CN-121997115-A - Multi-mode high-reflection product surface quality detection method and device

CN121997115ACN 121997115 ACN121997115 ACN 121997115ACN-121997115-A

Abstract

The invention discloses a method and a device for detecting the surface quality of a multi-mode high-reflection product, which belong to the field of quality detection, reject unqualified products caused by external dimensions and surface offset through preliminary detection, acquire a second two-dimensional image of the preliminary qualified products through a second two-dimensional camera, acquire a three-dimensional image of the preliminary qualified products through a three-dimensional camera, extract surface texture and color distribution characteristics through a convolutional neural network, capture surface microstructure and stress distribution information characteristics through a 3D convolutional network, splice surface texture and color distribution characteristics and surface microstructure and stress distribution information characteristic data, perform cross-modal interaction on the surface texture and color distribution characteristics, the surface microstructure and stress distribution information characteristics, output judging results and corresponding position coordinates through probability fusion, and not only can identify three-dimensional defects, but also can avoid data loss in a high-reflection area and have low cost through the steps.

Inventors

CHEN LIN
ZHANG CHENGSHU
Ge Weilun
SHEN DONGHUI
ZHAO LIYONG
LIU YUNFEI
GE DAWEI

Assignees

德创智控科技(苏州)有限公司
安徽财贸职业学院

Dates

Publication Date: 20260508
Application Date: 20251218

Claims (10)

1. The method for detecting the surface quality of the multi-mode high-reflection product is characterized by comprising the following steps of: The first two-dimensional camera acquires first two-dimensional images of a plurality of products to be detected on the vibration disc, and judges the material quality of the products, the outline dimension of the products to be detected and the surface contamination condition according to the first two-dimensional images; the first mechanical arm grabs the product to be detected, the preliminary quality of which is judged to be qualified, to a conveying structure, the conveying structure conveys the product to be detected to a material tray, and the first mechanical arm rejects the product to be detected, the preliminary quality of which is judged to be unqualified; Acquiring a second two-dimensional image of a product to be detected on a material tray through a second two-dimensional camera, and setting a light intensity sensor at the edge of the visual field of the second two-dimensional camera, wherein the light intensity sensor acquires the ambient light intensity and the spectrum distribution, and when the ambient light intensity exceeds a threshold value, the power of the second two-dimensional camera is reduced and the stroboscopic period is shortened; The method comprises the steps of precisely detecting three-dimensional image acquisition, namely arranging a polarizer at a light source of a three-dimensional camera, arranging an analyzer on the three-dimensional camera, rotating the analyzer through a motor drive, calculating an orthogonal extinction angle of the analyzer according to a material of a product, rotating the analyzer to the orthogonal extinction angle through the motor drive, and collecting a three-dimensional image of an orthogonal polarization mode of the product to be detected; And (3) multi-modal data fusion, namely carrying out coordinate system alignment on the second two-dimensional image and the three-dimensional image, carrying out multi-modal data normalization, adopting a convolutional neural network to extract surface textures and color distribution characteristics on the second two-dimensional image, capturing surface microstructure and stress distribution information characteristics of the three-dimensional image through a 3D convolutional network, splicing surface textures, color distribution characteristics and surface microstructure and stress distribution information characteristic data, carrying out cross-modal interaction on the surface textures and color distribution characteristics and the surface microstructure and stress distribution information characteristics, constructing a classification model, and outputting a judgment result and corresponding position coordinates through probability fusion.
2. The method for detecting the surface quality of the multi-mode highly reflective product according to claim 1, wherein the multi-mode data fusion step further comprises a logic check, and the logic check specifically comprises comparing a 2D detection result with a 3D detection result.
3. The method for detecting the surface quality of the multi-mode highly reflective product of claim 1, wherein the defect detection accuracy is improved by comparing the position coordinates of the outline defect in the 2D detection result with the position coordinates of the height abnormality in the 3D detection result.
4. The method for detecting the surface quality of the multi-mode high-reflectivity product of claim 1, wherein in the step of precisely detecting the two-dimensional image, the high-angle annular light source and the low-angle strip light source are adopted to trigger in a time sharing mode, and the second two-dimensional camera respectively collects multi-frame images and fuses the frame images to form a second two-dimensional image.
5. The method for detecting the surface quality of a multi-mode highly reflective product according to claim 1, wherein in the step of precisely detecting the three-dimensional image, a photoelectric encoder is arranged on the analyzer, and the photoelectric encoder detects the rotation angle of the analyzer and transmits the rotation angle to a control system to realize closed-loop control of the rotation angle of the analyzer.
6. The method for detecting the surface quality of the multi-mode highly reflective product according to claim 1, wherein in the step of precisely detecting the three-dimensional image, the product to be detected is located on a moving structure, the moving structure drives the product to be detected to move on a straight line, the three-dimensional camera is located above the moving structure, and the three-dimensional camera collects three-dimensional images of different angles of the product to be detected.
7. The multi-mode high-reflection product surface quality detection device is used for implementing the multi-mode high-reflection product surface quality detection method according to any one of claims 1-6 and is characterized by comprising a vibration disc, a first two-dimensional camera, a manipulator, a material disc, a second two-dimensional camera, a moving structure and a three-dimensional camera, wherein the vibration disc is used for enabling unordered products to be detected to finish gesture correction and uniform distribution through high-frequency vibration, the first two-dimensional camera is located above the vibration disc, the first two-dimensional camera is used for collecting first two-dimensional images of a plurality of products to be detected on the vibration disc and performing preliminary quality judgment on the products to be detected, the manipulator is used for grabbing the products to be detected to achieve transfer of the products to be detected, the second two-dimensional camera is located above the material disc to obtain second two-dimensional images of the products to be detected, the moving structure is used for driving the products to be detected to move, and the three-dimensional camera is located above the moving structure to obtain three-dimensional images of the products to be detected.
8. The device for detecting the surface quality of the multi-mode high-reflectivity product of claim 7, further comprising a base, wherein the vibration plate, the first two-dimensional camera, the manipulator, the second two-dimensional camera, the moving structure and the three-dimensional camera are fixed on the base, the material plate is rotatably installed on the base, the conveying structure is located between the vibration plate and the material plate, the number of the manipulators is two, the two manipulators are a first manipulator and a second manipulator, the first manipulator is located on one side of the vibration plate, and the second manipulator is located on one side of the material plate.
9. The device of claim 7, further comprising a light intensity sensor positioned at an edge of a field of view of the second two-dimensional camera, wherein the light intensity sensor is communicatively coupled to the second two-dimensional camera.
10. The device for detecting the surface quality of the multi-mode high-reflectivity product of claim 7, further comprising a polarizer, an analyzer, a motor and a photoelectric encoder, wherein the polarizer is arranged at a light source of the three-dimensional camera, the analyzer is arranged on the three-dimensional camera, the analyzer is in transmission connection with the motor, the photoelectric encoder is arranged on the analyzer, the motor drives the analyzer to rotate so that the analyzer rotates to an orthogonal extinction angle, and the photoelectric encoder detects the rotation angle of the analyzer to realize closed-loop control.

Description

Multi-mode high-reflection product surface quality detection method and device Technical Field The invention relates to the field of quality detection, in particular to a method and a device for detecting the surface quality of a multi-mode high-reflection product. Background With the rapid development of electronic devices in the miniaturization and portability directions, the importance of quality detection of button cells as key elements is increasingly highlighted. However, the current detection technology based on machine vision faces multiple challenges in practical application, namely firstly, the process difference adopted by different manufacturers leads to obvious differences in surface texture and size of the battery, secondly, the defects of surface scratches and the like are easy to generate in the production process, and furthermore, the high light reflection characteristic of the metal shell seriously affects the imaging quality. In addition, illumination variations and complex background disturbances of the production environment also present additional difficulties to the detection system. In the prior art, although a traditional 2D camera can acquire a plane image, the recognition capability of three-dimensional defects such as micro scratches and pits is limited, and a 3D line scanning camera can acquire morphological information, but data loss easily occurs when a high-reflection area is processed. If two sets of independent detection systems are adopted simultaneously, the equipment investment cost can be greatly increased, and the limited production line space can be occupied to influence the overall layout. Disclosure of Invention In order to overcome the defects in the prior art, one of the purposes of the invention is to provide a multi-mode high-reflection product surface quality detection method which can identify three-dimensional defects, can avoid data loss in a high-reflection area and has low cost. In order to overcome the defects in the prior art, the second aim of the invention is to provide a multi-mode high-reflection product surface quality detection device which can identify three-dimensional defects, has no data loss in a high-reflection area and is low in cost. One of the purposes of the invention is realized by adopting the following technical scheme: a multi-mode high-reflection product surface quality detection method comprises the following steps: The first two-dimensional camera acquires first two-dimensional images of a plurality of products to be detected on the vibration disc, and judges the material quality of the products, the outline dimension of the products to be detected and the surface contamination condition according to the first two-dimensional images; the first mechanical arm grabs the product to be detected, the preliminary quality of which is judged to be qualified, to a conveying structure, the conveying structure conveys the product to be detected to a material tray, and the first mechanical arm rejects the product to be detected, the preliminary quality of which is judged to be unqualified; Acquiring a second two-dimensional image of a product to be detected on a material tray through a second two-dimensional camera, and setting a light intensity sensor at the edge of the visual field of the second two-dimensional camera, wherein the light intensity sensor acquires the ambient light intensity and the spectrum distribution, and when the ambient light intensity exceeds a threshold value, the power of the second two-dimensional camera is reduced and the stroboscopic period is shortened; The method comprises the steps of precisely detecting three-dimensional image acquisition, namely arranging a polarizer at a light source of a three-dimensional camera, arranging an analyzer on the three-dimensional camera, rotating the analyzer through a motor drive, calculating an orthogonal extinction angle of the analyzer according to a material of a product, rotating the analyzer to the orthogonal extinction angle through the motor drive, and collecting a three-dimensional image of an orthogonal polarization mode of the product to be detected; And (3) multi-modal data fusion, namely carrying out coordinate system alignment on the second two-dimensional image and the three-dimensional image, carrying out multi-modal data normalization, adopting a convolutional neural network to extract surface textures and color distribution characteristics on the second two-dimensional image, capturing surface microstructure and stress distribution information characteristics of the three-dimensional image through a 3D convolutional network, splicing surface textures, color distribution characteristics and surface microstructure and stress distribution information characteristic data, carrying out cross-modal interaction on the surface textures and color distribution characteristics and the surface microstructure and stress distribution information characteristics, constructin