CN115855953A - A three-dimensional defect detection method and device for an engine cylinder head - Google Patents

Abstract

The invention provides a three-dimensional defect detection method and a three-dimensional defect detection device for an engine cylinder cover, which can automatically detect surface defects and an internal three-dimensional structure of the engine cylinder cover and quickly and efficiently switch six surfaces of the engine cylinder cover. The method comprises the following steps: the method comprises the following steps: grabbing an engine cylinder cover to be detected by a first industrial robot and sending the engine cylinder cover to a specified detection area; step two: the method comprises the following steps that a first industrial robot is matched with a second industrial robot, and a three-dimensional camera on the second industrial robot is used for obtaining three-dimensional images of six surfaces 1,2, 3, 4, A and B on an engine cylinder to be detected; step three: carrying out point cloud registration on the obtained three-dimensional image of the engine cylinder cover; step four: carrying out three-dimensional modeling on the detected point cloud registration data and comparing the point cloud registration data with a three-dimensional model of a standard engine cylinder cover in shape difference; step five: and comparing the compared shape difference value with a set numerical value.

Description

A three-dimensional defect detection method and device for an engine cylinder head

technical field

The invention relates to the technical field of engine cylinder head detection, in particular to a three-dimensional defect detection method and device for an engine cylinder head.

Background technique

The casting of engine block and cylinder head is the most difficult and most important part in engine production, and its quality plays a decisive role in the performance of engine power and fuel consumption. It can be said that the casting technology of the cylinder block and cylinder head reflects the manufacturing capacity of the engine, and the complete set of casting equipment for the cylinder block and cylinder head of the automobile and its manufacturing level are one of the important factors affecting the performance of the engine. At present, the detection method in the industry is mainly based on manual visual detection. The detection mainly depends on the recognition ability, experience, responsibility and carefulness of the human eye. The detection reliability is closely related to people. In the actual production process, due to the high speed of the production line, people are prone to fatigue, and it is necessary to inspect all sides of the entire casting, so missed inspections often occur.

In the existing patent database, there is a kind of engine cylinder head visual detection equipment whose publication number is CN216433934U. This kind of equipment utilizes a three-degree-of-freedom motion platform, a profiling torque gripper and a transmission platform to form a transmission inspection system. The engine cylinder head can be displayed in the range of the optical lens from multiple angles; the transmission inspection system cooperates with the optical detection system to greatly improve the detection efficiency of the engine cylinder head, and the optical camera replaces the naked eye inspection to further ensure the accuracy of detection ; The device has strong controllability, greatly reduces the economic cost, and realizes the purpose of automatic intelligent detection. However, this type of defect detection equipment and method can only take a two-dimensional plane image of the engine cylinder head, and cannot detect the internal three-dimensional structure of the engine cylinder head, and the existing engine cylinder head defect detection device mainly uses a conveyor belt. Transportation, it is impossible to quickly switch between different surfaces for detection, and the efficiency is low.

Contents of the invention

In this regard, the present invention aims to provide a three-dimensional defect detection method and device for an engine cylinder head, which can automatically detect surface defects and internal three-dimensional structures of the engine cylinder head, and quickly and efficiently switch six faces of the engine cylinder head.

To achieve the above object, the present invention provides the following technical solutions:

A three-dimensional defect detection method of an engine cylinder head, comprising the following steps:

Step 1: Use No. 1 industrial robot to grab the cylinder head of the engine to be inspected and send it to the designated inspection area;

Step 2: Cooperate with the No. 1 industrial robot and the No. 2 industrial robot, and use the 3D camera on the No. 2 industrial robot to obtain three-dimensional images of six surfaces 1, 2, 3, 4, A, and B on the engine cylinder to be detected;

Step 3: Perform point cloud registration on the obtained 3D image of the engine cylinder head;

Step 4: Perform 3D modeling of the detected point cloud registration data and compare the shape difference with the 3D model of the standard engine cylinder head;

Step 5: Compare the compared shape difference with the set value, if it is greater than the set value range or less than the set value range, it will be judged as the defect position.

Compared with prior art, the beneficial effect of the present invention is:

Through the above detection method, the problem of the two-dimensional image defect on the surface of the engine cylinder head and its three-dimensional internal structure can be detected at the same time only by a single photo data of each surface, and it is automated, with high accuracy and extremely high Improve detection efficiency and save labor.

To achieve the above object, the present invention also provides the following technical solutions:

A device for three-dimensional defect detection of an engine cylinder head, including a light source, No. 1 industrial robot, No. 2 industrial robot, a recognition camera, a 3D camera and a main control module. The light source is fixed on the top of the detection area, and the recognition camera is installed on the No. 1 robot Above, the No. 2 industrial robot is fixed with a 3D camera, and the light source, No. 1 industrial robot, No. 2 industrial robot, identification camera, and 3D camera are electrically connected to the main control module respectively.

Preferably, before step 1, the identification camera on the No. 1 industrial robot is used to identify the grasping point of the B-side of the cylinder head of the engine to be detected.

Compared with prior art, the beneficial effect of the present invention is:

Through the continuous switching of the positions of the No. 1 industrial robot and the No. 2 industrial robot, the organic combination of various functional components is realized, so as to realize the six-sided multi-angle switching when grasping and detecting the engine cylinder head, and the recognition camera can detect the engine cylinder head. The high-precision recognition of the grabbing point can also solve the problem of low recognition of the grabbing point in an environment with different light source angles, with high precision, and the 3D image obtained by the 3D camera is stored in the main control module for storage And processing, finally by calculating the detected point cloud registration data for 3D modeling and comparing the shape difference with the 3D model of the standard engine cylinder head, so as to detect and locate the defect position of the engine cylinder head.

Description of drawings

Fig. 1 is a flow chart of the positioning of the grabbing point of the engine cylinder head in a three-dimensional defect detection method of the engine cylinder head and its device according to the present invention.

Fig. 2 is a flow chart of acquiring three-dimensional images of six faces of an engine cylinder head in the present invention. .

Fig. 3 is a flow chart of defect detection of three-dimensional data in the present invention.

Fig. 4 is a projection diagram along the radial direction obtained by radially scanning along the radial direction with the centroid obtained by K-means as the initial center.

Detailed ways

The present invention is described in further detail below:

Refer to Figure 1 to Figure 4 for details:

A three-dimensional defect detection method of an engine cylinder head, comprising the following steps:

Step 1: Use No. 1 industrial robot to grab the cylinder head of the engine to be inspected and send it to the designated inspection area;

Step 2: Cooperate with the No. 1 industrial robot and the No. 2 industrial robot, and use the 3D camera on the No. 2 industrial robot to obtain three-dimensional images of six surfaces 1, 2, 3, 4, A, and B on the engine cylinder to be detected;

Step 3: Perform point cloud registration on the obtained 3D image of the engine cylinder head;

Step 4: Perform 3D modeling of the detected point cloud registration data and compare the shape difference with the 3D model of the standard engine cylinder head;

Step 5: Compare the compared shape difference with the set value, if it is greater than the set value range or less than the set value range, it will be judged as the defect position. Through the above detection method, the problem of the two-dimensional image defect on the surface of the engine cylinder head and its three-dimensional internal structure can be detected at the same time only by a single photo data of each surface, and it is automated, with high accuracy and extremely high Improve detection efficiency and save labor.

In this embodiment, the four sides of the engine cylinder head are respectively 1, 2, 3, and 4, the front side is A side, and the back side is B side.

In this embodiment, before step 1, the identification camera on the No. 1 industrial robot is used to identify the grasping point on the B side of the cylinder head of the engine to be detected, so that the consistency of the grasping position of each cylinder head of the engine to be detected can be guaranteed. , so as to ensure the detection accuracy.

Before step 1, identify the grabbing point on the B side of the engine cylinder head to be detected by the recognition camera, and use the combination of convolutional neural network and K-means algorithm to accurately locate the grabbing point for installation. First, use the convolutional neural network to The center area of the installation grab point is roughly positioned, because the accuracy is not enough, it is necessary to accurately grab the position of the round hole in the rough positioning area; the first step is to use the K-means clustering method to obtain the contour of the hole circle of the installation grab point, and other Centroid, used to initialize the approximate location of the center of the circle. The initial center point is selected as the center point of the ROI (Intersection over Union) of the installation capture point hole circle bounding box output by the convolutional neural network. The steps of the K-means algorithm are as follows:

c ⁽ⁱ⁾ = argmin||x ⁽ⁱ⁾ - μ _j ||

1) Select the initialized 2 samples as the initial cluster centers as 1, 2,...,.

2) For each sample in the data set, calculate its distance to two cluster centers and classify it into the class corresponding to the cluster center with the smallest distance;

3) For each category, recalculate its cluster center (that is, the centroid of all samples belonging to this category);

4) Repeat the above two steps of 23 until a certain termination condition (number of iterations, minimum error change, etc.) is reached.

The second step This application designs a method for finding edge points by scanning. Take the center of mass obtained by K-means as the initial center of the circle, and scan radially along the radial direction to obtain a projection map along the radial direction. You can see the obvious boundary points as shown in Figure 4:

In order to improve the positioning accuracy, the gradient of the projection map is calculated:

Calculate the gradient in the Y direction; for the two-dimensional discrete Y direction derivative, it can be obtained by convolving the following convolution kernel with the circle image. Finally, the point with the largest gradient is selected as the boundary point.

On the basis of obtaining a series of edge points, the least square method is used for precise circle detection, and the No. 1 robot grabs the engine cylinder head according to the precise position of the installation grab point.

In order to achieve the above object, the present invention also provides a device for detecting three-dimensional defects of an engine cylinder head, including a light source, a No. 1 industrial robot, a No. 2 industrial robot, a recognition camera, a three-dimensional camera and a main control module. The light source is fixed on the detection area. On the top, the identification camera is installed on the No. 1 robot, and the No. 2 industrial robot is fixed with a 3D camera. The light source, No. 1 industrial robot, No. 2 industrial robot, identification camera, and 3D camera are electrically connected to the main control module respectively. Through the continuous switching of the positions of the No. 1 industrial robot and the No. 2 industrial robot, the organic combination of various functional components is realized, so as to realize the six-sided multi-angle switching when grasping and detecting the engine cylinder head, and the recognition camera can detect the engine cylinder head. The high-precision recognition of the grabbing point can also solve the problem of low recognition of the grabbing point in an environment with different light source angles, with high precision, and the 3D image obtained by the 3D camera is stored in the main control module for storage And processing, finally by calculating the detected point cloud registration data for 3D modeling and comparing the shape difference with the 3D model of the standard engine cylinder head, so as to detect and locate the defect position of the engine cylinder head.

The specific operation process of this device is as follows:

After the No. 1 robot grabs the B side of the engine cylinder head, it transports the engine cylinder head to be inspected to the designated inspection area, and rotates the angle between the TCP (Tool Center Point) of the robot and the joint axis so that the first surface of the engine cylinder head is parallel to the horizontal. The set light source can fully illuminate the first surface of the cylinder head. The No. 2 robot carries the 3D camera to an appropriate height, and the running track of the 3D camera is also parallel to the horizontal plane. The field of view can completely cover one side of the engine cylinder head, and the 3D image is acquired and transferred to the main control module; after acquiring all the 3D images of one side, the No. parallel to the level. The set light source can fully illuminate the two sides of the cylinder head. The No. 2 robot carries the 3D camera to an appropriate height, and the running track of the camera is also parallel to the horizontal plane. Transfer to the main control module; after obtaining all the 3D images of the two sides, the No. 1 robot rotates the TCP angle by 90°, so that the three sides of the engine cylinder head are parallel to the horizontal. The set light source can fully illuminate the 3 sides of the cylinder head. The No. 2 robot carries the 3D camera to an appropriate height, and the running track of the camera is also parallel to the horizontal plane. Transfer to the main control module; after obtaining all three-dimensional images of the three sides, the No. 1 robot rotates the TCP angle by 90°, so that the four sides of the engine cylinder head are parallel to the horizontal. The set light source can fully illuminate the 4 sides of the cylinder head. The No. 2 robot carries the 3D camera to an appropriate height, and the running track of the camera is also parallel to the horizontal plane. Transfer to the main control module; at this time, all four sides of the engine cylinder head have been collected; the No. 2 robot rotates the TCP and the joint axis so that the running track of the 3D sensor is parallel to the A surface, and the central axis of the LED light source of the 3D sensor is parallel to the A surface. Vertical, and the field of view of the 3D camera can completely cover the A surface of the engine cylinder head to acquire the 3D image and transfer it to the main control module; the No. 1 robot places the engine cylinder head in the designated area, and the engine cylinder head B faces the base In the positive direction of the Z axis of the coordinate system, the set light source can fully illuminate the B surface of the cylinder head. The No. 2 robot carries the 3D camera to an appropriate height, and the running track of the 3D camera is also parallel to the horizontal plane. The central axis of the LED light source of the 3D camera is perpendicular to the B surface, and the field of view of the 3D camera can completely cover the B surface of the engine cylinder head. Get and dump to the main control module.

The main control module receives the X, Y, and Z coordinate data of the point cloud sent by the 3D sensor, converts the received 3D array into an image through the software in the C# environment, and finally outputs the corresponding brightness map and height map. The luminance images are stitched together. This application adopts the SURF algorithm for stitching, extracts and matches the feature points of the acquired luminance images, and performs image registration after obtaining the matching point sets of the two images to be stitched, that is, the two images Convert to the same coordinates, use the homography matrix function to obtain the transformation matrix, and realize the registration of the image; at this time, the splicing of the image is not natural, because the transition between the two images needs to be weighted due to the inconsistency of the illumination color Fusion is to add the pixels in the overlapping area of the image according to a certain weight to synthesize a new image. Repeating this step can make each entire surface of the engine cylinder head into a complete picture, and perform defect detection on the brightness map of the engine cylinder head through the trained neural network.

The three-dimensional structural defect detection of the engine cylinder head requires point cloud registration; the data type file is converted into a binary text file through the software simulator of the three-dimensional camera, and the binary text file is converted into an ASSIC code point cloud file through open source software. Because the amount of point cloud data is relatively large, it affects the processing efficiency of point cloud data, and at the same time, due to the interference of the external environment, a large number of noise points will be collected, so the point cloud data needs to be filtered. In this application, the VoxelGrid filter is used to down-sample the point cloud, and a three-dimensional voxel grid is created through the input point cloud data, and the center of gravity of all points in each voxel is used to approximate other points in the voxel after being accommodated. , so that all points in the voxel are finally represented by a barycenter point;

And before the registration, first use the standard engine cylinder head to scan through the three-dimensional sensor to establish a standard 3D model; when obtaining the three-dimensional data of the engine cylinder head, there must be a position error, and its measurement coordinate system will change with the position And every time there are subtle changes. This application firstly uses the iterative closest point (ICP) algorithm for point cloud fine registration, selects the minimum distance between two points as the corresponding point each time, and calculates the rotation transformation matrix according to the corresponding relationship obtained each time to obtain a new The position of the point cloud is iterated until the distance error between the point clouds reaches the minimum. Compare the produced or repaired engine cylinder head with the standard 3D model after registration, and calculate the shape difference between the mold parts and the standard 3D model, so as to detect and locate the defect position of the mold.

In this embodiment, the method and device can be extended to the detection of rigid objects with complex structures and multi-faceted detection requirements, not limited to engine cylinder heads.

In this embodiment, the robot can be replaced by a multi-axis manipulator with the same function.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications and substitutions can be made to these embodiments without departing from the principle and spirit of the present invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (3)

1. A three-dimensional defect detection method of an engine cylinder cover is characterized by comprising the following steps:

the method comprises the following steps: grabbing an engine cylinder cover to be detected by a first industrial robot and sending the engine cylinder cover to an appointed detection area;

step two: the method comprises the following steps that a first industrial robot is matched with a second industrial robot, and a three-dimensional camera on the second industrial robot is used for obtaining three-dimensional images of six surfaces 1,2, 3, 4, A and B on an engine cylinder to be detected;

step three: carrying out point cloud registration on the obtained three-dimensional image of the engine cylinder cover;

step four: carrying out three-dimensional modeling on the detected point cloud registration data and comparing the point cloud registration data with a three-dimensional model of a standard engine cylinder cover in shape difference;

step five: and comparing the compared shape difference value with a set value, and judging as the defect position if the shape difference value is larger than the set value range or smaller than the set value range.

2. The method for detecting the three-dimensional defects of the engine cylinder cover according to the claim 1, characterized in that before the step one, a recognition camera on a first industrial robot is used for recognizing a grabbing point of the B surface of the engine cylinder cover to be detected.

3. The utility model provides a device that three-dimensional defect of engine cylinder lid detected, its characterized in that, includes light source, industrial robot, no. two industrial robot, discernment camera, three-dimensional camera and host system, and the light source is fixed at the top of detection area, and the discernment camera is installed on the robot, is fixed with the three-dimensional camera on No. two industrial robot, light source, industrial robot, no. two industrial robot, discernment camera, three-dimensional camera are connected with host system is automatically controlled respectively.

Images