CN112414623A

CN112414623A - Method and system for detecting part air tightness leakage defect based on artificial intelligence

Info

Publication number: CN112414623A
Application number: CN202011216000.7A
Authority: CN
Inventors: 周婷婷; 李俊
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-11-04
Filing date: 2020-11-04
Publication date: 2021-02-26

Abstract

The invention discloses a method and a system for detecting the air tightness leakage defect of a part based on artificial intelligence, and relates to the field of artificial intelligence. The method comprises the following steps: placing the part in a transparent water tank, inflating the part, and enhancing bubble characteristics in the water body by using linear array laser to enable the bubble characteristics to generate bright points due to total reflection; acquiring image information of a water body in the transparent water tank, wherein the image information comprises bubble bright spot characteristics in the water body; after the image information is processed by a deep neural network, outputting a bright spot image, wherein the bright spot image comprises the bubble bright spot characteristics and the position information thereof; acquiring the number of the positions of the bubble bright spots by utilizing a plurality of frames of bright spot images and combining the position information; and screening out the leaked bubbles by combining the flicker times of the bubble bright points at all positions with a preset threshold. The automatic detection of small leakage bubbles is realized, and the interference of the adhesion bubbles carried by the part to the detection structure is eliminated.

Description

Method and system for detecting part air tightness leakage defect based on artificial intelligence

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a method and a system for detecting the air tightness leakage defect of a part based on artificial intelligence.

Background

In production and life, the problems of air tightness leakage of air tightness parts of a plurality of devices can occur due to corrosion, high temperature and high pressure, cracks, abrasion, aging, long-time operation and the like. At present, in the process of performing an air tightness experiment test on a pressure vessel, high-pressure gas is generally directly filled in the pressure vessel, and then a tester manually or by using image processing equipment observes whether bubbles exist or not to judge whether parts leak or not.

However, some parts have smaller leakage holes, the generated leakage bubbles are smaller, the judgment error is larger through human eyes or a camera, and the judgment result is interfered by the adhesion bubbles carried by the parts entering the water body, so that the judgment result is wrong.

Disclosure of Invention

In order to solve the technical problems, the invention aims to provide a method and a system for detecting the part air tightness leakage defect based on artificial intelligence, and the adopted technical scheme is as follows:

in a first aspect, an embodiment of the present invention provides an artificial intelligence-based method for detecting a defect of hermetic leakage of a part, including the following steps:

placing the part in a transparent water tank, inflating the part, and enhancing bubble characteristics in the water body by using linear array laser to enable the bubble characteristics to generate bright points due to total reflection;

acquiring image information of a water body in the transparent water tank, wherein the image information comprises bubble bright spot characteristics in the water body;

after the image information is processed by a deep neural network, outputting a bright spot image, wherein the bright spot image comprises the bubble bright spot characteristics and the position information thereof;

acquiring the number of the positions of the bubble bright spots by utilizing a plurality of frames of bright spot images and combining the position information;

and screening out the leaked bubbles by combining the flicker times of the bubble bright points at all positions with a preset threshold.

Preferably, the obtaining the number of the positions of the bubble bright points by using a plurality of frames of the bright point images and combining the position information comprises the following steps:

carrying out binarization processing on the single-frame bright spot image to obtain a bright spot binary image;

superposing each frame of the bright spot binary images according to a time sequence to obtain a superposed image;

clustering the superposed graph, wherein the positions of the bubble bright spots are the number;

and superposing the same type pixel values of each position to obtain the flicker times of the bubble bright spots of the position.

Preferably, the method further comprises the step of acquiring the leakage speed of the leakage hole in the part by combining the sampling time, and the method comprises the following steps:

acquiring the bubble bright spot clustering center, setting a searching radius and establishing a searching area;

detecting bubble bright spots in the search area frame by frame according to a time sequence, and marking a time sequence;

obtaining a binary sequence according to the marking result, and calculating the average frame number between the bubble bright spot marks;

and combining the sampling frequency of the camera to obtain interval time, calculating the flicker frequency corresponding to the bubble bright spot, and taking the flicker frequency as the measurement value of the leakage speed corresponding to the leakage hole.

Preferably, the method further comprises positioning each leakage position of the part by combining the pose of the line laser, and the method comprises the following steps:

carrying out distortion correction on the image information to obtain a corrected image;

processing the corrected image through a semantic segmentation network to obtain an edge image, wherein the edge image comprises edge line characteristics of the transparent water tank;

according to the screened leakage bubble bright spots, combining the internal reference of a camera and the sideline characteristics, obtaining the positions (x, y) of the leakage bubble bright spots on the edge images;

taking one angular point in the transparent water tank as an original point, and combining with a sideline connected with the original point to establish a three-dimensional coordinate system;

according to the prior length of the water tank and the position of the leakage bubble bright spot on the edge image, obtaining the actual projection position (X, Y) of the leakage bubble bright spot on the main view angle of the transparent water tank, wherein the main view angle is the view angle when a camera collects the image information;

and acquiring a Z-direction coordinate of the leakage bubble bright point by combining the pose of the linear array laser, and acquiring the position (X, Y, Z) of the leakage bubble bright point in a three-dimensional coordinate system, wherein the leakage position corresponding to the leakage bubble bright point is (X, 0, Z).

In a second aspect, another embodiment of the present invention provides an artificial intelligence-based part airtightness leakage defect detection system, including:

the characteristic enhancement module is used for placing the parts in a transparent water tank, inflating the parts, and enhancing bubble characteristics in the water body by utilizing linear array laser so as to enable the bubble characteristics to generate bright points due to total reflection;

the image acquisition module is used for acquiring image information of the water body in the transparent water tank, wherein the image information comprises bubble bright spot characteristics in the water body;

the deep neural network module is used for outputting a bright spot image after the image information is processed by the deep neural network, wherein the bright spot image comprises the bubble bright spot characteristics and the position information thereof;

the bright spot analysis module is used for acquiring the number of the positions of the bubble bright spots by utilizing the bright spot images of the plurality of frames and combining the position information;

and the leakage bubble determination module is used for screening out leakage bubbles according to the flicker times of the bubble bright points at each position in combination with a preset threshold value.

Preferably, the bright spot analyzing module includes:

the binarization unit is used for carrying out binarization processing on the single-frame bright spot image to obtain a bright spot binary image;

the superposition map acquisition unit is used for superposing each frame of the bright spot binary images according to a time sequence to obtain a superposition map;

the position number acquisition unit is used for clustering the superposed graph and the position number of the bubble bright spots;

and the flicker frequency acquisition unit is used for superposing the same type of pixel values of each position to obtain the flicker frequency of the bubble bright spot at the position.

Preferably, the system further comprises a leakage rate obtaining module for obtaining a leakage rate of the leakage hole in the part by combining the sampling time, wherein the leakage rate obtaining module comprises:

the searching area establishing unit is used for acquiring the bubble bright point clustering center, setting a searching radius and establishing a searching area;

a time sequence marking unit for detecting the bubble bright spots in the search area frame by frame according to the time sequence and marking the time sequence;

the average frame number acquisition unit is used for acquiring a binary sequence according to the marking result and calculating the average frame number between the bubble bright spot marks;

and the leakage speed acquisition unit is used for obtaining interval time by combining the sampling frequency of the camera, calculating the flicker frequency corresponding to the bubble bright point, and taking the flicker frequency as the measurement value of the leakage speed corresponding to the leakage hole.

Preferably, the system further comprises a leak position locating module for locating each leak position of the part in combination with the pose of the line laser, the leak position locating module comprising:

the image correction unit is used for carrying out distortion correction on the image information to obtain a corrected image;

the semantic segmentation unit is used for processing the corrected image through a semantic segmentation network to obtain an edge image, and the edge image comprises side line characteristics of the transparent water tank;

the edge image bright spot positioning unit is used for acquiring the position (x, y) of the leaked bubble bright spot in the edge image according to the screened leaked bubble bright spot in combination with the internal reference and the sideline characteristics of the camera;

the world coordinate system establishing unit is used for combining a sideline connected with an origin point by taking one corner point in the transparent water tank as the origin point to establish a three-dimensional coordinate system;

the bright spot actual projection acquisition unit is used for acquiring the actual projection position (X, Y) of the leaked bubble bright spot at the main view angle of the transparent water tank according to the prior length of the water tank and the position of the leaked bubble bright spot at the edge image, wherein the main view angle is the view angle when the camera acquires the image information;

and the leakage position acquisition unit is used for acquiring a Z-direction coordinate of the leakage bubble bright point by combining the pose of the linear array laser, and acquiring the position (X, Y, Z) of the leakage bubble bright point in a three-dimensional coordinate system, wherein the leakage position corresponding to the leakage bubble bright point is (X, 0, Z).

The invention has at least the following beneficial effects:

according to the invention, a part is placed in a transparent water tank, the interior of the part is inflated, and then the linear array laser is used for enhancing the bubble characteristics in the water body, so that a bubble bright spot is obtained, wherein the bubble bright spot is generated by the total reflection of the bubble characteristics after the bubble characteristics are enhanced by the linear array laser. And then collecting image information of the water body in the transparent water tank, processing the image information by a neural network, and outputting a bright spot image, wherein the bright spot image comprises the bubble bright spot characteristics and the position information thereof. And finally, screening out the leaked bubbles according to the number of the flickering times of the bubble bright points at each position by combining a preset threshold value. The automatic detection of small leakage bubbles is realized, and the interference of the adhesion bubbles carried by the part to the detection structure is eliminated.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.

FIG. 1 is a schematic diagram of a method for detecting a defect of hermetic leakage of a part based on artificial intelligence according to an embodiment of the present invention;

FIG. 2 is a flowchart of a method for detecting a defect of hermetic leakage of a part based on artificial intelligence according to an embodiment of the present invention;

FIG. 3 is a view of a transparent water tank and a linear array laser pose graph in a part airtight leakage defect detection method based on artificial intelligence according to an embodiment of the present invention;

fig. 4 is a block diagram of a system for detecting a defect of hermetic leakage of a part based on artificial intelligence according to an embodiment of the present invention.

Detailed Description

To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description will be given to a method and a system for detecting a defect of air-tightness leakage of a part based on artificial intelligence according to the present invention, with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

The following describes a specific scheme of the part airtightness leakage defect detection method and system based on artificial intelligence in detail with reference to the accompanying drawings.

Referring to fig. 1, a schematic diagram of a method for detecting a defect of hermetic leakage of a part based on artificial intelligence according to an embodiment of the present invention is shown; FIG. 2 is a flow chart of a method for detecting a defect of hermetic leakage of a part based on artificial intelligence according to an embodiment of the present invention; fig. 3 shows a transparent water tank view and a linear array laser pose view in an artificial intelligence-based part airtight leakage defect detection method provided by an embodiment of the invention.

A method for detecting the air tightness leakage defect of a part based on artificial intelligence is shown in a flow chart of figure 1 and comprises the following steps:

step 1: placing the part in a transparent water tank, inflating the part, and enhancing bubble characteristics in the water body by using linear array laser to enable the bubble characteristics to generate bright points due to total reflection;

step 2: acquiring image information of a water body in a transparent water tank, wherein the image information comprises bubble bright spot characteristics in the water body;

and step 3: after the image information is processed by a deep neural network, a bright spot image is output, wherein the bright spot image comprises bubble bright spot characteristics and position information thereof;

and 4, step 4: acquiring the position number of the bubble bright spots by utilizing the multi-frame bright spot images and combining the position information;

and 5: and screening out the leaked bubbles according to the flicker times of the bubble bright points at each position and a preset threshold value.

In summary, the embodiment of the invention provides an artificial intelligence-based method for detecting the air tightness leakage defect of a part, which comprises the steps of placing the part in a transparent water tank, inflating the part, and then enhancing bubble characteristics in the water body by using linear array laser to obtain a bubble bright spot, wherein the bubble bright spot is a bright spot generated by total reflection after the bubble characteristics are enhanced by the linear array laser. And then collecting image information of the water body in the transparent water tank, processing the image information through a neural network, and outputting a bright spot image, wherein the bright spot image comprises bubble bright spot characteristics and position information thereof. And finally, screening out leaked bubbles by combining the flicker times of the bubble bright spots at each position with a preset threshold value. The automatic detection of small leakage bubbles is realized, and the interference of the adhesion bubbles carried by the part to the detection structure is eliminated.

Specifically, in step 1 of this embodiment, the characteristics of the bubbles in the water are enhanced by the linear array laser, and when the laser is emitted from the water to the bubbles, the laser is reflected and refracted on the surfaces of the bubbles due to the change of the medium, and the change in vision is a highlighted colored spot.

It should be noted that, when the bubbles are detected by using computer vision, the characteristics of the bubbles are often not obvious enough due to turbidity of the water body and inappropriate illumination intensity, and the bubbles are difficult to detect and segment. And after detecting the bubble, it is difficult to locate the defect through a single viewing angle. Due to the influence of a water body and a glass cylinder body, the RGB-D camera enables the obtained depth value information to be seriously lost and have larger error, and is difficult to use.

Specifically, in step 3 of the present embodiment, a deep learning neural network is used to identify and locate the bubble bright spots in the image information. Due to the fact that the background, the turbidity, the illumination depth and the laser wavelength of the water body are different in different scenes, a common image processing mode is difficult to adapt to different situations, and the generalization capability is limited. The method for learning the neural network deeply can enable the neural network to learn through a large amount of training data, then encode image information to extract image characteristics, and then restore the image through a decoding process to obtain the relative position of a key point in the image, thereby improving the generalization capability of the system. Specifically, the method comprises the following steps:

the deep neural network takes collected image information as input, firstly decodes, namely extracts spatial domain features in the image information in the process of downsampling the image information by using convolution and pooling operations, and then outputs feature vectors in the image information after processing by an Encoder (Encoder).

And processing the output feature vector by a first decoder (Dncoder1), and outputting a key point thermodynamic diagram with a certain scaling, wherein the number of channels of the key point thermodynamic diagram is two, one channel is a background channel, and the other channel is a bright point key point channel.

And after the key point thermodynamic diagram is subjected to post-processing, a bright point image is obtained, and the bright point image can reflect the specific coordinate position of the bubble bright point in the image.

Specifically, in the present embodiment, the deep neural network adopts an Encoder-Decoder1 structure, and the loss function in the network adopts a mean square error loss function. The scene that this network trained used data set collection is for the camera to press close to transparent jar, has the part because the bubble is emitted to the gas tightness problem in the jar, and when the bubble contacted laser, the surface took place to launch for demonstrate with the highlight point of laser with the colour. The label used for training is a key point hot spot label image corresponding to the test image, the position of a high-bright point in the image is marked firstly, and then the Gaussian kernel is used for fuzzy processing to obtain the corresponding key point hot spot.

It should be noted that there are many implementations of the encoder and decoder in the deep neural network, including the network structures of the Hourglass network Hourglass, Mask-RCNN, deplaybv 3, etc. In this embodiment, a decaplabv 3 network structure implementation manner is adopted.

Specifically, in step 4 and step 5 of this embodiment, since a large amount of adhesion bubbles exist in the part, the floating of the adhesion bubbles also generates a bright spot when contacting the laser, and thus the detected laser bright spot is not completely generated by the leakage bubbles. And the deep neural network is adopted for identification, the accuracy of the deep neural network is influenced by the background, so that the condition of false detection sometimes occurs. According to the invention, the position number of the bright points and the flickering times of the bright points at each position are firstly obtained in a multi-frame overlapping mode, so that the real leakage bubble point is judged. Specifically, the method comprises the following steps:

and carrying out binarization processing on the single-frame bright spot image to obtain a bright spot binary image. Specifically, the pixel value at the bubble bright point position is set to 1, and the pixel values at the other positions are set to 0.

And superposing the bright spot binary images of each frame according to the time sequence to obtain a superposed image.

Clustering the superposed graph to obtain the total number of the bright spots; specifically, because the position coordinates of the same bubble bright point determined by each frame of image fluctuate, the statistical result map IM is clustered to obtain the category number, which is the position number of the bubble bright point. In this embodiment, the clustering method uses K-MEANS.

And screening out the leaked bubbles by combining the flicker times of the bubble bright points at each position with a preset threshold.

Further, the method further comprises the step of obtaining the leakage rate of the part by combining the sampling time, so as to reflect the severity of the defect, specifically:

the bubble bright point cluster center is obtained and the search radius is set, and in this embodiment, the maximum distance from each point in the category to the center point is used as the search radius R.

And (4) combining the cluster center as an origin with the search radius, detecting bright spots in the range frame by frame according to the time sequence, and marking the time sequence. Specifically, detection is performed frame by frame from a start frame according to a time sequence, a clustering center of the bubble is taken as an origin, a radius R is taken as a search radius, whether a bright point exists in a detection range or not is determined, if yes, the time sequence is recorded as 1, and if not, the time sequence is 0.

And obtaining a binary sequence according to the detection result, and calculating the average frame number p of the adjacent bubbles touching the laser line. Specifically, the average frame number is an average number of 0 s spaced between 1 and 1 in the obtained binary sequence.

Combining the camera sampling frequency f to obtain an interval time t: t is p f. Calculating the flicker frequency q of the corresponding bubble bright point: q is 1/t, and the flicker frequency reflects the leakage rate of the corresponding leakage hole.

Further, the method also comprises the step of positioning the leakage position of the part by the pose of the linear array laser, as shown in fig. 3, in the implementation, a three-dimensional coordinate system is established by taking the corner point of the lower left corner under the visual angle of the transparent water tank main view 20 as an origin O, and under the visual angle of the transparent water tank right view 21, the linear array laser 23 is installed in a manner that the longitudinal side line of the left side of the transparent water tank is taken as a starting point 24 and is uniformly installed to the corner position of the water tank in an inclined manner, the included angle between the formed straight line and the horizontal line is theta, and the distance between the starting point and the bottom edge. According to this mounting manner, the position (X, 0, Z) of the part leakage hole can be located by acquiring the Z-axis distance Z of the bubble 22 from the origin point O according to the longitudinal distance Y of the bubble from the origin point O in the perspective of the front view 20 of the transparent water tank, and then combining the lateral distance X of the bubble from the origin point O. Specifically, the method comprises the following steps:

and carrying out distortion correction on the image information to obtain a corrected image. It should be noted that the distortion correction method of the image is a common method known to those skilled in the art, and will not be described in detail herein.

And processing the corrected image by a semantic segmentation network to obtain an edge image, wherein the edge image comprises the sideline characteristics of the transparent water tank. Specifically, in this embodiment, the semantic segmentation network includes a second decoder (decoder 2) and an Encoder (Encoder) shared with the deep neural network, and the semantic segmentation network is implemented in the same manner as the deep neural network in the structural form of Encoder-decoder 2.

And acquiring the position (x, y) of the leakage bubble in the edge image according to the screened leakage bubble in combination with the internal reference and side line characteristics of the camera. Specifically, the method comprises the following steps:

x＝um*dx

y＝vm*dy

wherein um is the pixel distance of bubble bright spot to marginal image left side sideline in the image, and vm is the pixel distance of bubble bright spot to marginal image bottom sideline in the image. dx and dy are internal parameters of the camera, dx represents the horizontal actual physical quantity length of a pixel point, and dy represents the vertical actual physical length of the pixel point.

And establishing a three-dimensional coordinate system by taking one corner point in the transparent water tank as an original point and combining a sideline connected with the original point.

And obtaining the actual projection position (X, Y) of the leakage bubble bright spot at the main visual angle of the transparent water tank according to the prior length of the water tank and the position of the leakage bubble at the edge image. Specifically, the method comprises the following steps:

X＝x*L/f

Y＝y*L/f

wherein, L is the distance from the optical center of the camera to the front view glass surface of the transparent water tank, and f is the focal length of the camera.

And acquiring a Z-direction coordinate of the leakage bubble bright point and acquiring the position (X, Y, Z) of the leakage bubble bright point in a three-dimensional coordinate system by combining the pose of the linear array laser, wherein the leakage position corresponding to the leakage bubble bright point is (X, 0, Z). Specifically, as shown in fig. 3, the position and posture of the linear array laser in the embodiment are utilized, and a geometric model is combined to solve the Z-direction coordinate of the leakage hole:

Z＝(Y-d)/tanθ

it should be noted that the part to be detected is located in the transparent water tank and at the bottom of the transparent water tank, so that when the position of the leakage hole of the part is located, the leakage position on the surface of the part can be obtained only by locating the coordinate position of the leakage hole of the part on the bottom surface of the water tank. In this case, the Y-axis coordinate position in the three-dimensional coordinate system is 0, i.e., the coordinates (X, 0, Z) indicate the leakage position on the surface of the part. The actual position of the leakage hole of the part under the coordinate is calculated by combining the position of the bubble in the image with the pose of the linear array laser, so that the positioning of the leakage hole under the single visual angle is realized.

Based on the same inventive concept as the method embodiment, the embodiment of the invention also provides a part air tightness leakage defect detection system based on artificial intelligence.

Referring to fig. 4, the system 100 for detecting a defect of hermetic leakage of a part includes a feature enhancement module 101, an image acquisition module 102, a deep neural network module 103, a bright spot analysis module 104, and a leakage bubble determination module 105.

Specifically, the characteristic enhancement module is used for placing the parts in a transparent water tank, inflating the parts, and enhancing bubble characteristics in the water body by utilizing the linear array laser, so that the bubble characteristics generate bright points due to total reflection. The image acquisition module is used for acquiring image information of the water body in the transparent water tank, and the image information comprises bubble bright spot characteristics in the water body. The deep neural network module is used for outputting a bright spot image after the image information is processed by the deep neural network, and the bright spot image comprises bubble bright spot characteristics and position information thereof. The bright spot analysis module is used for obtaining the position number of the bubble bright spots by utilizing a plurality of frames of bright spot images and combining the position information. And the leakage bubble determination module is used for screening out leakage bubbles according to the flicker times of the bubble bright points at each position in combination with a preset threshold value.

Further, the bright spot analysis module comprises a binarization unit, a superimposed graph acquisition unit, a position number acquisition unit and a flicker frequency acquisition unit.

Specifically, the binarization unit is configured to perform binarization processing on the single-frame bright point image to obtain a bright point binary image. The overlay image acquisition unit is used for overlaying each frame of bright spot binary image according to a time sequence to obtain an overlay image. The position number acquiring unit is used for clustering the superposed graph and the position number of the bubble bright spots. The flicker frequency acquisition unit is used for superposing the same type of pixel values of each position to obtain the flicker frequency of the bubble bright spot at the position.

Further, the system also comprises a leakage speed acquisition module which is used for acquiring the leakage speed of the leakage hole in the part by combining with the sampling time, wherein the leakage speed acquisition module comprises a search area establishing unit, a time sequence marking unit, an average frame number acquisition unit and a leakage speed acquisition unit.

Specifically, the search area establishing unit is configured to obtain a bubble bright point cluster center, set a search radius, and establish a search area. The time sequence marking unit is used for detecting the bubble bright spots in the search area frame by frame according to the time sequence and marking the time sequence. The average frame number acquisition unit is used for acquiring a binary sequence according to the marking result and calculating the average frame number between the bubble bright spot marks. The leakage speed acquisition unit is used for obtaining interval time by combining with the sampling frequency of the camera, calculating the flicker frequency of the corresponding bubble bright point, and taking the flicker frequency as the measurement value of the leakage speed of the corresponding leakage hole.

Furthermore, the system also comprises a leakage position positioning module which is used for positioning each leakage position of the part by combining the pose of the linear array laser, wherein the leakage position positioning module comprises an image correction unit, a semantic segmentation unit, an edge image bright point positioning unit, a world coordinate system establishing unit, a bright point actual projection acquisition unit and a leakage position acquisition unit.

Specifically, the image correction unit is configured to perform distortion correction on the image information to obtain a corrected image. The semantic segmentation unit is used for correcting the image and processing the image through a semantic segmentation network to obtain an edge image, and the edge image comprises edge line characteristics of the transparent water tank. The edge image bright spot positioning unit is used for acquiring the position (x, y) of the leaked bubble bright spot in the edge image according to the screened leaked bubble bright spot in combination with the internal reference and the sideline characteristics of the camera. The world coordinate system establishing unit is used for combining a sideline connected with an origin point by taking one corner point in the transparent water tank as the origin point to establish a three-dimensional coordinate system. The bright spot actual projection acquisition unit is used for acquiring actual projection positions (X, Y) of the leaked bubble bright spots at a main visual angle of the transparent water tank according to the prior length of the water tank and the positions of the leaked bubble bright spots at the edge images, wherein the main visual angle is the visual angle when the camera acquires image information. The leakage position acquisition unit is used for acquiring Z-direction coordinates of a leakage bubble bright point by combining the pose of the linear array laser, and acquiring the position (X, Y, Z) of the leakage bubble bright point in a three-dimensional coordinate system, wherein the leakage position corresponding to the leakage bubble bright point is (X, 0, Z).

It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A part air tightness leakage defect detection method based on artificial intelligence is characterized by comprising the following steps:

2. The method for detecting the airtightness leakage defect of the artificial intelligence-based part as claimed in claim 1, wherein said obtaining the number of positions of said bright spots of the bubbles by using a plurality of frames of said bright spot images in combination with said position information comprises the steps of:

3. The method for detecting the part airtightness leakage defect based on the artificial intelligence, as claimed in claim 1, further comprising obtaining the leakage rate of the leakage hole in the part in combination with the sampling time, and including the steps of:

4. The method for detecting the airtightness leakage defect of the artificial intelligence-based part, according to claim 1, further comprising positioning each leakage position of the part in combination with the pose of the line laser, including:

5. The utility model provides a part gas tightness leakage defect detecting system based on artificial intelligence which characterized in that includes:

6. The system as claimed in claim 5, wherein the bright spot analyzing module comprises:

7. The system of claim 5, further comprising a leakage rate obtaining module for obtaining a leakage rate of a leakage hole in the part according to the sampling time, wherein the leakage rate obtaining module comprises:

8. The system for detecting the defect of the airtight leakage of the part based on the artificial intelligence, as recited in claim 5, further comprising a leakage position locating module for locating each leakage position of the part in combination with the pose of the line laser, wherein the leakage position locating module comprises: