CN113902728A

CN113902728A - System for detecting product quality

Info

Publication number: CN113902728A
Application number: CN202111227532.5A
Authority: CN
Inventors: 陈新华; 唐顺海; 刘泉; 李星辉; 范余银
Original assignee: Hunan Aerospace Tianlu New Material Testing Co ltd
Current assignee: Hunan Aerospace Tianlu New Material Testing Co ltd
Priority date: 2021-10-21
Filing date: 2021-10-21
Publication date: 2022-01-07

Abstract

The invention relates to a system for detecting product quality, which adopts an image processing method of coarse positioning and fine analysis, and comprises the steps of firstly, obtaining a first image by coarsely scanning a product to be detected to judge whether the product has a defect, positioning a defect position when the product has the defect, and shifting the defect position to the center of a visual field of an image acquisition module; and then, a second image is obtained by accurately scanning the defect position, and the specific numerical value of the defect position is analyzed to more accurately judge the severity of the defect, so that a plurality of defects of the traditional mode are avoided: 1. compared with the traditional manual mode, the image processing method avoids subjectivity and time delay of manual participation, has higher automation degree, improves judgment accuracy and shortens judgment time; 2. compared with the mode of directly shooting the whole product to be detected to perform image processing, the mode of rough positioning and fine analysis further improves the judgment accuracy.

Description

System for detecting product quality

Technical Field

The invention relates to the field of image analysis, in particular to an image analysis system for detecting product quality.

Background

Casting is the basis of equipment manufacturing, the quality of castings directly influences the technical performance, service life, use safety and the like of the equipment. Taking a lot of key components of aerospace, weaponry and light alloy as examples, the detection of internal defects is required in the production process, and in order to keep the integrity of the appearance and the function of a workpiece from being damaged, the industrial CT nondestructive detection mode is mostly adopted.

In the traditional industrial CT nondestructive test, firstly, a casting is integrally scanned to obtain a CT image, and then, a professional judges the CT image manually. On one hand, the casting is integrally scanned, so that the time consumption and the long-acting rate are low, no pertinence is caused, and the accuracy is not high; on the other hand, the identification mode mainly depends on the decision-making of the experience of the detection personnel, the detection result is often interfered by the subjective judgment of the detection personnel, the result is not objective enough, the reliability is poor, and the phenomena of false detection and missed detection are easily caused.

Therefore, how to improve the quality detection efficiency and accuracy of the products, especially the degree of automation, is a technical problem to be solved urgently at present.

Disclosure of Invention

To solve the above technical problem, the present invention provides a system for detecting product quality, comprising: the system comprises an image acquisition module 100, a first analysis module 200, a positioning module 300, a second analysis module 400 and a judgment module 500;

the image acquisition module 100 is configured to acquire a first image of a product to be detected and send the first image to the first analysis module 200;

the first analysis module 200 is configured to receive the first image, analyze whether the product to be detected has a defect according to the first image, and determine a defect position when the product to be detected has the defect;

the positioning module 300 is configured to adjust the position of the image acquisition module 100 or the product to be detected according to the defect position, and shift the defect position to the center of the field of view of the image acquisition module 100;

the image acquisition module 100 is further configured to acquire a second image of the defect position, and send the second image to the second analysis module 400;

the second analysis module 400 is configured to receive the second image, and analyze a defect value of the defect position according to the second image;

the judging module 500 is configured to judge whether the product to be detected has a defect or not and the defect value, and whether the quality of the product to be detected is good or not.

Further, the positioning module 300 includes:

a control unit 310, connected to the first analysis module 200, for sending a shift control signal according to the defect position;

the shifting unit 320 is connected to the control unit 310, and configured to adjust the position of the image capturing module 100 or the product to be detected according to the shifting control signal, and shift the defect position to the center of the field of view of the image capturing module 100.

Further, the shift unit 320 includes: a lateral movement assembly 321 and a longitudinal movement assembly 322;

the transverse moving assembly 321 is configured to move the image capturing module 100 or the product to be detected in the X direction; the shift control signal comprises an X-bit adjusting signal;

the longitudinal moving assembly 322 is configured to move the image capturing module 100 or the product to be detected in the Y direction; the shift control signal comprises a Y-bit adjusting signal.

Further, the control unit 310 includes:

a small frame-cutting unit 311, configured to cut a frame image including the defect position with the defect position as a center;

a first small calculating unit 312, configured to calculate an X-bit difference value and a Y-bit difference value between the reference point of the square image and the reference point of the image acquisition module 100 or the product to be detected;

the first signal generating small unit 313 generates the X-bit adjustment signal and the Y-bit adjustment signal according to the X-bit difference value and the Y-bit difference value.

Further, the shift unit 320 further includes: an axial moving assembly 323 for moving the image capturing module 100 or the product to be detected in the Z direction; the shift control signal comprises a Z-bit adjusting signal.

Further, the control unit 310 further includes:

the second small calculating unit 314 is configured to calculate a Z-position adjustment value of the image acquisition module 100 or the product to be detected according to a preset visual field specification and the size of the square image;

the second signal generating unit 315 generates the Z-bit adjustment signal according to the Z-bit adjustment value.

Further, the shift unit 320 further includes: a rotation moving assembly 324, configured to rotate the image capturing module 100 around the product to be detected, or rotate the product to be detected.

Further, the first analysis module 200 comprises:

a constructing unit 210 for constructing a deep learning neural network detection model of the product, for learning whether the product is defective and a defective position;

the training unit 220 is connected with the constructing unit 210, and is used for inputting the sample set of the product into the deep learning neural network detection model, training the deep learning neural network detection model, and obtaining the trained deep learning neural network detection model;

the detecting unit 230 is connected to the training unit 220 and the image acquisition module 100, and is configured to input the first image of the product to be detected into the trained deep learning neural network detection model, so as to obtain whether the product has defects and the defect positions.

Further, the second analysis module 400 includes:

a segmentation unit 410, configured to perform defect segmentation on the second image by using a maximum inter-class variance dual-threshold algorithm;

and a calculating unit 420, connected to the dividing unit 410, for calculating relevant parameters of the divided defects as the defect values.

Further, the determining module 500 includes:

a first determining unit 510, configured to determine that the quality of the product to be detected is a first grade if the product to be detected is defect-free;

a first judging unit 520, if the product to be detected has defects, judging whether the defect value is larger than a set threshold value;

a second determining unit 530, configured to determine that the quality of the product to be detected is a second grade if the defect value is not greater than the set threshold;

a third determining unit 540, for determining the quality of the product to be detected as a third grade if the defect value is greater than the set threshold value

The invention provides a system for detecting product quality, which adopts an image processing method of coarse positioning and fine analysis, and comprises the steps of firstly, obtaining a first image by coarsely scanning a product to be detected to judge whether the product has a defect, positioning a defect position when the product has the defect, and shifting the defect position to the center of a visual field of an image acquisition module; and then, a second image is obtained by accurately scanning the defect position, and the specific numerical value of the defect position is analyzed to more accurately judge the severity of the defect, so that a plurality of defects of the traditional mode are avoided: 1. compared with the traditional manual mode, the image processing method avoids subjectivity and time delay of manual participation, has higher automation degree, improves judgment accuracy and shortens judgment time; 2. compared with the mode of directly shooting the whole product to be detected to perform image processing, the mode of rough positioning and fine analysis further improves the judgment accuracy.

Drawings

FIG. 1 is a block diagram of one embodiment of a system for product quality inspection of the present invention;

FIG. 2 is a block diagram of one embodiment of a first analysis module of the system for product quality detection of the present invention;

FIG. 3 is a block diagram of one embodiment of a positioning module of the system for product quality detection of the present invention.

Detailed Description

As shown in fig. 1, the present invention provides a system for product quality inspection, comprising: an image acquisition module 100, a first analysis module 200, a positioning module 300, a second analysis module 400, and a determination module 500.

The image acquisition module 100 is connected with the first analysis module, acquires a first image of a product to be detected, and sends the first image to the first analysis module 200; the first analysis module 200 is used for receiving the first image, analyzing whether the product to be detected has defects according to the first image, and determining the positions of the defects when the product to be detected has the defects; a positioning module 300, connected to the first analysis module 200, for adjusting the position of the image capturing module 100 or the product to be detected according to the defect position (fig. 1, which lists an exemplary example, but not limited to, that the positioning module 300 is connected to the image capturing module 100 to adjust the image capturing module 100), and shifting the defect position to the center of the field of view of the image capturing module 100; the image acquisition module 100 is further connected with the second analysis module 400, acquires a second image of the defect position, and sends the second image to the second analysis module 400; the second analysis module 400 receives the second image and analyzes the defect value of the defect position according to the second image; finally, the judging module 500 judges whether the product to be detected has defects or not according to the analyzed values of the defects and the quality of the product to be detected. It should be noted that each module is merely a division in functional configuration, and is not strictly limited in physical configuration. As can be appreciated by those skilled in the art, the modules can be arranged in a combined manner or in a split manner. For example, the image capturing module 100 may optionally, but not limited to, capture the first image or the second image with the same device in time-division and function-division manner, or capture the first image and the second image with different devices respectively disposed at suitable positions. The first analysis module 200, the second analysis module 400, the judgment module 500 and the subsequent control units are optionally but not limited to be centrally arranged to be functional modules of terminals such as an upper computer, a remote control center, a human-computer interface and the like, and through centralized arrangement, operation and control of a detector are facilitated, and shot image information, analyzed results whether the defects exist, specific defect data, defect types, subsequent judgment results and the like can be checked in real time.

In the embodiment, a system for detecting the product quality is provided, and an image processing method of coarse positioning and fine analysis is adopted, wherein a first image is obtained by coarse scanning of a product to be detected to judge whether the product has a defect, the position of the defect is positioned when the product has the defect, and the position of the defect is shifted to the center of the visual field of an image acquisition module; and then, a second image is obtained by accurately scanning the defect position, and the specific numerical value of the defect position is analyzed to more accurately judge the severity of the defect, so that a plurality of defects of the traditional mode are avoided: 1. compared with the traditional manual mode, the image processing method avoids subjectivity and time delay of manual participation, has higher automation degree, improves judgment accuracy and shortens manual judgment time; 2. compare and treat that the product is whole to carry out image processing in direct shooting, this coarse positioning + fine analysis's mode has shortened the time that directly sweeps, sweeps defect position moreover, has further improved the judgement precision.

Specifically, the method comprises the following steps:

1. image acquisition module

The image acquisition module 100, which may be but is not limited to a CT scanning device, first acquires a global DR image of a product to be detected as a first image, determines whether the product to be detected has a defect or not by viewing a global overview of the product to be detected, and searches for a specific position of the defect; and then mechanically moving the defect position of the product to be detected to the visual field center of the CT scanning equipment, and collecting the multi-angle and multi-dimensional local CT image of the specific defect position.

In the embodiment, the defect position is positioned through the DR image, so that the advantages of high imaging speed and small radiation quantity of the DR image are fully utilized, the pixel requirement of positioning the defect can be met, and the balance between the imaging effect and the time cost is achieved. And subsequently, CT scanning is adopted, the defect numerical value of the defect position is analyzed through the multi-angle and multi-dimensional local CT scanning image, high-resolution scanning of the defect position is realized, the severity of the defect can be further accurately analyzed, and the quality of the product is further deeply judged. It is noted that industrial CT is often used for quality inspection of castings, especially for inspection of light alloys such as aerospace and weaponry. However, the method for detecting the quality of the product is not limited to this field, the CT apparatus as the image acquisition module and the quality detection for the casting are only exemplary, and the quality detection method of the invention can be applied to any image analysis and product detection field, and the technical solution of the invention under the concept of coarse positioning and fine scanning is within the protection scope of the invention.

2. First analysis Module 200

The first analysis module 200, optionally but not limited to, includes:

a constructing unit 210 for constructing a deep learning neural network detection model of the product to learn whether the product is defective and a defective position;

the training unit 220 is connected with the construction unit 210, and is used for inputting the sample set of the product into the deep learning neural network detection model, training the deep learning neural network detection model and obtaining the trained deep learning neural network detection model;

and the detection unit 230 is connected with the training unit 220 and the image acquisition module 100, and is used for inputting the first image of the product to be detected into the trained deep learning neural network detection model to obtain whether the product has defects and the defect positions.

In this embodiment, a specific example of how the first analysis module 200 determines whether the product to be detected in the first image is defective, and how the position of the defect is known when it is defective is given, by constructing a single-input double-output deep learning neural network model and a sample set (marked with non-defective products, optional but not limited to high-quality products and the like, and defective products, defective positions, types and the like, optional but not limited to product images of good products, defective products, waste products and the like to form training data), then the sample set, the training data is sent into a deep learning neural network model, the deep learning neural network model after training is obtained after multiple parameter adjustments, and finally the defect detection is carried out on the first image (preferably the global DR image) through the trained neural network model so as to know whether the first image has defects and the specific defect position. It should be noted that the way of detecting whether there is a defect and the specific defect location by the deep learning neural network is only a preferred example of the inventor, and the image processing method for quality detection of the present invention is not limited thereto.

More preferably, as shown in fig. 2, the first analysis module 200 further includes a preprocessing unit 240, configured to preprocess the first image of the product to be detected before inputting the first image into the trained deep learning neural network detection model. More specifically, the preprocessing may be, but not limited to, performing preprocessing of image denoising and image enhancement on the first image, so as to further improve the accuracy of determining the neural network model, and further avoid the possibility of missing detection and false detection. More specifically, the preprocessing unit 240, optionally but not limited to being connected to the detection unit 230, is shown only as an illustrative example, since the preprocessing step can be performed at any step prior to inputting the first image of the product to be detected into the trained deep learning neural network detection model.

More preferably, the deep learning neural network detection model is also used for learning the defect type of the product; and then obtaining the defect types (air holes, looseness and the like) of the product for subsequent judgment of the quality of the product to be detected.

In this embodiment, the defect type of the product is further considered to further improve the multidimensional evaluation of the product quality. Illustratively, the defect type may optionally, but not exclusively, include one or more of inclusions, pores, porosity, shrinkage porosity, and the like.

3. Positioning module 300

As shown in fig. 1 and 3, the positioning module 300 may optionally, but not exclusively, include: a control unit 310 and a shift unit 320. The control unit 310 is connected to the first analysis module 200, and configured to send a shift control signal according to a defect position; and the shifting unit 320 is connected with the control unit 310 and is used for adjusting the position of the image acquisition module 100 or the product to be detected according to the shifting control signal and shifting the defect position to the center of the visual field of the image acquisition module 100.

In this embodiment, a specific embodiment of how the positioning module 300 adjusts the position of the image capturing module 100 or the product to be detected and shifts the defect position to the center of the image capturing module field of view is provided, so that the automatic operation of shifting is realized, the positioning is faster and more accurate, the rapidity and accuracy of image analysis are further improved, and the speed and accuracy of quality detection are improved.

More specifically, the shifting unit 320, optionally but not limited to, includes a transverse moving assembly 321, a longitudinal moving assembly 322, an axial moving assembly 323, and a rotational moving assembly 324, which enable the image capturing module 100 or the product to be detected to move in four dimensions, i.e., transverse, longitudinal, axial, and rotational directions. Wherein, the defect position is accurately positioned to the center of the visual field of the image acquisition module 100 by moving transversely and longitudinally; axially moving to enlarge or reduce the pixels so as to fully cover the defect position and simultaneously reach the maximum pixels (just covering the whole defect position, not only can the defect overall situation be summarized, but also the clearest fine scanning can be realized); and the defect position is scanned in multiple angles by rotating and moving, so that more accurate multidimensional analysis is realized. Therefore, in the embodiment, the four-dimensional arrangement of the mobile unit can further improve the precision scanning of the defect position, further improve the precision of quality precision detection, and avoid false detection and missing detection.

More specifically, the lateral moving assembly 321 is configured to move the image capturing module 100 or the product to be detected in the X direction; the corresponding shift control signal includes an X-bit adjustment signal. A longitudinal movement assembly 322 for moving the image acquisition module 100 or the product to be inspected in the Y direction; the corresponding shift control signal includes a Y-bit adjustment signal. An axial moving assembly 323 for moving the image acquisition module 100 or the product to be detected in the Z direction; the corresponding shift control signal comprises a Z-bit adjusting signal. The rotating and moving assembly 324 is used for rotating the image capturing module 100 around the product to be detected, or rotating the product to be detected, so as to place different angles of the product to be detected in the field of view of the image capturing module 100.

More specifically, taking the product to be detected as an example, the transverse moving assembly 321 may optionally but not limited to include an object stage for placing the product to be detected, a first chute (arranged along the X direction) arranged at the bottom of the object stage, a first sliding block arranged in the first chute and connected to the object stage, a first push rod connected to the first sliding block, and a first driving member connected to the first push rod. When a product to be detected needs to move in the X direction, the first driving piece can be started, the first push rod is driven to move in the first sliding groove in the X direction in a linkage mode, and therefore the object stage and the product to be detected arranged on the object stage are driven to move in the X direction. Similarly, the longitudinal moving component optionally, but not limited to, includes a second sliding chute (along the Y direction), a vertical rod disposed in the second sliding chute and connected to the side surface of the stage, a second push rod connected to the vertical rod, and a second driving member connected to the second push rod. When a product to be detected needs to move in the Y direction, the second driving assembly can be started, the second push rod is driven to move in the second sliding groove in the Y direction in a linkage mode, and therefore the object stage and the product to be detected arranged on the object stage are driven to move in the Y direction. More similarly, the axial moving component, optionally but not limited to, includes a third sliding groove (along the Z direction) disposed on the vertical rod, a third sliding block disposed in the third sliding groove and connected to the stage, a third push rod connected to the third sliding block, and a third driving member connected to the third push rod. When a product to be detected needs to move in the Z direction, the third driving piece can be started, the third push rod is driven to move in the third sliding groove in the Z direction in a linkage mode, and therefore the object stage and the product to be detected arranged on the object stage are driven to move in the Z direction. It should be noted that, in this embodiment, only the preferred mode of each moving component is given, and the method for implementing the four-dimensional movement, which can be understood by those skilled in the art, especially those skilled in the mechanical field, is within the scope of the present invention. The specific four-dimensional moving mode is only a specific example of the present invention, and the core inventive idea is how to implement a quality detection method of coarse positioning and fine scanning.

More specifically, for the lateral displacement assembly and the longitudinal displacement assembly, the control unit 310 may optionally, but not exclusively, include: a truncated frame small unit 311, a first calculation small unit 312 and a first signal generation small unit 313. First, the small frame-cutting unit 311 cuts a frame image including the defect position with the defect position as the center, and the specific size, form, and the like of the frame image can be arbitrarily set by those skilled in the art according to the parameters of the image capturing module 100, such as the field of view, the pixel size, and the like. Then, the first small calculating unit 312 calculates an X-bit difference value and a Y-bit difference value between the reference point of the square frame image and the reference point of the image acquisition module 100 or the product to be detected; taking the product to be detected as an example, the method can select, but is not limited to select, any point of the center point or any point of the four corners of the square frame image as the reference point of the square frame image, similarly select, as the reference point of the product to be detected, any corresponding point of the global center point or any point of the four corners of the product to be detected, and calculate the difference between the center point and the corresponding point in the X direction and the difference between the center point and the corresponding point in the Y direction. So that the first signal generating small unit 313 generates the X-bit adjustment signal and the Y-bit adjustment signal to the first driving member and the second driving member based thereon. More specifically, taking the central point as a reference point (an origin point on a coordinate system) as an example, when the difference value in the X direction is equal to 0, the representation does not need to move in the X direction, when the difference value in the X direction is less than 0, the representation needs to move in the opposite direction of X (left movement), and when the difference value in the X direction is greater than 0, the representation needs to move in the positive direction of X (right movement); similarly, when the difference in the Y direction is equal to 0, the token needs not to move in the Y direction, when the difference in the Y direction is less than 0, the token needs to move in the reverse direction of Y (backward), and when the difference in the Y direction is greater than 0, the token needs to move in the positive direction of Y (forward).

More specifically, for the axial moving assembly, the control unit 310 further includes: a second calculation small unit 314 and a second signal generation small unit 315. Firstly, the second small calculating unit 314 calculates the Z-position adjusting value of the image acquisition module (100) or the product to be detected according to the preset visual field specification and the size of the square image; and then the second signal generating small unit 315 generates the Z position adjusting signal to the third driving member according to the Z position adjusting value. Specifically, for example, when the product to be detected is moved, the preset visual field specification of the image detection module is compared with the size or multiple of the square frame image, when the preset visual field specification is equal to the size of the square frame image, the movement in the Z direction is not required, when the preset visual field rule is larger than the size of the square frame image, the movement (upward movement) in the Z positive direction is required, and when the preset visual field rule is smaller than the size of the square frame image, the movement (downward movement) in the Z negative direction is required. The specific Z-position adjustment value can be determined according to the size ratio of the preset visual field specification and the square frame image and the length of the vertical rod.

More specifically, for the rotation and displacement assembly, the control unit 310 further includes a third signal generating unit 316, which generates a rotation adjusting signal according to preset parameters such as speed, angle, interval time, and the like, so as to control the rotation parameters of the product to be detected or the image capturing module 100.

4. Second analysis Module 400

As shown in fig. 1, the second analysis module 400, optionally but not limited to, includes: a segmentation unit 410 and a calculation unit 420. The segmentation unit 410 is configured to perform defect segmentation on the second image by using a maximum inter-class variance dual-threshold algorithm; and the calculating unit 420 is connected with the dividing unit 410 and is used for calculating the area or/and the perimeter or/and the center of mass of the divided defects as the relevant parameters of the defect values.

In this embodiment, a specific example of how the second analysis module 400 processes the second image to analyze the specific value of the defect location is given. It should be noted that the relevant parameters may be selected, but not limited to, parameters including area, perimeter, and centroid of the defect, and one of the area, perimeter, and centroid, two of the area, perimeter, and centroid may be selected or other parameters may be added by those skilled in the art according to the evaluation criteria of the current quality (such as the quality requirement, factory requirement, etc. set forth in the application of the batch of products), or the specific type of the defect determined in advance, etc.

5. Judging module

As shown in fig. 1, the determining module 500 may optionally, but not limited to, include:

and the third judging unit 540 judges that the quality of the product to be detected is the third grade if the defect value is greater than the set threshold value.

In this embodiment, a determination module 500 is provided, which determines a specific evaluation manner of the product quality according to whether the detected product has defects and a specific value of the defects. Taking a product to be delivered from a factory as an example, when the product is inspected by the inspection system of the invention, if any defect is not detected after the first image is shot, the product can be directly judged as a first grade and is qualified product without subsequent inspection; if the defects are detected after the first image is shot, subsequent detection is needed, omnibearing scanning is carried out, a second image of a specific defect position of high-pixel accurate scanning is obtained, specific numerical values of the defects are specifically analyzed, if the numerical values are not larger than a set threshold value, the defects are represented to be small and not serious, the defects belong to an acceptable range, and the defects are determined to be a second grade, and a person skilled in the art can understand the defects as good products and calculate the defects as qualified products; if the value is greater than the set threshold, the defect is characterized as being very serious, belonging to an intolerable range, and is considered as a third grade, and the person skilled in the art can understand that the defect is poor product and not qualified product. Specifically, the set threshold value may be arbitrarily set by a person skilled in the art according to an index such as a quality requirement.

Of course, it should be noted that the specific evaluation criteria can be specifically set by those skilled in the art according to the application field, the application range and the quality requirement of the batch of products. For example, assuming that the product is used in the fields of aerospace, national defense and the like, only the first-grade product, i.e., the product without defects, is considered to be a qualified product; if the product is used for daily life, the products of the first grade and the second grade can be selected but not limited to be qualified products, namely, a little impurity, small air holes and the like can be ignored; the flexible evaluation standard of the specific quality can further reduce the probability of false detection and missed detection, and avoid the product quality from not meeting the requirements of users while reducing the production cost.

More specifically, if the relevant parameter of the defect value is one of the area, the perimeter, and the centroid, or two or three of the area, the perimeter, and the centroid, or other parameters are added, then the corresponding number needs to be set for the subsequent threshold setting value to correspond to the area, the perimeter, the centroid, and the like.

More specifically, a plurality of hierarchical setting thresholds may be set with respect to each parameter of the area, the circumference, the centroid, and the like. Taking the area as an example, a first area threshold (first defect threshold), a second area threshold (second defect threshold), and a third area threshold (third defect threshold) may be optionally set, wherein the first area threshold is smaller than the second area threshold, and the second area threshold is smaller than the third area threshold. If the product to be detected has defects, judging whether the defect numerical value of the product to be detected is larger than a first defect threshold value; if the quality is not greater than the first defect threshold, judging that the quality of the product to be detected is of a second grade, namely, the product is a good product (the product has defects, but the defects are small and can be ignored and be regarded as a qualified product); if the defect value is larger than the first defect threshold value, judging whether the defect value of the product to be detected is larger than a second defect threshold value (whether the defect value is between the first defect threshold value and the second defect threshold value); if the defect value is not greater than the second defect threshold (the defect value is between the first defect threshold and the second defect threshold), judging that the quality of the product to be detected is a defective product in a third grade, wherein the defect is selected but not limited to be only a slight defect by a person skilled in the art, and the product can be discounted for sale or sold in occasions with low requirements; as exemplified in the present application, it is not applicable to the fields of aerospace, national defense, but can be used in the field of daily life; if the defect value is larger than the second defect threshold (the defect value is not between the first defect threshold and the second defect threshold), judging whether the defect value of the current product to be detected is larger than a third set threshold (whether the defect value is between the second defect threshold and the third defect threshold); if the defect value is not greater than a third set threshold (the defect value is between the second defect threshold and the third defect threshold), judging that the quality of the product to be detected is a defective product in a third grade, wherein the quality of the product to be detected can be selected and not limited by the person skilled in the art to be sold at a lower discount or on a occasion with lower requirements, or scrapped; if the defect value is not less than the third defect threshold value (the defect value is not between the second defect threshold value and the third defect threshold value, and exceeds the third defect threshold value), the quality of the product to be detected is judged to be a waste product, and the product to be detected can be directly scrapped by a person skilled in the art.

In summary, the invention provides a system for detecting product quality based on image processing, which uses a core idea of coarse positioning (preferably performing defect positioning and classification identification based on a fast imaging DR technology and a deep learning neural network model), and fine analysis (preferably calculating a specific defect value based on defect segmentation of a multidimensional CT scanning technology for accurately scanning defect positions and a maximum inter-class variance dual-threshold algorithm), so that the judgment time is shortened, the false detection probability and the missed detection probability are reduced, and the system has the characteristics of high judgment speed and high precision.

The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A system for detecting product quality, comprising: the system comprises an image acquisition module (100), a first analysis module (200), a positioning module (300), a second analysis module (400) and a judgment module (500);

the image acquisition module (100) is used for acquiring a first image of a product to be detected and sending the first image to the first analysis module (200);

the first analysis module (200) is used for receiving the first image, analyzing whether the product to be detected is defective or not according to the first image, and determining the position of the defect when the product to be detected is defective;

the positioning module (300) is used for adjusting the position of the image acquisition module (100) or the product to be detected according to the defect position and shifting the defect position to the center of the visual field of the image acquisition module (100);

the image acquisition module (100) is further configured to acquire a second image of the defect position and send the second image to the second analysis module (400);

the second analysis module (400) is used for receiving the second image and analyzing the defect value of the defect position according to the second image;

the judging module (500) is used for judging the quality of the product to be detected according to whether the product to be detected has defects or not and the defect numerical value.

2. The system according to claim 1, wherein the positioning module (300) comprises:

the control unit (310) is connected with the first analysis module (200) and is used for sending out a displacement control signal according to the defect position;

the shifting unit (320) is connected with the control unit (310) and is used for adjusting the position of the image acquisition module (100) or the product to be detected according to the shifting control signal and shifting the defect position to the center of the visual field of the image acquisition module (100).

3. The system according to claim 2, wherein the displacement unit (320) comprises: a transverse moving assembly (321) and a longitudinal moving assembly (322);

the transverse moving assembly (321) is used for moving the image acquisition module (100) or the product to be detected in the X direction; the shift control signal comprises an X-bit adjusting signal;

the longitudinal moving assembly (322) is used for moving the image acquisition module (100) or the product to be detected in the Y direction; the shift control signal comprises a Y-bit adjusting signal.

4. The system according to claim 3, wherein the control unit (310) comprises:

a small frame-cutting unit (311) for cutting a frame image including the defect position with the defect position as a center;

the first small calculating unit (312) is used for calculating an X-bit difference value and a Y-bit difference value of the reference point of the square image and the reference point of the image acquisition module (100) or the product to be detected;

and a first signal generation small unit (313) that generates the X-bit adjustment signal and the Y-bit adjustment signal based on the X-bit difference value and the Y-bit difference value.

5. The system of claim 4, wherein the displacement unit (320) further comprises: an axial movement assembly (323) for moving the image acquisition module (100) or the product to be inspected in the Z direction; the shift control signal comprises a Z-bit adjusting signal.

6. The system of claim 5, wherein the control unit (310) further comprises:

the second small calculating unit (314) is used for calculating the Z-position adjusting value of the image acquisition module (100) or the product to be detected according to the preset visual field specification and the size of the square image;

and a second signal generation small unit (315) for generating the Z-bit adjustment signal according to the Z-bit adjustment value.

7. The system of claim 6, wherein the displacement unit (320) further comprises: a rotational movement assembly (324) for rotating the image acquisition module (100) around or around the product to be inspected.

8. The system according to any one of claims 1 to 7, characterized in that said first analysis module (200) comprises:

a construction unit (210) for constructing a deep learning neural network detection model of the product, for learning whether the product is defective and a defective position;

the training unit (220) is connected with the construction unit (210), inputs the sample set of the product into the deep learning neural network detection model, trains the deep learning neural network detection model, and obtains the trained deep learning neural network detection model;

and the detection unit (230) is connected with the training unit (220) and the image acquisition module (100) and is used for inputting the first image of the product to be detected into the trained deep learning neural network detection model to obtain whether the product has defects and the defect position.

9. The system according to claim 8, wherein the second analysis module (400) comprises:

a segmentation unit (410) for performing defect segmentation on the second image by using a maximum between-class variance dual-threshold algorithm;

and the calculating unit (420) is connected with the dividing unit (410) and is used for calculating relevant parameters of the divided defects as the defect numerical values.

10. The system according to claim 9, wherein said decision module (500) comprises:

a first judging unit (510) for judging the quality of the product to be detected to be a first grade if the product to be detected is defect-free;

a first judging unit (520) for judging whether the defect value is larger than a set threshold value if the product to be detected has defects;

a second judging unit (530) for judging the quality of the product to be detected to be a second grade if the defect value is not greater than the set threshold value;

and a third judging unit (540) for judging the quality of the product to be detected to be a third grade if the defect numerical value is larger than the set threshold.