A vision detection method, system and application compatible with multiple models of workpieces to be measured

Abstract

The present invention discloses a vision detection method compatible with multiple models of workpieces to be measured, which obtains reference data and robot measurement trajectories: during measurement, a vision positioning sensor collects Feature I and calculates the measured coordinates; if the workpiece to be measured is a new model: the RT is obtained by using the theoretical coordinates and the measured coordinates of Feature I 移动 ; RT is selected from the reference database 定位 and RT 基准 , and R is calculated 实际 ; through the positioning of RT 实际 , the measured point coordinates are converted to the coordinate system of the workpiece to be measured; and a new set of reference data is added; if the workpiece to be measured is an old model, the reference coordinates, reference transformation matrix and robot measurement trajectories of this model are found; the position offset matrix is calculated by using the measured coordinates and the reference coordinates to correct the robot measurement trajectories, and the current workpiece to be measured is positioned by using the reference transformation matrix, and the measured point coordinates are converted to the coordinate system of the workpiece to be measured; the method of the present invention not only ensures the positioning accuracy but also improves the positioning speed. Compared with the existing methods, the detection time is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the field of visual positioning, and particularly to a visual detection method, system and application thereof that are compatible with multiple models of workpieces to be measured. Background Art

[0002] At present, visual detection solutions have been widely used in the processing and manufacturing industries, and are characterized by high precision and high automation. The visual detection solution aims to install "eyes" at the end of the robot, and uses the visual detection sensor installed at the end of the robot to obtain the features on the workpiece to be measured, so as to realize the intelligent monitoring of the workpiece. In high-precision manufacturing fields such as automobiles, ships, and aerospace, in order to ensure the detection accuracy, it is necessary to ensure the relative position between the workpiece to be measured and the visual detection sensor is fixed. To achieve this goal, the existing technical solutions can be divided into the following two types:

[0003] Solution 1: Install a high-precision positioning tooling in the detection station to fix the workpiece to be measured. Rely on the machining accuracy of the positioning tooling to ensure that the positions of different workpieces to be measured do not shift after being placed; for example: multiple positioning pins will be set on the high-precision positioning tooling. Using the positioning holes of the workpiece to be measured as the positioning reference, when in use, insert them into the positioning holes of the workpiece to be measured, and the pin-hole fit restricts the degrees of freedom of the workpiece to be measured, and controls the six degrees of freedom of the linear motion of the workpiece to be measured in the X, Y, and Z axes and the rotational motion around X, Y, and Z.

[0004] This solution has the following problems: high cost. For multiple models of workpieces to be measured, it is necessary to match and machine multiple positioning toolings, which requires huge investment, especially for positioning toolings suitable for large workpieces (such as: positioning toolings for vehicle bodies, aircraft fuselage skeletons, aircraft fuselage skins, etc.).

[0005] Solution 2: Add accessories to the positioning tooling. Refer to the method for accurately positioning a workpiece to be measured using a three-dimensional measurement system in patent document CN 109443273 B. It realizes the accurate positioning of the workpiece to be measured by calculating the conversion relationship between the coordinate systems of the workpiece to be measured before and after being installed on the fixture (positioning tooling) and the theoretical coordinate system. This solution solves the technical problem of the positioning deviation of multiple workpieces of the same model.

[0006] This solution has the following problems. After each workpiece to be measured is placed, it is necessary to measure the accessories using a three-dimensional measurement system (laser tracker, coordinate measuring machine, three-dimensional scanner), and the operation is complex and time-consuming.

[0007] It should be emphasized that neither Solution 1 nor Solution 2 involves the improvement of the visual positioning scheme for multi-model workpieces. In the prior art, when adding a new model of workpiece to be measured, it is necessary to use standard instruments (such as laser trackers, coordinate measuring machines, etc.) to collect the features on the new model of workpiece to be measured, establish the conversion relationship between the coordinate system of the workpiece to be measured and the coordinate system of the standard instrument, and then use the standard instrument to calibrate the robot to establish the conversion relationship between the robot base coordinate system and the standard instrument coordinate system. Through the transfer of the standard instrument, the conversion relationship between the coordinate system of the workpiece to be measured and the robot base coordinate system is established to achieve the positioning of the new model of workpiece. This method is cumbersome, requires the installation of features (such as tracker target balls) at the end of the robot, and requires the robot to change multiple poses. The standard instrument collects feature information and constructs the conversion relationship, which usually takes more than 6 hours, with a long time consumption and occupying the production rhythm. Summary of the Invention

[0008] To solve the above problems, the present invention provides a visual detection method, system and its application compatible with multi-model workpieces to be measured. Different positioning mechanisms are designed for new and old workpieces to be measured. The corresponding conversion relationship is calculated by collecting Feature I through a visual positioning sensor, which not only ensures the positioning accuracy but also improves the positioning speed. For new models of workpieces to be measured, there is no need for standard instruments to participate in the positioning process, nor is it necessary to install expensive positioning tooling. Compared with the existing methods, the detection time is greatly reduced.

[0009] The technical solution of the present invention is as follows:

[0010] A visual detection method compatible with multi-model workpieces to be measured, wherein the workpieces to be measured are placed in a detection area, and a robot with a visual detection sensor installed at the end and a visual positioning sensor are fixedly arranged around the detection area; the visual detection sensor is used to collect images of the workpieces to be measured; the visual positioning sensor is used to collect Feature I on the workpieces to be measured; Feature I is an inherent feature hole, feature ball or feature point on the workpiece to be measured; there are multiple Feature Is.

[0011] The following process is pre-performed to obtain reference data and the robot measurement trajectory:

[0012] Select any one of the multi-model workpieces to be measured as a reference workpiece, place it in the detection area, the visual positioning sensor collects Feature I on the reference workpiece, and calculates the coordinates of each Feature I in the coordinate system of the visual positioning sensor, denoted as reference coordinates; the conversion matrix between the robot base coordinate system and the reference workpiece coordinate system calibrated by using a standard instrument is denoted as the reference conversion matrix; combining the relationship between the calibrated robot base coordinate system and the visual positioning sensor, the conversion matrix between the visual positioning sensor coordinate system and the reference workpiece coordinate system is obtained, denoted as the positioning matrix.

[0013] Store the model number, reference coordinates, reference transformation matrix, and positioning matrix of the reference workpiece in the reference database as a set of reference data;

[0014] Based on the detection requirements of the vision detection sensor, determine the robot trajectory during the detection of workpieces to be measured of each model, obtain the transformation relationship between the vision detection sensor coordinate system and the robot base coordinate system at each image acquisition position, and store it in association with the model number of the workpiece to be measured and the robot trajectory;

[0015] During measurement, use the following steps to perform vision detection on the workpiece to be measured:

[0016] Obtain the model number of the current workpiece to be measured; place the workpiece to be measured in the detection area, use the vision positioning sensor to collect the feature I on the workpiece to be measured, and calculate the coordinates of each feature I in the vision positioning sensor coordinate system, denoted as the measured coordinates;

[0017] Determine whether the current workpiece to be measured and the previous workpiece to be measured are of the same model; if not, process using method A, if so, process using method B:

[0018] Method A:

[0019] Denote the coordinates of the feature I on the current workpiece to be measured in the workpiece theoretical digital mock-up coordinate system as the theoretical coordinates; use the theoretical coordinates and the measured coordinates to obtain the transformation matrix between the workpiece theoretical digital mock-up coordinate system and the vision positioning sensor coordinate system, denoted as RT 移动 ; Select the positioning matrix and the reference transformation matrix corresponding to the workpiece to be measured of any model from the reference database, denoted as RT 定位 and RT 基准 respectively, and use the following formula to obtain the transformation matrix R between the current workpiece coordinate system and the robot base coordinate system 实际 =RT 移动 ×(RT 定位 ) -1 ×RT 基准 ;

[0020] Retrieve the corresponding robot trajectory according to the model number of the current workpiece to be measured; along the robot trajectory, the vision detection sensor performs image acquisition on the workpiece to be measured at the preset image acquisition position, and locates the current workpiece to be measured through RT 实际 and convert the measured point coordinates in the acquired workpiece image to the current workpiece coordinate system;

[0021] Meanwhile, denote the current measured coordinates as the reference coordinates; denote RT 移动 and RT 实际 as the positioning matrix and the reference transformation matrix of the current workpiece to be measured respectively;

[0022] Add the model of the current workpiece to be measured, its corresponding reference coordinates, positioning matrix, reference transformation matrix, and the taught robot measurement trajectory as a set of reference data and store them in the reference database;

[0023] Method B:

[0024] Based on the model of the current workpiece to be measured, search in the reference database for the reference coordinates, reference transformation matrix, and robot measurement trajectory corresponding to this model;

[0025] Calculate the position offset matrix of the workpiece to be measured using the measured coordinates and the reference coordinates, send it to the robot controller to correct the robot measurement trajectory, and the vision detection sensor uses the corrected robot measurement trajectory to collect images of the workpiece to be measured;

[0026] Use the found reference transformation matrix to position the current workpiece to be measured, and convert the measurement point coordinates in the workpiece image to the coordinate system of the current workpiece to be measured.

[0027] Furthermore, use the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system calibrated by a standard instrument, denoted as the reference transformation relationship; the calibration process is as follows:

[0028] Use a standard instrument to collect multiple Features II on the reference workpiece, establish a reference workpiece coordinate system, and obtain the transformation matrix between the reference workpiece coordinate system and the standard instrument coordinate system;

[0029] Use a standard instrument to collect multiple Features II at the end of the robot in each posture, and combine the robot kinematic model to calculate the transformation matrix between the robot base coordinate system and the standard instrument coordinate system;

[0030] The Feature II is a feature hole, a feature sphere, or a feature point;

[0031] Combine the above two transformation matrices to obtain the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system, denoted as the reference transformation relationship.

[0032] Preferably, the standard instrument includes a coordinate measuring machine, a laser tracker, a photogrammetry system, a total station, a theodolite, or an articulated arm measuring machine as the standard measuring instrument.

[0033] Preferably, both Feature I and Feature II include at least 4 and are not coplanar.

[0034] Preferably, when the workpiece to be measured is large, there are multiple robots and vision detection sensors. First, use a standard instrument to calibrate the relative position relationship between each robot, and then arbitrarily select one robot to calculate the relationship between the robot base coordinate and the coordinate system of the workpiece to be measured.

[0035] Preferably, there are 2 to 8 vision positioning sensors.

[0036] The present invention also discloses a vision detection system based on the above-mentioned vision detection method for compatible multi-model workpieces to be measured, including: a reference workpiece processing module, a reference data storage module, a model judgment module, a new model positioning module, an old model positioning module, and a detection information storage module;

[0037] The reference data storage module is respectively connected to the reference workpiece processing module, the new model positioning module, and the old model positioning module, and is used to store reference data and robot trajectory information;

[0038] The reference workpiece processing module is used to obtain the reference coordinates, reference transformation matrix, and positioning matrix of the reference workpiece; and based on the detection requirements of the vision detection sensor, determine the robot trajectory during the detection of workpieces to be measured of each model;

[0039] The model judgment module is respectively connected to the new model positioning module and the old model positioning module, and is used to judge whether the current workpiece to be measured is of the same model as the previous workpiece to be measured, and classify it into the corresponding module (connected to the new model positioning module or the old model positioning module) for processing;

[0040] The new model positioning module is used to obtain the reference transformation matrix and reference coordinates corresponding to the new model workpiece to be measured, and store them together with the model in the reference data storage module; at the same time, use the newly obtained reference transformation matrix to position the current workpiece to be measured, and convert the coordinate points in the workpiece image to the current workpiece coordinate system, and send them to the detection information storage module;

[0041] The old model positioning module is used to obtain the offset matrix corresponding to the old model workpiece to be measured and correct the robot measurement trajectory; at the same time, find the reference data with the same model in the reference data storage module, use the reference transformation matrix in the reference data to position the current workpiece to be measured, and convert the coordinate points in the workpiece image to the current workpiece coordinate system, and send them to the detection information storage module;

[0042] The detection information storage module is respectively connected to the new model positioning module and the old model positioning module, and is used to store the coordinates converted to the current workpiece coordinate system.

[0043] The present invention also discloses a method for accuracy evaluation using the above-mentioned vision detection method for compatible multi-model workpieces to be measured, including: recording the coordinates of the measurement points in the workpiece image after being converted to the current workpiece coordinate system as the measurement coordinates, subtracting the measurement coordinates from their coordinates in the theoretical digital model coordinate system of the workpiece, and judging whether the difference is within a preset interval. If so, the machining accuracy of the workpiece to be measured meets the requirements; if not, the machining accuracy of the workpiece to be measured does not meet the requirements.

[0044] Preferably, the method for obtaining the model of the current workpiece to be measured is:

[0045] Set in the detection area: a workpiece model radio frequency reader or a workpiece model image acquisition sensor, and identify the model of the current workpiece to be measured according to the radio frequency signal or the model image.

[0046] The solution of the present invention has the following advantages:

[0047] During actual detection, the reference data can be directly called without occupying the beat.

[0048] Fix a vision positioning sensor around the detection area, rely on it to collect Feature I (feature holes, balls, points), obtain the conversion relationship using rigid body transformation according to the coordinates of Feature I, and design different positioning mechanisms for new and old workpieces to be measured, which not only ensures the positioning accuracy but also improves the positioning speed, and the detection process is fast and effective.

[0049] It should be noted that in this solution, for the positioning process of new model workpieces to be measured, there is no need for standard instruments (tracking instruments, coordinate measuring machines, etc.) to participate anymore, but it is obtained by relying on coordinate system conversion, which takes a short time. It only takes 30 minutes to position new model workpieces. Compared with the existing methods, the positioning and detection time is greatly reduced.

[0050] In addition, in this solution, the workpiece to be measured only needs to be placed in the detection area, and there is no need to install it on a complex and expensive positioning tooling / bracket, which not only saves the processing cost but also saves the installation time. Description of the Drawings

[0051] Figure 1 is the system structure block diagram of the invention;

[0052] Figure 2 is the installation layout schematic diagram of the vision positioning sensor and the robot in the specific embodiment. Specific Embodiment

[0053] The technical solution of the present invention will be described in detail below in conjunction with the drawings and specific embodiments.

[0054] A vision detection method compatible with multiple models of workpieces to be measured, the workpiece to be measured is placed in the detection area, and a robot with a vision detection sensor installed at the end and a vision positioning sensor are fixedly arranged around the detection area; the vision detection sensor is used to collect images of the workpiece to be measured; the vision positioning sensor is used to collect Feature I on the workpiece to be measured; Feature I is an inherent feature hole, feature ball or feature point on the workpiece to be measured; there are multiple Feature Is;

[0055] The following process is carried out in advance to obtain the reference data and the robot measurement trajectory:

[0056] Select any one of the workpieces to be measured with multiple models as the reference workpiece, place it in the detection area, and the vision positioning sensor collects the feature I on the reference workpiece, and calculates the coordinates of each feature I in the coordinate system of the vision positioning sensor, which is recorded as the reference coordinate; use the conversion matrix between the robot base coordinate system and the reference workpiece coordinate system calibrated by the standard instrument, which is recorded as the reference conversion matrix; combine the relationship between the calibrated robot base coordinate system and the vision positioning sensor to obtain the conversion matrix between the vision positioning sensor coordinate system and the reference workpiece coordinate system, which is recorded as the positioning matrix;

[0057] Store the model, reference coordinates, reference conversion matrix and positioning matrix of the reference workpiece in the reference database as a set of reference data;

[0058] Based on the detection requirements of the vision detection sensor, determine the robot trajectory during the detection of the workpieces to be measured with each model, obtain the conversion relationship between the vision detection sensor coordinate system and the robot base coordinate system at each image acquisition position, and store it in association with the model of the workpiece to be measured and the robot trajectory;

[0059] During measurement, use the following steps to perform vision detection on the workpiece to be measured:

[0060] Obtain the model of the current workpiece to be measured; in this embodiment, the method for obtaining the model of the current workpiece to be measured is:

[0061] Set a workpiece model radio frequency reader or a workpiece model image acquisition sensor in the detection station, and identify the model of the current workpiece to be measured according to the radio frequency signal or the image.

[0062] Place the workpiece to be measured in the detection area, use the vision positioning sensor to collect the feature I on the workpiece to be measured, and calculate the coordinates of each feature I in the coordinate system of the vision positioning sensor, which is recorded as the measured coordinate;

[0063] Judge whether the current workpiece to be measured and the previous workpiece to be measured are of the same model; if not, the current workpiece to be measured is a new model and is processed by method A, if so, the current workpiece to be measured is an old model and is processed by method B:

[0064] Method A:

[0065] Record the coordinates of the feature I on the current workpiece to be measured in the workpiece theoretical digital model coordinate system as the theoretical coordinate; use the theoretical coordinate and the measured coordinate to obtain the conversion matrix between the workpiece theoretical digital model coordinate system and the vision positioning sensor coordinate system, which is recorded as RT 移动 ; Select the positioning matrix and the reference conversion matrix corresponding to the workpiece to be measured of any model from the reference database, which are respectively recorded as RT 定位 and RT 基准 Use the following formula to obtain the conversion matrix R between the current workpiece coordinate system and the robot base coordinate system实际 = RT 移动 × (RT 定位 ) -1 × RT 基准 ;

[0066] Retrieve the corresponding robot trajectory according to the model of the workpiece to be measured currently; along the robot trajectory, the vision detection sensor collects an image of the workpiece to be measured at a preset image acquisition position, and locates the current workpiece to be measured through RT 实际 Convert the measured point coordinates in the collected workpiece image to the current workpiece coordinate system;

[0067] Meanwhile, record the current measured coordinates as the reference coordinates; record RT 移动 、RT 实际 as the positioning matrix and the reference transformation matrix of the current workpiece to be measured respectively;

[0068] Add the model of the current workpiece to be measured and its corresponding reference coordinates, positioning matrix, reference transformation matrix, and the taught robot measurement trajectory as a set of reference data and store it in the reference database;

[0069] Method B:

[0070] Based on the model of the current workpiece to be measured, search in the reference database for the corresponding reference coordinates, reference transformation matrix, and robot measurement trajectory of this model;

[0071] Calculate the position offset matrix of the workpiece to be measured using the measured coordinates and the reference coordinates, send it to the robot controller to correct the robot measurement trajectory, and the vision detection sensor uses the corrected robot measurement trajectory to collect the workpiece image;

[0072] Locate the current workpiece to be measured using the found reference transformation matrix, and convert the measured point coordinates in the workpiece image to the current workpiece coordinate system.

[0073] Among them, the standard instrument includes a standard measuring instrument such as a coordinate measuring machine, a laser tracker, a photogrammetry system, a total station, a theodolite, or an articulated arm measuring machine.

[0074] When the workpiece to be measured is large, there are multiple robots and vision detection sensors. First, use a standard instrument to calibrate the relative position relationship between each robot, and then arbitrarily select a robot to calculate the relationship between the robot base coordinates and the workpiece coordinate system.

[0075] In order to collect the workpiece to be measured more comprehensively and ensure that the Feature I is not less than 4, preferably, 2 to 8 visual positioning sensors are provided. In practical applications, for workpieces of the same category but different models, the appearance differences are often less than 30%, such as different styles of car models. At this time, 2 to 8 visual positioning sensors can meet the positioning requirements.

[0076] When the differences between different models of workpieces to be measured are large, more positioning sensors can be set to collect Feature I at various angles and positions of the workpiece.

[0077] As shown in the figure, a station layout of one of the detection areas is shown: 4 robots and visual detection sensors are arranged on both sides of the workpiece to be measured, and 4 visual positionings are correspondingly arranged at the four corners of the detection area.

[0078] Such as Figure 1 As shown, based on the above method, this embodiment provides a visual detection system compatible with multiple models of workpieces to be measured, including: a reference workpiece processing module, a reference data storage module, a model judgment module, a new model positioning module, an old model positioning module, and a detection information storage module;

[0079] The reference data storage module is respectively connected to the reference workpiece processing module, the new model positioning module, and the old model positioning module, and is used to store reference data and robot trajectory information;

[0080] The reference workpiece processing module is used to obtain the reference coordinates, reference transformation matrix, and positioning matrix of the reference workpiece; and based on the detection requirements of the visual detection sensor, determine the robot trajectory during the detection of workpieces of each model;

[0081] The model judgment module is respectively connected to the new model positioning module and the old model positioning module, and is used to judge whether the current workpiece to be measured is of the same model as the previous workpiece to be measured, and classify it into the corresponding module (connected to the new model positioning module or the old model positioning module) for processing;

[0082] The new model positioning module is used to obtain the reference transformation matrix and reference coordinates corresponding to the new model workpiece to be measured, and store them together with the model in the reference data storage module; at the same time, use the newly obtained reference transformation matrix to position the current workpiece to be measured, and convert the coordinate points in the workpiece image to the current workpiece coordinate system, and send them to the detection information storage module;

[0083] The old model positioning module is used to obtain the offset matrix corresponding to the old model workpiece to be measured and correct the robot measurement trajectory; at the same time, find the reference data with the same model in the reference data storage module, use the reference transformation matrix in the reference data to position the current workpiece to be measured, and convert the coordinate points in the workpiece image to the current workpiece coordinate system, and send them to the detection information storage module;

[0084] The detection information storage module is respectively connected to the new model positioning module and the old model positioning module, and is used to store the coordinates converted to the current workpiece coordinate system to be measured.

[0085] The present invention also discloses a method for precision evaluation using a vision detection method compatible with multiple models of workpieces to be measured, including: recording the coordinates of the measurement points in the workpiece image after being converted to the current workpiece coordinate system to be measured as the measurement coordinates, taking the difference between the measurement coordinates and their coordinates in the theoretical digital model coordinate system of the workpiece, and judging whether the difference is within a preset interval. If so, the machining precision of the workpiece to be measured meets the requirements; if not, the machining precision of the workpiece to be measured does not meet the requirements.

[0086] The following takes the detection of automobile bodies of different models as an example for exemplary elaboration:

[0087] A vision detection method compatible with multiple models of automobile bodies, the automobile body is placed in the detection area, as Figure 2 shown, there are 4 robots with vision detection sensors installed at the end and 4 vision positioning sensors fixedly arranged around the detection area; the vision detection sensors are used to collect the images of the automobile body; the vision positioning sensors are respectively in four directions and are used to collect the feature I on the vehicle body; the feature I is an inherent feature hole on the vehicle body (such as the hole on the wheel hub); there are multiple feature Is;

[0088] Taking the laser tracker as an example of a standard instrument, the conversion relationship between each robot base coordinate system and the tracker coordinate system is calibrated in advance by the laser tracker, and then the conversion relationship between each robot base coordinate is established through the tracker coordinate system.

[0089] The following process is carried out in advance to obtain the reference data and the robot measurement trajectory:

[0090] Select any one of the multiple models of automobile bodies as the reference workpiece, place it in the detection area, the vision positioning sensor collects the feature I on the reference workpiece, and calculates the coordinates of each feature I in the vision positioning sensor coordinate system, which is recorded as the reference coordinates; the conversion matrix between the robot base coordinate system and the reference workpiece coordinate system calibrated by the laser tracker is recorded as the reference conversion matrix; combining the relationship between the calibrated robot base coordinate system and the vision positioning sensor (hand-eye calibration combined with the robot motion model), the conversion matrix between the vision positioning sensor coordinate system and the reference workpiece coordinate system is obtained, which is recorded as the positioning matrix;

[0091] The model, reference coordinates, reference conversion matrix and positioning matrix of the reference workpiece are stored in the reference database as a set of reference data;

[0092] During specific implementation, the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system obtained by calibration using a laser tracker is denoted as the reference transformation relationship. The calibration process is as follows:

[0093] Use a laser tracker to collect multiple features II on the reference workpiece, establish a reference workpiece coordinate system, and obtain the transformation matrix between the reference workpiece coordinate system and the laser tracker coordinate system;

[0094] Use a laser tracker to collect multiple features II at the end of the robot in each posture, and combine the robot kinematic model to calculate the transformation matrix between the robot base coordinate system and the laser tracker coordinate system;

[0095] Feature II is a feature hole, a feature sphere or a feature point; such as the tracker target ball installed at the end of the robot;

[0096] Combine the above two transformation matrices to obtain the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system, denoted as the reference transformation relationship.

[0097] Among them, in order to accurately calculate the transformation matrix, in this embodiment, both feature I and feature II include at least 4 and are not coplanar.

[0098] Next, based on the detection requirements of the vision detection sensor, determine the robot trajectory during the detection of each model of automobile body, obtain the transformation relationship between the vision detection sensor coordinate system and the robot base coordinate system at each image acquisition position, and store it in association with the automobile body model and the robot trajectory;

[0099] The above process is an offline process, which is carried out before the formal online detection and only needs to be carried out once.

[0100] During measurement, use the following steps to perform vision detection on the automobile body:

[0101] Set a workpiece model radio frequency reader in the detection area, and obtain the model of the current vehicle body according to the radio frequency signal;

[0102] Place the vehicle body in the detection area, use the vision positioning sensor to collect feature I on the vehicle body, and calculate the coordinates of each feature I in the vision positioning sensor coordinate system, denoted as the measured coordinates;

[0103] Judge whether the current vehicle body and the previous vehicle body are of the same model; if not, the current vehicle body is a new model and is processed by method A, if so, the current vehicle body is an old model and is processed by method B:

[0104] Method A:

[0105] Record the coordinates of feature I on the current vehicle body in the workpiece theoretical digital model coordinate system as the theoretical coordinates; use the theoretical coordinates and the measured coordinates to obtain the transformation matrix between the vehicle body theoretical digital model coordinate system and the vision positioning sensor coordinate system, denoted as RT 移动 ; Select the positioning matrix and the reference transformation matrix corresponding to the vehicle body of any model from the reference database, denoted as RT 定位 and RT 基准 respectively; use the following formula to obtain the transformation matrix R between the current vehicle body coordinate system and the robot base coordinate system 实际 = RT 移动 ×(RT 定位 ) -1 ×RT 基准 ;

[0106] According to the model of the current vehicle body, retrieve the corresponding robot trajectory; along the robot trajectory, the vision detection sensor collects images of the vehicle body at the preset image acquisition positions, and locates the current vehicle body through RT 实际 , and convert the measured point coordinates in the collected vehicle body image to the current vehicle body coordinate system;

[0107] Meanwhile, record the current measured coordinates as the reference coordinates; record RT 移动 , RT 实际 as the positioning matrix and the reference transformation matrix of the current vehicle body respectively;

[0108] Add the model of the current vehicle body and its corresponding reference coordinates, positioning matrix, reference transformation matrix, and the taught robot measurement trajectory as a set of reference data and store them in the reference database;

[0109] If there are 3 models of vehicles to be detected, the first model (model 1) is the same as the reference workpiece, and it can be directly positioned and detected using the reference data. Model 2 is a new model that appears for the first time, and model 3 is a new model that appears for the second time. The calculation of the new model R 实际 is as follows:

[0110] Model 2: R 车型2实际 = RT 车型2移动 ×(RT 车型1定位 ) -1 ×RT 车型1基准

[0111] When adding new reference data, perform the following data update: record RT 车型2移动 , RT 车型2实际 as the positioning matrix RT 车型2定位 and the reference transformation matrix RT 车型2基准 of model 2 respectively;

[0112] Model 3: When selecting the positioning matrix and the reference transformation matrix corresponding to the vehicle body of model 2 from the reference database:

[0113] R 车型3实际 = RT 车型3移动 × (RT 车型2定位 ) -1 × RT 车型2基准

[0114] When selecting the positioning matrix and the reference transformation matrix corresponding to the body of vehicle model 1 from the reference database:

[0115] R 车型3实际 = RT 车型3移动 × (RT 车型1定位 ) -1 × RT 车型1基准

[0116] For the sake of easy understanding, the above formula annotates the vehicle models in the following table. Among them, R 车型2实际 represents R of vehicle model 2 实际 , RT 车型2移动 represents RT of vehicle model 2 移动 , RT 车型1定位 represents RT of vehicle model 1 定位 , RT 车型1基准 represents RT of vehicle model 1 基准 , R 车型3实际 represents RT of vehicle model 3 实际 , RT 车型3移动 represents RT of vehicle model 3 移动 .

[0117] Method B:

[0118] Based on the model of the current body, search in the reference database for the reference coordinates, the reference transformation matrix, and the robot measurement trajectory corresponding to this model;

[0119] Use the measured coordinates and the reference coordinates to calculate the position offset matrix of the body, send it to the robot controller to correct the robot measurement trajectory, and the vision detection sensor uses the corrected robot measurement trajectory to collect the body image;

[0120] Use the found reference transformation matrix to position the current body and convert the measured point coordinates in the body image to the current body coordinate system.

[0121] To evaluate the body processing accuracy, record the coordinates obtained by converting the measured point coordinates in the image of the workpiece to be measured to the current workpiece coordinate system to be the measurement coordinates, calculate the difference between the measurement coordinates and their coordinates in the theoretical digital model coordinate system of the workpiece, and determine whether the difference is within the preset interval. If so, the processing accuracy of the workpiece to be measured meets the requirements; if not, the processing accuracy of the workpiece to be measured does not meet the requirements.

[0122] The method of the present invention designs different positioning mechanisms for new and old workpieces to be measured, which not only ensures the positioning accuracy but also improves the positioning speed. Compared with the existing methods, the detection time is greatly reduced.

[0123] The foregoing description of the specific exemplary embodiments of the present invention has been presented for purposes of illustration and description. The foregoing description is not intended to be exhaustive or to limit the invention to the precise forms disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain the particular principles of the invention and its practical application to enable others skilled in the art to make and utilize the various exemplary embodiments of the invention and their various alternatives and modifications. The scope of the invention is intended to be defined by the appended claims and their equivalents.

Claims

1. A visual inspection method compatible with multiple models of workpieces to be measured. The workpiece to be measured is placed in the inspection area, and a robot with a visual inspection sensor mounted at its end and a visual positioning sensor are fixedly arranged around the inspection area. The visual inspection sensor is used to collect images of the workpiece to be measured. The visual positioning sensor is used to collect Feature I on the workpiece to be measured. Feature I is an inherent feature hole, feature ball or feature point on the workpiece to be measured. There are multiple Feature Is. It is characterized in that: The following processes are pre - carried out to obtain reference data and the robot measurement trajectory: Select any one of the multiple models of workpieces to be measured as the reference workpiece, place it in the inspection area, the visual positioning sensor collects Feature I on the reference workpiece, and calculates the coordinates of each Feature I in the coordinate system of the visual positioning sensor, denoted as reference coordinates. Use the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system calibrated by a standard instrument, denoted as the reference transformation matrix. Combined with the relationship between the calibrated robot base coordinate system and the visual positioning sensor, obtain the transformation matrix between the visual positioning sensor coordinate system and the reference workpiece coordinate system, denoted as the positioning matrix. Store the model of the reference workpiece, the reference coordinates, the reference transformation matrix and the positioning matrix in the reference database as a set of reference data. Based on the detection requirements of the visual inspection sensor, determine the robot trajectory during the inspection of each model of workpiece to be measured, obtain the transformation relationship between the visual inspection sensor coordinate system and the robot base coordinate system at each image - acquisition position, and store it in association with the model of the workpiece to be measured and the robot trajectory. During measurement, the following steps are used to perform visual inspection on the workpiece to be measured: Obtain the model of the current workpiece to be measured. The workpiece to be measured is placed in the inspection area, and the visual positioning sensor is used to collect Feature I on the workpiece to be measured, and calculate the coordinates of each Feature I in the coordinate system of the visual positioning sensor, denoted as the measured coordinates. Judge whether the current workpiece to be measured and the previous workpiece to be measured are of the same model. If not, process it using Method A. If so, process it using Method B: Method A: Record the coordinates of Feature I on the current workpiece to be measured in the coordinate system of the workpiece's theoretical digital model as the theoretical coordinates; use the theoretical coordinates and the measured coordinates to obtain the transformation matrix between the coordinate system of the workpiece's theoretical digital model and the coordinate system of the vision positioning sensor, denoted as RT 移动 ; Select the positioning matrix and the reference transformation matrix corresponding to any model of the workpiece to be measured from the reference database, denoted as RT 定位 and RT 基准 respectively. Use the following formula to obtain the transformation matrix R between the coordinate system of the current workpiece to be measured and the coordinate system of the robot base 实际 = RT 移动 ×(RT 定位 ) -1 ×RT 基准 ; Retrieve the corresponding robot trajectory according to the model of the current workpiece to be measured; along the robot trajectory, the vision detection sensor collects an image of the workpiece to be measured at a preset image acquisition position, and locates the current workpiece to be measured through RT 实际 Convert the measured point coordinates in the collected workpiece image to the current workpiece coordinate system Meanwhile, record the current measured coordinates as the reference coordinates; record RT 移动、 RT 实际 as the positioning matrix and the reference transformation matrix of the current workpiece to be measured, respectively; Add the model of the current workpiece to be measured, its corresponding reference coordinates, positioning matrix, reference transformation matrix and the taught robot measurement trajectory as a new set of reference data and store it in the reference database. Method B: Based on the model of the current workpiece to be measured, search in the reference database for the corresponding reference coordinates, reference transformation matrix and robot measurement trajectory of this model. Use the measured coordinates and the reference coordinates to calculate the position offset matrix of the workpiece to be measured, send it to the robot controller to correct the robot measurement trajectory, and the visual inspection sensor uses the corrected robot measurement trajectory to collect images of the workpiece to be measured. Use the found reference transformation matrix to position the current workpiece to be measured, and convert the measured point coordinates in the workpiece image to the coordinate system of the current workpiece to be measured.

2. The visual inspection method for compatible with multiple models of workpieces to be measured according to claim 1, characterized in that, Use the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system calibrated by a standard instrument, denoted as the reference transformation relationship. The calibration process is as follows: Use a standard instrument to collect multiple Feature IIs on the reference workpiece, establish the reference workpiece coordinate system, and obtain the transformation matrix between the reference workpiece coordinate system and the standard instrument coordinate system. Collect multiple features II at the end of the robot in each posture by using standard instruments, and calculate the transformation matrix between the robot base coordinate system and the standard instrument coordinate system in combination with the robot kinematic model; The feature II is a feature hole, a feature sphere or a feature point; Combine the above two transformation matrices to obtain the transformation matrix between the robot base coordinate system and the reference workpiece coordinate system, which is denoted as the reference transformation relationship.

3. The visual inspection method for compatible multi-model workpieces to be measured according to claim 2, wherein The standard instrument includes a standard measuring instrument, which is a coordinate measuring machine, a laser tracker, a photogrammetry system, a total station, a theodolite or an articulated arm measuring machine.

4. The visual inspection method for compatible with multiple models of workpieces to be measured according to claim 2, characterized in that, Both the feature I and the feature II include at least 4 and are not coplanar.

5. The visual inspection method for compatible multi-model workpieces to be measured according to claim 1 or 2, characterized in that When the size of the workpiece to be measured is large, there are multiple robots and vision detection sensors. First, use a standard instrument to calibrate the relative position relationship between each robot, and then select any one robot to calculate the relationship between the robot base coordinate and the workpiece coordinate system to be measured.

6. The visual inspection method for compatible multi-model workpieces to be measured according to claim 1, characterized in that, There are 2 to 8 visual positioning sensors.

7. A vision inspection system for compatible multi-model workpieces to be measured based on the vision inspection method for compatible multi-model workpieces to be measured according to claim 1, characterized in that, It includes: A reference workpiece processing module, a reference data storage module, a model judgment module, a new model positioning module, an old model positioning module and a detection information storage module; The reference data storage module is respectively connected to the reference workpiece processing module, the new model positioning module and the old model positioning module, and is used to store reference data and robot trajectory information; The reference workpiece processing module is used to obtain the reference coordinates, the reference transformation matrix and the positioning matrix of the reference workpiece; and based on the detection requirements of the visual detection sensor, determine the robot trajectory during the detection of the workpiece to be measured of each model; The model judgment module is respectively connected to the new model positioning module and the old model positioning module, and is used to judge whether the current workpiece to be measured is of the same model as the previous workpiece to be measured, and classify it into the corresponding new model positioning module or old model positioning module for processing; The new model positioning module is used to obtain the reference transformation matrix and reference coordinates corresponding to the new model workpiece to be measured, and store them together with the model in the reference data storage module; at the same time, use the newly obtained reference transformation matrix to position the current workpiece to be measured, and convert the coordinate points in the image of the workpiece to be measured to the current workpiece coordinate system, and send them to the detection information storage module; The old model positioning module is used to obtain the offset matrix corresponding to the old model workpiece to be measured and correct the robot measurement trajectory; at the same time, find the reference data with the same model in the reference data storage module, use the reference transformation matrix in the reference data to position the current workpiece to be measured, and convert the coordinate points in the image of the workpiece to be measured to the current workpiece coordinate system, and send them to the detection information storage module; The detection information storage module is respectively connected to the new model positioning module and the old model positioning module, and is used to store the coordinates converted to the current workpiece coordinate system.

8. A method for accuracy evaluation using the visual inspection method for compatible multi-model workpieces to be measured as described in claim 1, characterized in that, It includes: Record the coordinates of the measurement points in the image of the workpiece to be measured after being converted to the current workpiece coordinate system as the measurement coordinates, subtract the measurement coordinates from their coordinates in the workpiece theoretical digital model coordinate system, and judge whether the difference is within a preset interval. If so, the machining accuracy of the workpiece to be measured meets the requirements; if not, the machining accuracy of the workpiece to be measured does not meet the requirements.

9. The visual inspection method for compatible multi-model workpieces to be measured according to claim 1, characterized in that, The method for obtaining the model of the current workpiece to be measured is: Set in the detection area: a workpiece model radio frequency reader or a workpiece model image acquisition sensor, and identify the model of the current workpiece to be measured according to the radio frequency signal or the model image.

Images