality, reduce the defective rate, and improve the production efficiency and product reliability.

[0129] Example:

[0130] After our factory adopted this technical solution, compared with the original solution, the comparison data is shown in Table 1:

[0131] Table 1 Comparison table of data between the new and old solutions

[0132]

[0133] To sum up:

[0134] Mounting accuracy: The existing solution has significantly improved in mounting accuracy, from the original 15 ± 3μm to 5 ± 1μm.

[0135] Field of view: The field of view of the original solution was fixed at 400 mm² (20x20 mm), while the existing solution can theoretically achieve infinite expansion.

[0136] Processing time: The processing time of the existing solution has been significantly reduced, from the original 60 seconds / substrate to 30 seconds / substrate.

[0137] Analysis of error sources: The existing solution divides errors into systematic errors and random errors, and effective compensation has been carried out respectively, resulting in a significant reduction in both types of errors.

[0138] Error compensation strategy: The existing solution introduces compensation strategies for systematic errors and random errors, improving the accuracy of placement.

[0139] System complexity: By simplifying the process and reducing the number of coordinate system switches, the system complexity of the existing solution has been significantly reduced.

[0140] Defective rate: The improvement of placement accuracy has directly led to a significant reduction in the defective rate, from the original 5% to 1%.

[0141] Maximum substrate size for processing: The existing solution can process larger-sized chip substrates and is theoretically not limited by the field of view.

[0142] Calibration time and accuracy: The calibration time of the existing solution has been reduced by half, and the calibration accuracy has also been improved.

[0143] Long-term stability: The existing solution has also been significantly improved in terms of long-term stability, and the trouble-free operation time has increased from the original 100 hours to 500 hours.

[0144] At the same time, the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0145] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0146] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for compensating errors by infinitely expanding the vision field of a mounter camera, characterized in that Including the following steps: S1. Define the chip substrate coordinate system and the camera coordinate system, automatically extract the feature points with scale and rotation invariance in the chip substrate image using the SIFT algorithm, and obtain the coordinates of the feature points in the two coordinate systems; S2. Establish the homogeneous coordinate matrix H to describe the transformation relationship from the chip substrate coordinate system to the camera coordinate system, transform the points in the chip substrate coordinate system to the camera coordinate system through the matrix H, and considering the camera optical distortion, use the polynomial distortion model to correct the transformed coordinates to eliminate the distortion effect; S3. Continuously select new feature points in the chip substrate coordinate system, repeat the coordinate transformation process, and achieve infinite expansion of the camera's field of view to process the chip substrate beyond the actual field of view of the camera; S4. Analyze the deviation between the actual position and the theoretical position of the chip after placement, and judge the nature of the error; for systematic errors, recalibrate the parameters of the H matrix; for random errors, use the method of taking the average of multiple measurements; apply the error compensation strategy to the coordinate transformation process to correct the transformed coordinates; S5. Output the error-compensated camera coordinates to the placement machine moving module for chip placement work.

2. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, according to claim 1, characterized in that: The specific content of S1 includes coordinate system establishment and feature point selection: S1.1 Establish the coordinate system: Define the coordinate system of the chip substrate , the origin O b is selected at one of the corner points of the chip substrate, and the X b axis and the Y b axis are respectively along two sides of the chip substrate; Define the camera coordinate system , the origin O c is the optical center of the camera, and the X c axis and the Y c axis are parallel to the horizontal and vertical directions of the camera imaging plane respectively; Feature point selection: Use the SIFT feature point detection algorithm to process the chip substrate image and automatically extract the feature points with scale invariance and rotation invariance; Suppose that n feature points are detected from the chip substrate image, and their coordinates in the chip substrate coordinate system are expressed as , and their corresponding coordinates in the camera coordinate system are expressed as .

3. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, as claimed in claim 2, wherein: The homogeneous coordinate matrix H in S2 is: ； Among them, λ is the scaling factor, representing the scale transformation ratio from the chip substrate coordinate system to the camera coordinate system; θ is the rotation angle, representing the rotation relationship between the two coordinate systems; and are the translation amounts of the feature points in the X and Y directions of the two coordinate systems, respectively.

4. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, according to claim 3, characterized in that: The specific content of S3 includes coordinate transformation and field of view expansion: S3.1 Coordinate transformation: For any point in the chip substrate coordinate system , the homogeneous coordinate representation is , and the coordinates transformed to the camera coordinate system through the homogeneous coordinate matrix H are : ； Considering the optical distortion of the camera, use the polynomial distortion model for correction; assume the distortion model is: ； ； Among them, , k1, k2, k3... are distortion coefficients, which are determined by calibration experiments; S3.2 Field of view expansion: Continuously select new feature points in the chip substrate coordinate system and repeat the above coordinate transformation process to achieve infinite expansion of the camera's field of view.

5. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, as claimed in claim 1, wherein: The error analysis steps in S4 include: S4.1 Analyze the error source by actually measuring the deviation between the chip position after mounting and the theoretical position; assume the theoretical position is , and the actually measured position is , then the error , ; S4.2 Analyze the nature of the error and judge whether it is a systematic error or a random error; systematic errors are regular, and random errors are random.

6. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, as described in claim 5, characterized in that: The error compensation strategy in S4 includes: For systematic errors, if a fixed deviation related to the H matrix is found, it is corrected by recalibrating the parameters in the H matrix; let the deviation of the translation amount be ΔT x and ΔT y , then the corrected homogeneous coordinate matrix is: ； For random errors, use the methods of taking the average of multiple measurements and data smoothing to reduce the influence; Error compensation implementation: Apply the formulated error compensation strategy to the coordinate transformation process to correct the transformed coordinates.

7. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, as claimed in claim 1, wherein: The specific content of S5 includes: S5.1 Coordinate Output: Output the coordinates (x c ′, y c ′) in the camera coordinate system after error compensation and distortion processing to the movement module of the mounter; S5.2 Moving and placing: The placement machine controls the moving module to accurately place the chip at the specified position on the substrate according to the received coordinate information.

8. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, according to claim 5, characterized in that: In the error nature analysis, the analysis of error regularity includes: S4.2.1 Data collection: Multiple measurements: Under the same experimental conditions, perform m measurements on the same mounting position and record the actual position coordinates of each measurement ; S4.2.2 Analyze the error regularity: Calculate the mean and variance of the error: Mean: ； ； Variance: ； ； Plot the error distribution diagram: Use the number of measurements \(i\) as the horizontal axis and the error and as the vertical axis to plot the graph of the error variation with the number of measurements, or plot the histogram or scatter diagram of the error to observe the error distribution; Calculate the autocorrelation coefficient: The autocorrelation coefficient used to reflect the correlation of errors between different time points or measurement times: ； Among them, represents Δx i and covariance, represents standard deviation, and a represents the lag order.

9. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, as claimed in claim 8, wherein: Systematic error judgment includes: If the mean error or is significantly non-zero, and the error distribution plot shows an obvious trend or periodic change in the error, or the autocorrelation coefficient is significantly non-zero at a certain lag order a, it indicates the existence of systematic error.

10. A method for compensating errors by infinitely expanding the field of view of a camera of a chip mounter, as claimed in claim 8, wherein: Random error judgment includes: If the error mean or is close to zero, and the error distribution graph shows that the errors are randomly distributed without obvious trends or periodic changes, and at the same time the autocorrelation coefficient is close to zero at all lag orders a, it indicates the existence of random errors.

Images