Dynamically adjustable cross light path lens system

Abstract

The invention discloses a dynamic adjustable cross light path lens system, which comprises a light path recombination module, and is characterized in that the light path recombination module comprises a double-asymmetric optical wedge assembly, a transmission module, a reflection module and a stroboscopic control unit; the double-asymmetric optical wedge assembly splits incident light into a plurality of cross light beams, and cross points cover a thin film detection plane; each of the transmission module and the reflection module comprises 16 LED lamp beads which are symmetrically distributed on two sides of a thin film detection plane; the stroboscopic control unit controls the LED lamp beads to be lightened independently or in groups according to a preset time sequence to generate nine detection images; the system has the beneficial effects that nine multi-dimensional images are generated through transmission / reflection double modules and stroboscopic control, multi-angle scratch signals are comprehensively captured, and the omission ratio is reduced to 0.1% or below; in combination with cross light paths and multi-image fusion, the gray difference stability between longitudinal scratches and the background is larger than 50 and is improved by 70%-180% compared with a traditional scheme; key parameters such as lens curvature radius, LED spacing and the like are determined, and the installation and adjustment difficulty is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of optical detection, and particularly relates to a dynamic adjustable cross light path lens system. BACKGROUND

[0002] In thin film surface defect detection, longitudinal scratch is a core defect type that needs to be accurately identified; the prior art mainly adopts a single illumination module, which has the following limitations:

[0003] A single light path can only capture scratch scattering signals at a specific angle, and the coverage of multi-angle and weak scratches is insufficient, which is easy to cause missed detection due to signal loss;

[0004] There is a lack of multi-modal illumination control mechanism, and scratch features cannot be enhanced through image comparison of different brightness and positions, so that background interference and pseudo-defects are difficult to effectively distinguish;

[0005] The lens system is not associated with module lamp bead layout parameters and light path crossing characteristics, resulting in low matching degree of light path optimization and defect capture;

[0006] The detection algorithm only relies on single image processing, does not combine multi-image fusion technology, cannot integrate multi-dimensional scratch information, and the recognition accuracy of weak defects is limited. SUMMARY

[0007] The purpose of the application is to provide a dynamic adjustable cross light path lens system, which solves the problems of "single light path coverage deficiency, no stroboscopic multi-image comparison, and lack of multi-image fusion" in the existing thin film longitudinal scratch detection, and realizes high-precision detection of multi-angle and weak scratches.

[0008] To achieve the above purpose, the application provides the following technical scheme: a dynamic adjustable cross light path lens system, comprising

[0009] A light path reorganization module, the light path reorganization module comprises a double-asymmetric optical wedge assembly, a transmission module, a reflection module and a stroboscopic control unit; the double-asymmetric optical wedge assembly divides the incident light into cross multi-path light rays, and the intersection point covers the thin film detection plane; the transmission module and the reflection module each contain 16 LED lamp beads, which are symmetrically distributed on both sides of the thin film detection plane; the stroboscopic control unit controls the LED lamp beads to be lit alone or in groups according to a preset time sequence, and generates 9 detection images;

[0010] A detection algorithm module, the detection algorithm module comprises a preprocessing unit, a feature extraction unit, a contrast enhancement unit, an image fusion unit and a segmentation and edge optimization unit, which is used for processing 9 detection images and extracting scratch features;

[0011] An AI analysis module, the AI analysis module receives a detection result, generates a light path parameter adjustment instruction, and feeds back to the light path reorganization module to realize adaptive optimization.

[0012] The intelligent detection interface is connected with the light path reorganization module, the detection algorithm module and the AI analysis module respectively, and is used for collecting image data, transferring light path parameters and realizing bidirectional communication.

[0013] As a preferred technical solution of the present application, the preset timing of the stroboscopic control unit is: the first scanning transmits the LED full-bright generating TBF image, the second to fifth scanning respectively lights up the four groups of lamp beads of the transmission module to generate TDF0-TDF3 images, and the sixth to ninth scanning respectively lights up the four groups of lamp beads of the reflection module to generate RBF0-RBF3 images.

[0014] As a preferred technical solution of the present application, the nine detection images include one full-illumination reference image TBF, four transmission group-illumination images TDF0-TDF3 and four reflection group-illumination images RBF0-RBF3.

[0015] As a preferred technical solution of the present application, the image fusion unit of the detection algorithm module adopts a weighted average algorithm to integrate the gray scale features of the nine images.

[0016] As a preferred technical solution of the present application, the reflection module further includes a mirror with an inclination angle of 45°, which is used to reflect the LED light to the film detection plane and form cross coverage with the light of the transmission module.

[0017] As a preferred technical solution of the present application, the light path reorganization module further includes an auxiliary positioning laser unit, the auxiliary positioning laser is a 650nm red light, the light path is coaxial with the detection sub-beam, the cross point imaging is captured by a CCD camera, and the positions of the two non-symmetrical optical wedges are feedback adjusted.

[0018] As a preferred technical solution of the present application, the AI analysis module is built-in with a trained optimization model to analyze the detection results.

[0019] Compared with the prior art, the present application has the following beneficial effects:

[0020] Through the transmission / reflection dual-module + stroboscopic control, nine multi-dimensional images are generated, and multi-angle scratch signals are comprehensively captured, and the missed detection rate is reduced to below 0.1%;

[0021] Combined with cross light path and multi-image fusion, the gray scale difference between longitudinal scratches and background is stable> 50, which is 70%-180% higher than the traditional scheme;

[0022] The key parameters such as lens curvature radius and LED spacing are determined, and the difficulty of assembly and adjustment is reduced;

[0023] The stroboscopic frequency, light intensity and algorithm filter template weight are synchronously adjusted, and the recognition rate of weak scratches is improved;

[0024] Adapt to light transmission materials such as adaptive film, photovoltaic, flexible screen, and compatible with various detection equipment such as industrial 8K line scan camera. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 It is a schematic diagram of a double asymmetric optical wedge structure of the present application;

[0026] Figure 2 It is a schematic diagram of two double asymmetric optical wedge structures of the present application;

[0027] Figure 3 It is a schematic diagram of multiple double asymmetric optical wedge structures of the present application;

[0028] Figure 4 It is a transmission single mode diagram of the present application;

[0029] Figure 5 It is a reflection module diagram of the present application;

[0030] Figure 6 It is an image with different generated positions and uneven light and shade of the present application. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0032] Please refer to Figures 1-6 The present application provides a dynamic adjustable cross optical path lens system, comprising

[0033] An optical path reorganization module, the optical path reorganization module comprising a double asymmetric optical wedge assembly, a transmission module, a reflection module and a stroboscopic control unit; the double asymmetric optical wedge assembly splits the incident light into cross multi-path light, and the intersection covers the film detection plane; the transmission module and the reflection module each contain 16 LED lamp beads, symmetrically distributed on both sides of the film detection plane; the stroboscopic control unit controls the LED lamp beads to light up individually or in groups according to the preset time sequence, generating 9 detection images:

[0034] A detection algorithm module, the detection algorithm module comprising a preprocessing unit, a feature extraction unit, a contrast enhancement unit, an image fusion unit and a segmentation and edge optimization unit, for processing 9 detection images and extracting scratch features;

[0035] An AI analysis module, the AI analysis module receiving detection results, generating optical path parameter adjustment instructions, and feeding back to the optical path reorganization module to realize adaptive optimization;

[0036] Intelligent detection interface, the intelligent detection interface is connected with the optical path reorganization module, the detection algorithm module and the AI analysis module respectively, and is used for collecting image data, transmitting optical path parameters and realizing bidirectional communication.

[0037] Specifically as follows:

[0038] The optical path reorganization module comprises:

[0039] The double asymmetric optical wedge assembly is made of fused quartz, and the curvature radii are R1=50 mm±0.5 mm and R2=80 mm±0.5 mm respectively. The incident light is split into crossed multiple light rays by the optical wedge, and the intersection point is calibrated by an auxiliary positioning laser (wavelength 650 nm red light), and the deviation from the film detection plane is ≤±0.1 mm.

[0040] The transmission module and the reflection module:

[0041] The transmission module contains 16 evenly arranged LED lamp beads, and the lamp bead spacing is 5 mm±0.1 mm. The distance between the module and the film detection plane is 50-200 mm (locked by a mechanical guide rail).

[0042] The reflection module is symmetrical to the transmission module and also contains 16 LED lamp beads, which are distributed on both sides of the film detection plane at 180° with the transmission module.

[0043] The stroboscopic control unit is connected with the transmission / reflection module, controls the LED lamp beads to be lit alone or in groups according to the preset time sequence, and generates 9 detection images with different brightness and positions.

[0044] The intelligent detection interface comprises:

[0045] The data acquisition interface is compatible with an industrial 8K line scan camera, receives 9 detection image data, and the acquisition rate is 1000 frames / s.

[0046] The parameter linkage unit collects the parameters of the optical path reorganization module in real time (crossing angle 0°-30°, stroboscopic frequency 50-100 Hz, LED light intensity 500-2000 lux), and transmits them to the detection algorithm module.

[0047] The communication protocol unit connects the AI analysis module by using the TCP / IP protocol, realizes bidirectional transmission of detection results and optical path adjustment instructions.

[0048] The detection algorithm module comprises:

[0049] The preprocessing unit adopts a 1×3 longitudinal long strip mask mean filter operator (mean_image) to suppress Gaussian noise and background non-uniformity, and the signal-to-noise ratio is improved to more than 30 dB.

[0050] Feature extraction unit: based on the background average gray value G, the threshold 0.8G is set to extract the low gray interference area (stain, shadow), and the interference is eliminated by the gray filling operator (paint_image_gray); contrast enhancement unit: the histogram equalization operator (equ_histo_image) is adopted to expand the gray dynamic range from 50-150 to 20-230;

[0051] Image fusion unit: the weighted average fusion is carried out on 9 detection images, the scratch features under different illuminations are integrated, the gray consistency signal of the scratch area is strengthened, and the random noise in a single image is suppressed;

[0052] Segmentation and edge optimization unit: the scratch area is segmented by the local dynamic threshold value of the 3*3 neighborhood mean, after the morphological opening operation (noise with an area <5 pixels is removed), the edge sharpening operator (emphasize) is adopted to make the scratch boundary gradient value increase by 40%.

[0053] The AI analysis module has a built-in reinforcement learning optimization model, receives the "scratch contrast, edge gradient value, and missed / missed detection number" output by the detection algorithm module, and generates an optical path adjustment instruction:

[0054] If the scratch contrast is <50, increase the cross angle of the optical wedge (such as from 10° to 15°);

[0055] If the false detection rate is >0.5%, reduce the LED light intensity by 10%-20%;

[0056] If the weak scratch is missed, adjust the frequency of the stroboscopic frequency to 80Hz, and optimize the multi-image fusion weight,

[0057] The application has a new detection method:

[0058] In a set of structures, there are two modules, which are a transmission single module and a reflection module;

[0059] On the projection module, there are a total of 16 lamp beads; in the application, the mode of stroboscopic is used for control, the first scanning controls 1-16 lamp beads to be fully lit to generate a TBF image; the second scanning controls the 1st, 5th, 9th and 13th to be lit to generate a TDF0 image; the third scanning controls the 2nd, 6th, 10th and 14th to be lit to generate a TDF1 image, the fourth scanning controls the 3rd, 7th, 11th and 15th to be lit to generate a TDF2 image, and the fifth scanning controls the 4th, 8th, 12th and 16th to be lit to generate a TDF3 image, so that the transmission module generates a total of 5 projection images;

[0060] On the reflection module, there are 16 lamp beads in total; in this application, a stroboscopic mode is used for control, the first, fifth, ninth and thirteenth are controlled to light up in the sixth scan to generate an RBF0 image; the second, sixth, tenth and fourteenth are controlled to light up in the seventh scan to generate an RBF1 image, the third, seventh, eleventh and fifteenth are controlled to light up in the eighth scan to generate an RBF2 image, and the fourth, eighth, twelfth and sixteenth are controlled to light up in the ninth scan to generate an RBF3 image, so that the transmission module generates a total of four projection images;

[0061] In this way, a total of nine images with different positions and uneven brightness are generated, and finally a new image is fused through the following image fusion algorithm;

[0062] Background noise suppression and directional filtering, mean filter preprocessing

[0063] For the image characteristics of the Haidongqing optical system, a longitudinal strip-shaped mask (1x3 template) is used to perform a mean filter operation (mean_image);

[0064] Mechanism of action: suppress Gaussian noise and background non-uniformity interference through directional filtering, while retaining the longitudinal structure characteristics of scratches, and improving the background smoothness and signal-to-noise ratio;

[0065] Feature region enhancement and interference suppression, feature extraction based on gray scale

[0066] The scratch area presents a bright area feature in the Haidongqing system. According to the background average gray value, a threshold is set to extract the interference area with a gray value lower than the threshold, and a gray scale filling (paint_image_gray) is used;

[0067] Technical purpose: eliminate low gray scale interference (such as stains, shadows), highlight the saliency of the scratch target, and reduce the interference of false defects;

[0068] Weak defect contrast enhancement, histogram equalization enhancement

[0069] For the low contrast characteristics of slight scratches, a gray scale histogram equalization enhancement (equ_histo_image) operator is used;

[0070] Optimization effect: expand the gray scale dynamic range, enhance the contrast between the scratch area and the background, and solve the problem of missed detection caused by uneven lighting of weak defects;

[0071] Defect precise segmentation and edge optimization, dynamic threshold segmentation and edge enhancement

[0072] The preprocessed image extracts the scratch area through local dynamic threshold segmentation, and combines morphological operation to remove noise residues;

[0073] An emphasize operator is used to perform edge sharpening enhancement on the segmented scratch area, so as to strengthen the gradient feature of the scratch boundary and improve the subsequent quantitative detection accuracy.

[0074] The effect of the experiment is shown in the following table:

[0075]

[0076] Significant contrast improvement: through the cross light path design, the gray scale difference between the longitudinal scratch and the background is greater than 50, which is 70%-180% higher than the traditional scheme (within 10 gray scale difference);

[0077] Strong adaptability: the cross angle of the light path can be dynamically adjusted between 0-30°, which is suitable for multi-angle longitudinal scratch detection and solves the problem of poor adaptability of single oblique incidence;

[0078] Improved signal-to-noise ratio: the cross light path makes the longitudinal scratch produce bidirectional diffraction enhancement, and the signal-to-noise ratio is improved by more than 40% compared with the vertical illumination scheme;

[0079] Linkage optimization: the intelligent detection interface realizes the linkage of the light path parameters and the detection algorithm, the algorithm dynamically adapts to the change of the light path, and further improves the detection accuracy of the weak defects;

[0080] Wide compatibility: compatible with surface defect detection of various light-transmitting materials such as thin film, photovoltaic and flexible screen, and expands the application scenarios.

[0081] Although the embodiments of the present application have been shown and described in detail, as described above, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A dynamically adjustable cross-optical-path lens system, characterized in that: include The optical path reconstruction module includes a dual asymmetric optical wedge assembly, a transmission module, a reflection module, and a strobe control unit. The dual asymmetric optical wedge assembly splits the incident light into multiple intersecting light rays, with the intersection points covering the thin film detection plane. The transmission module and the reflection module each contain 16 LED beads, symmetrically distributed on both sides of the thin film detection plane. The strobe control unit controls the LED beads to light up individually or in groups according to a preset timing sequence, generating 9 detection images. The detection algorithm module includes a preprocessing unit, a feature extraction unit, a contrast enhancement unit, an image fusion unit, and a segmentation and edge optimization unit, which are used to process nine detection images and extract scratch features. The AI ​​analysis module receives the detection results, generates optical path parameter adjustment instructions, and feeds them back to the optical path reconstruction module to achieve adaptive optimization. The intelligent detection interface is connected to the optical path reconstruction module, the detection algorithm module, and the AI ​​analysis module, respectively, and is used to acquire image data, transmit optical path parameters, and realize bidirectional communication.

2. The dynamically adjustable cross-optical-path lens system according to claim 1, characterized in that: The preset timing sequence of the strobe control unit is as follows: the first scan illuminates all LEDs of the transmission module to generate a TBF image; the second to fifth scans illuminate four groups of LEDs in the transmission module to generate TDF0-TDF3 images respectively; and the sixth to ninth scans illuminate four groups of LEDs in the reflection module to generate RBF0-RBF3 images respectively.

3. The dynamically adjustable cross-optical-path lens system according to claim 2, characterized in that: The nine detection images include one full illumination reference image (TBF), four transmission group illumination images (TDF0-TDF3), and four reflection group illumination images (RBF0-RBF3).

4. The dynamically adjustable cross-optical-path lens system according to claim 3, characterized in that: The image fusion unit of the detection algorithm module uses a weighted average algorithm to integrate the grayscale features of the nine images.

5. The dynamically adjustable cross-optical-path lens system according to claim 1, characterized in that: The reflection module also includes a reflector with a tilt angle of 45°, which is used to reflect the LED light onto the thin film detection plane, forming a cross-coverage with the light from the transmission module.

6. The dynamically adjustable cross-optical-path lens system according to claim 1, characterized in that: The optical path reconstruction module also includes an auxiliary positioning laser unit. The auxiliary positioning laser is 650nm red light, and the optical path is coaxial with the detection sub-beam. The cross-point imaging is captured by a CCD camera, and the position of the double asymmetric optical wedges is adjusted based on feedback.

7. The dynamically adjustable cross-optical-path lens system according to claim 1, characterized in that: The AI ​​analysis module has a built-in trained and optimized model to analyze the detection results.

Images