WO2022227103A1 - 一种零部件智能检测与标签系统及方法 - Google Patents

一种零部件智能检测与标签系统及方法 Download PDF

Info

Publication number
WO2022227103A1
WO2022227103A1 PCT/CN2021/092103 CN2021092103W WO2022227103A1 WO 2022227103 A1 WO2022227103 A1 WO 2022227103A1 CN 2021092103 W CN2021092103 W CN 2021092103W WO 2022227103 A1 WO2022227103 A1 WO 2022227103A1
Authority
WO
WIPO (PCT)
Prior art keywords
parts
detection
camera
identification code
aperture
Prior art date
Application number
PCT/CN2021/092103
Other languages
English (en)
French (fr)
Inventor
王国栋
Original Assignee
中奥智能工业研究院(南京)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中奥智能工业研究院(南京)有限公司 filed Critical 中奥智能工业研究院(南京)有限公司
Publication of WO2022227103A1 publication Critical patent/WO2022227103A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component

Definitions

  • the invention relates to the technical field of smart factories, in particular to an intelligent detection and labeling system for parts and components.
  • Industry 4.0 is dominated by intelligent manufacturing and utilizes the fourth industrial revolution of cyber-physical systems, Internet of Things, Industrial Internet of Things, cloud computing, cognitive computing and artificial intelligence technologies.
  • Smart factory is an important part of realizing Industry 4.0. Based on a series of scientific management practices in the manufacturing industry, it deeply integrates automation technology, information and communication technology and intelligent science and technology, and combines data, information and knowledge to build a more competitive new industry. A generation of manufacturing companies and their ecosystems. Among them, the intelligent and automated transformation of various production processes in the factory is the first step to promote the smart factory.
  • the process of defect detection, labeling, and information storage of parts in traditional factories is mainly implemented manually, which is time-consuming and labor-intensive, and the detection accuracy and missed detection rate are easily interfered by individual subjective factors.
  • a system that can integrate the process of factory parts classification, hole diameter detection, thread detection and label marking into an automated process is required.
  • the purpose of the present invention is to propose a system. It is an automatic intelligent detection and labeling system that realizes the inspection, labeling and information storage process of factory parts.
  • the present invention proposes an intelligent parts detection and labeling system, which can realize the automation of the inspection, labeling, and information storage processes of factory parts and improve the intelligence level.
  • an intelligent parts detection and labeling system of the present invention includes a conveyor belt for conveying parts, and the conveyor belt is provided with a part category camera, a part feature camera, a laser printer and an identification code along the direction of parts transmission.
  • the identification camera also includes a central processing module for processing the acquired data and outputting information.
  • the central processing module includes a communication module, and the central processing module establishes signal communication with the part category camera, the part feature camera, the laser printer and the identification code identification camera through the communication module, and the part category camera, the part feature camera and the identification code identification camera are respectively
  • an image acquisition card collects and converts the captured images into digital signals and transmits them to the central processing module through the communication module.
  • Client UI for interacting with the user.
  • an algorithm model is run on the central processing module, and the processing algorithm includes a component classification model, an aperture identification model, a thread detection model, and an equipment identification code identification model.
  • the central processing module also has a function module, and the function module includes an identification code generation module. , identification code verification module and interface status detection module.
  • the parts classification model abstracts the problem of classification of parts into an image classification problem, uses a CNN-based deep learning method to classify images, and customizes different convolution kernels and network structures according to specific scenarios, extracting specific scenarios. Component features.
  • the aperture recognition model abstracts the aperture detection problem of parts into a target detection problem.
  • the detection target is whether it contains apertures.
  • Connected domain calculate the area ratio of the inner and outer contours, and determine whether the aperture is qualified according to the comparison with the set threshold.
  • the thread detection model abstracts the thread defect detection problem as a target detection problem, uses Gaussian filtering to denoise the image and uses the method of histogram equalization to enhance the image, uses darknet50 to extract the features of the image, and uses yolo to Identify and detect defects, output the location coordinates of defects and defect types.
  • the types of defects include inner diameter, tapping, outer scratches, and foreign objects. If the detection results contain one or more types of defects, the component is judged as a substandard product. , otherwise it is a qualified product.
  • the equipment identification code recognition model converts the identification problem of parts and equipment identification codes into OCR text recognition problems, detects the area where the text is located through the text detection algorithm, and recognizes the content of the text through the text recognition algorithm. Compare the parts list information. If the corresponding code can be searched, it means the recognition is correct and the recognition result is returned; otherwise, the recognition fails and an error message is returned.
  • the present invention also proposes a method for intelligent detection and labeling of parts, comprising the following steps:
  • S3 Input the picture taken by the part feature camera, call the thread detection module, and judge whether the thread is qualified or not.
  • S4 Input the picture taken by the part feature camera, call the device identification code identification module to identify the device identification code that comes with the part, then the identification code generation module generates the identification code, and the laser printer marks the parts;
  • S5 Input the identification code to identify the picture taken by the camera, the identification code verification module identifies the label printed by the laser printer, and compares it with the label data in the system.
  • the intelligent detection and labeling system for parts of the present invention can integrate the classification, aperture detection, thread detection and label marking of factory parts into an automatic process, so as to realize the process of factory parts detection, labeling and information storage. of automation.
  • Fig. 1 is a system architecture diagram of a component intelligent detection and labeling system according to a preferred embodiment of the present invention
  • FIG. 2 is a hardware cooperation diagram of a component intelligent detection and labeling system according to a preferred embodiment of the present invention
  • FIG. 3 is a flowchart of a method for intelligent detection and labeling of parts according to a preferred embodiment of the present invention
  • Figure 4 is an end view of the component aperture.
  • an intelligent parts detection and labeling system includes a conveyor belt for conveying parts, and the conveyor belt is provided with a part category camera, part features along the direction of parts transmission. Cameras, laser printers, and identification code identification cameras, and also include a central processing module for processing collected data and outputting information.
  • the central processing module includes a communication module.
  • the central processing module establishes signal communication with the part category camera, the part feature camera, the laser printer and the identification code identification camera through the communication module.
  • the part category camera, the part feature camera and the identification code identification camera There are image capture cards on the cameras.
  • the image capture cards capture and capture images, convert them into digital signals, and transmit them to the central processing module through the communication module.
  • the central processing module runs on edge devices, mobile devices, or servers with X86 architecture. Modules have a client UI for interacting with the user. Users can interact with the system through a browser or client software.
  • the parts category camera includes one industrial camera, which takes pictures of the parts as a whole and is used to distinguish different parts.
  • the part feature camera consists of 5 industrial cameras, distributed on the left, right, front, rear and above the conveyor belt, for photographing parts from different angles.
  • the part feature camera will take pictures according to the part type identified by the part category camera and the preset camera angle, and only take pictures of the surface containing the inspection target.
  • the laser printer is used to print the generated identification code on the parts, and the side that prints the identification code needs to face up. It includes an industrial camera, which captures the identification code on the parts, identifies the identification code from the picture, and compares the identified content with the content built into the system to determine whether the label is qualified.
  • the central processing module There is an algorithm model running on the central processing module, and the processing algorithm includes a component classification model, an aperture identification model, a thread detection model, and an equipment identification code identification model.
  • the central processing module also has a functional module, which includes an identification code generation module, identification code Verification module and interface status detection module.
  • the identification code adopts DMC/QR/Bar code.
  • the aperture shapes of parts are various, and the parts classification model abstracts the part classification problem into an image classification problem, uses a CNN-based deep learning method to classify images, and customizes different volumes according to specific scenes Accumulate kernel and network structure to extract component features in specific scenarios.
  • the aperture For the detection of the aperture, it can be determined whether the aperture is qualified by detecting the contour and the connected domain, and calculating the area ratio of the inner and outer contours. First detect the contour area area inner of the aperture, and then detect the area outer of the rectangle or circle where the aperture is located, and calculate When the ratio is within a certain range, it is considered that the component aperture is qualified, otherwise it is unqualified.
  • For thread defect detection first use Gaussian filtering to denoise the image and use histogram equalization to enhance the image, then use darknet50 to extract the features of the image, use yolo to identify and detect defects, and finally output the defect's Position coordinates and defect categories, where the categories of defects include inner diameter, tapping, outer scratches, and foreign objects. If the inspection result contains one or more defects, the component is judged to be an unqualified product, otherwise it is a good product.
  • the equipment identification code recognition model converts the identification code recognition problem of parts and equipment into OCR text recognition problems. First, the text area is detected by the text detection algorithm, and then the text content is recognized by the text recognition algorithm. It supports horizontal, vertical, arc, etc. Layout text recognition. Compare the identified content with the parts list information at the time of storage. If the corresponding code can be found, it means the identification is correct and the identification result will be returned; otherwise, the identification fails and an error message will be returned.
  • a method for intelligent detection and labeling of parts of the present invention includes the following steps:
  • S3 Input the picture taken by the part feature camera, call the thread detection module, and judge whether the thread is qualified or not.
  • S4 Input the picture taken by the part feature camera, call the device identification code identification module to identify the device identification code that comes with the part, then the identification code generation module generates the identification code, and the laser printer marks the parts;
  • S5 Input the identification code to identify the picture taken by the camera, the identification code verification module identifies the label printed by the laser printer, and compares it with the label data in the system.
  • An intelligent detection and labeling system for parts of the present invention constructs the entire process of factory parts classification, aperture detection, thread detection and label marking into an automated system. First design the position and angle of the image capture camera, then use image classification, object detection and scene character recognition algorithms to classify parts, identify part apertures and calculate sizes, identify thread defects in parts, identify codes in parts, detect defects Qualified label codes are finally combined with industrial cameras, laser printers, image capture cards, intelligent edge devices and other auxiliary equipment to automate the process of factory parts inspection, labeling, and information storage.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Multimedia (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

一种零部件智能检测与标签系统及方法，该系统包括用于输送零部件的传送带，传送带上沿零部件传输的方向设有零件类别照相机、零件特征照相机、激光打印机和识别码识别相机，还包括用于处理采集数据和输出信息的中央处理模块。该系统能够实现零部件智能检测、打标检测的自动化，有利于提高智能化程度，降低人工参与度。

Description

一种零部件智能检测与标签系统及方法 技术领域
本发明涉及智慧工厂技术领域，具体涉及一种零部件智能检测与标签系统。
背景技术
工业4.0是以智能制造为主导，利用了信息物理系统、物联网、工业物联网、云计算、认知计算和人工智能技术的第四次工业革命。智慧工厂是实现工业4.0的重要一环，是在制造业一系列科学管理实践的基础上，深度融合自动化技术、信息通信技术和智能科学技术，结合数据、信息和知识建立更具竞争力的新一代制造业企业及其生态系统。其中，对工厂的各个生产流程的智能化和自动化改造是推进智慧工厂的第一步。
传统工厂中零部件的缺陷检测、打标签、信息入库的过程，主要依靠手工实现，既耗时又耗力，且检测准确率和漏检率易受个人主观因素的干扰。为实现零部件打标的智能化和自动化，需要一种能够将工厂零部件的分类、孔径检测、螺纹检测以及标签打标的过程整合为一个自动化流程的系统，本发明的目的即在于提出一种实现工厂零部件的检测、打标签、信息入库流程的自动化智能检测与标签系统。
发明内容
为克服现有技术的不足，本发明提出一种零部件智能检测与标签系统，能够实现工厂零部件的检测、打标签、信息入库流程的自动化，提升智能化水平。
为实现上述目的，本发明的一种零部件智能检测与标签系统，包括用于输送零部件的传送带，传送带上沿零部件传输的方向设有零件类别照相机、零件特征照相机、激光打印机和识别码识别相机，还包括用于处理采集数据和输出信息的中央处理模块。
进一步地，中央处理模块包括通信模块，中央处理模块通过通信模块与零件类别照相机、零件特征照相机、激光打印机和识别码识别相机建立信号通信，零件类别照相机、零件特征照相机和识别码识别相机上分别设有图像采集卡，图像采集卡采集拍摄图像转换为数字信号并通过通信模块传送给中央处理模块，中央处理模块运行在边缘端设备、移动端设备或X86架构的服务器上，中央处理模块具有用于与用户交互的客户端UI。
进一步地，中央处理模块上运行有算法模型，处理算法包括零部件分类模型、孔径识别模型、螺纹检测模型、设备标识码识别模型，中央处理模块上还具有功能模块，功能模块包括识别码生成模块、识别码校验模块以及接口状态检测模块。
进一步地，零部件分类模型将零部件类别判别问题抽象为图像分类问题，使用基于CNN的深度学习方法对图像进行分类，并根据特定的场景定制不同的卷积核和网络结构，抽取特定场景下零部件特征。
进一步地，孔径识别模型将零部件的孔径检测问题抽象为目标检测问题，首先通过目标检测算法如检测零部件中的孔径，检测目标为是否包含孔径，在含有孔径的图片上，通过检测轮廓和连通域，计算内外轮廓的面积比值，根据和设置的阈值进行对比确定孔径是否合格。
进一步地，检测孔径的轮廓面积area inner，再检测孔径所在矩形或者圆形的面积area outer，计算
Figure PCTCN2021092103-appb-000001
当比值在规定范围内认为部件孔径是合格的，否则不合格，判断逻辑如下
Figure PCTCN2021092103-appb-000002
进一步地，螺纹检测模型将螺纹缺陷检测问题抽象为目标检测问题，使用高斯滤波对图像进行降噪处理并使用直方图均衡的方法对图片进行增强，使用darknet50对图像的特征进行提取，使用yolo对缺陷进行识别和检测，输出缺陷的位置坐标以及缺陷类别，缺陷的类别包括内径、攻牙、外侧划痕、异物，若检测的结果中包含一类及以上缺陷则判断该零部件为不合格产品，否则为合格品。
进一步地，设备标识码识别模型将零部件设备标识码识别问题转换成OCR文字识别问题，通过文字检测算法检测出文字所在区域，通过文字识别算法识别文字的内容，将识别的内容和入库时零部件清单信息进行对比，若能搜索到对应的编码，说明识别正确，返回识别结果，否则识别失败，返回错误提示
本发明还提出一种零部件智能检测与标签方法，包括以下步骤：
S1：将零部件放上传送带，经过零件类别照相机，零件类别照相机拍摄产品图片，调用零部件分类模块判断是否是已知的零部件，若不是则结束，若是进入下一步；
S2：零部件经过零件特征照相机，零件特征照相机拍摄产品图片，调用孔径识别模型，首先检测零部件孔径，然后检测孔径是否合格，若不合格则结束，若合格进入 下一步；
S3：输入零件特征照相机拍摄的图片，调用螺纹检测模块，判断螺纹是否合格，若不合格则结束，若合格进入下一步；
S4：输入零件特征照相机拍摄的图片，调用设备标识码识别模块识别零部件自带的设备识别码，然后由识别码生成模块生成识别码，由激光打印机在零部件上打标；
S5：输入识别码识别相机拍摄的图片，识别码校验模块识别激光打印机打印的标签，并和系统中的标签数据进行对比，若不合格，则结束，若合格则进行零部件信息入库。
本发明的一种零部件智能检测与标签系统能够将工厂零部件的分类、孔径检测、螺纹检测以及标签打标的整合为一个自动化流程，实现工厂零部件的检测、打标签、信息入库流程的自动化。
附图说明
下面结合附图对本发明作进一步描写和阐述。
图1是本发明首选实施方式的一种零部件智能检测与标签系统的系统架构图；
图2是本发明首选实施方式的一种零部件智能检测与标签系统的硬件协作图；
图3是本发明首选实施方式的一种零部件智能检测与标签方法的流程图；
图4是零部件孔径的端面视图。
具体实施方式
下面将结合附图、通过对本发明的优选实施方式的描述，更加清楚、完整地阐述本发明的技术方案。
如图1和图2所示，本发明首选实施方式的一种零部件智能检测与标签系统，包括用于输送零部件的传送带，传送带上沿零部件传输的方向设有零件类别照相机、零件特征照相机、激光打印机和识别码识别相机，还包括用于处理采集数据和输出信息的中央处理模块。
如图1所示，中央处理模块包括通信模块，中央处理模块通过通信模块与零件类别照相机、零件特征照相机、激光打印机和识别码识别相机建立信号通信，零件类别照相机、零件特征照相机和识别码识别相机上分别设有图像采集卡，图像采集卡采集拍摄图像转换为数字信号并通过通信模块传送给中央处理模块，中央处理模块运行在边缘端设备、移动端设备或X86架构的服务器上，中央处理模块具有用于与用户交互 的客户端UI。用户可通过浏览器或客户端软件与系统进行交互。
其中，零件类别照相机包含1台工业相机，从整体对零部件进行拍摄，用于区分不同的零部件。零件特征照相机包含5台工业相机，分布在左、右、前、后和传送带上方，用于从不同的角度拍摄零部件。零件特征照相机会根据零件类别照相机识别的部件类型和预先设定的拍照角度进行拍摄，并只对包含检测目标的面进行拍照。激光打印机用于将生成的识别码，打印到零部件上，打印识别码的面需要朝上放置。包含1台工业相机，拍摄零部件上的识别码，并从图片中识别识别码，将识别的内容和系统内置的内容进行对比，从而判断所打的标签是否合格。
中央处理模块上运行有算法模型，处理算法包括零部件分类模型、孔径识别模型、螺纹检测模型、设备标识码识别模型，中央处理模块上还具有功能模块，功能模块包括识别码生成模块、识别码校验模块以及接口状态检测模块。识别码采用DMC/QR/Bar码。
如图4所示，零部件的孔径形状多样，零部件分类模型将零部件类别判别问题抽象为图像分类问题，使用基于CNN的深度学习方法对图像进行分类，并根据特定的场景定制不同的卷积核和网络结构，抽取特定场景下零部件特征。
针对孔径的检测，可以通过检测轮廓和连通域，计算内外轮廓的面积比值来确定孔径是否合格。首先检测孔径的轮廓面积area inner，再检测孔径所在矩形或者圆形的面积area outer，计算
Figure PCTCN2021092103-appb-000003
当比值在一定范围时，则认为部件孔径是合格的，否则不合格。
检测孔径的轮廓面积area inner，再检测孔径所在矩形或者圆形的面积area outer，计算
Figure PCTCN2021092103-appb-000004
当比值在规定范围内认为部件孔径是合格的，否则不合格，判断逻辑如下
Figure PCTCN2021092103-appb-000005
针对螺纹缺陷检测，首先使用高斯滤波对图像进行降噪处理并使用直方图均衡的方法对图片进行增强，然后使用darknet50对图像的特征进行提取，使用yolo对缺陷进行识别和检测，最后输出缺陷的位置坐标以及缺陷类别，其中缺陷的类别包括内径、攻牙、外侧划痕以及异物等。若检测的结果中包含一类及其以上缺陷，则判断该零部 件为不合格产品，否则为良品。
设备标识码识别模型将零部件设备标识码识别问题转换成OCR文字识别问题，首先通过文字检测算法检测出文字所在区域，然后再通过文字识别算法识别文字的内容，支持横向、纵向、弧形等布局的文字识别。将识别的内容和入库时零部件清单信息进行对比，若能搜索到对应的编码，说明识别正确，返回识别结果，否则识别失败，返回错误提示。
如图3所示，本发明的一种零部件智能检测与标签方法，包括以下步骤：
S1：将零部件放上传送带，经过零件类别照相机，零件类别照相机拍摄产品图片，调用零部件分类模块判断是否是已知的零部件，若不是则结束，若是进入下一步；
S2：零部件经过零件特征照相机，零件特征照相机拍摄产品图片，调用孔径识别模型，首先检测零部件孔径，然后检测孔径是否合格，若不合格则结束，若合格进入下一步；
S3：输入零件特征照相机拍摄的图片，调用螺纹检测模块，判断螺纹是否合格，若不合格则结束，若合格进入下一步；
S4：输入零件特征照相机拍摄的图片，调用设备标识码识别模块识别零部件自带的设备识别码，然后由识别码生成模块生成识别码，由激光打印机在零部件上打标；
S5：输入识别码识别相机拍摄的图片，识别码校验模块识别激光打印机打印的标签，并和系统中的标签数据进行对比，若不合格，则结束，若合格则进行零部件信息入库。
本发明的一种零部件智能检测与标签系统将工厂零部件的分类、孔径检测、螺纹检测以及标签打标的整个流程构建为一个自动化系统。首先设计图像采集相机的位置和角度，然后使用图像分类、目标检测以及场景字符识别算法，对零件进行分类、识别零件孔径并计算大小、识别零件中螺纹的缺陷、识别零件中的编码、检测不合格的标签码，最终结合工业相机、激光打印机、图像采集卡、智能边缘端设备以及其他辅助设备，实现工厂零部件的检测、打标签、信息入库流程的自动化。
上述具体实施方式仅仅对本发明的优选实施方式进行描述，而并非对本发明的保护范围进行限定。在不脱离本发明设计构思和精神范畴的前提下，本领域的普通技术人员根据本发明所提供的文字描述、附图对本发明的技术方案所作出的各种变形、替代和改进，均应属于本发明的保护范畴。本发明的保护范围由权利要求确定。

Claims (9)

  1. 一种零部件智能检测与标签系统，其特征在于，包括用于输送零部件的传送带，所述传送带上沿零部件传输的方向设有零件类别照相机、零件特征照相机、激光打印机和识别码识别相机，还包括用于处理采集数据和输出信息的中央处理模块。
  2. 根据权利要求1所述的一种零部件智能检测与标签系统，其特征在于，所述中央处理模块包括通信模块，所述中央处理模块通过通信模块与零件类别照相机、零件特征照相机、激光打印机和识别码识别相机建立信号通信，所述零件类别照相机、零件特征照相机和识别码识别相机上分别设有图像采集卡，所述图像采集卡采集拍摄图像转换为数字信号并通过通信模块传送给中央处理模块，所述中央处理模块运行在边缘端设备、移动端设备或X86架构的服务器上，所述中央处理模块具有用于与用户交互的客户端UI。
  3. 根据权利要求2所述的一种零部件智能检测与标签系统，其特征在于，所述中央处理模块上运行有算法模型，所述处理算法包括零部件分类模型、孔径识别模型、螺纹检测模型、设备标识码识别模型，所述中央处理模块上还具有功能模块，所述功能模块包括识别码生成模块、识别码校验模块以及接口状态检测模块。
  4. 根据权利要求3所述的一种零部件智能检测与标签系统，其特征在于，所述零部件分类模型将零部件类别判别问题抽象为图像分类问题，使用基于CNN的深度学习方法对图像进行分类，并根据特定的场景定制不同的卷积核和网络结构，抽取特定场景下零部件特征。
  5. 根据权利要求3所述的一种零部件智能检测与标签系统，其特征在于，所述孔径识别模型将零部件的孔径检测问题抽象为目标检测问题，首先通过目标检测算法如检测零部件中的孔径，检测目标为是否包含孔径，在含有孔径的图片上，通过检测轮廓和连通域，计算内外轮廓的面积比值，根据和设置的阈值进行对比确定孔径是否合格。
  6. 根据权利要求5所述的一种零部件智能检测与标签系统，其特征在于，检测孔径的轮廓面积area inner，再检测孔径所在矩形或者圆形的面积area outer，计算
    Figure PCTCN2021092103-appb-100001
    当比值在规定范围内认为部件孔径是合格的，否则不合格，判断逻辑如下
    Figure PCTCN2021092103-appb-100002
  7. 根据权利要求3所述的一种零部件智能检测与标签系统，其特征在于，所述螺 纹检测模型将螺纹缺陷检测问题抽象为目标检测问题，使用高斯滤波对图像进行降噪处理并使用直方图均衡的方法对图片进行增强，使用darknet50对图像的特征进行提取，使用yolo对缺陷进行识别和检测，输出缺陷的位置坐标以及缺陷类别，缺陷的类别包括内径、攻牙、外侧划痕、异物，若检测的结果中包含一类及以上缺陷，则判断该零部件为不合格产品，否则为合格品。
  8. 根据权利要求3所述的一种零部件智能检测与标签系统，其特征在于，所述设备标识码识别模型将零部件设备标识码识别问题转换成OCR文字识别问题，通过文字检测算法检测出文字所在区域，通过文字识别算法识别文字的内容，将识别的内容和入库时零部件清单信息进行对比，若能搜索到对应的编码，说明识别正确，返回识别结果，否则识别失败，返回错误提示。
  9. 一种零部件智能检测与标签方法，其特征在于，包括以下步骤：
    S1：将零部件放上传送带，经过零件类别照相机，零件类别照相机拍摄产品图片，调用零部件分类模块判断是否是已知的零部件，若不是则结束，若是进入下一步；
    S2：零部件经过零件特征照相机，零件特征照相机拍摄产品图片，调用孔径识别模型，首先检测零部件孔径，然后检测孔径是否合格，若不合格则结束，若合格进入下一步；
    S3：输入零件特征照相机拍摄的图片，调用螺纹检测模块，判断螺纹是否合格，若不合格则结束，若合格进入下一步；
    S4：输入零件特征照相机拍摄的图片，调用设备标识码识别模块识别零部件自带的设备识别码，然后由识别码生成模块生成识别码，由激光打印机在零部件上打标；
    S5：输入识别码识别相机拍摄的图片，识别码校验模块识别激光打印机打印的标签，并和系统中的标签数据进行对比，若不合格，则结束，若合格则进行零部件信息入库。
PCT/CN2021/092103 2021-04-29 2021-05-07 一种零部件智能检测与标签系统及方法 WO2022227103A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110474725.4 2021-04-29
CN202110474725.4A CN113066087B (zh) 2021-04-29 2021-04-29 一种零部件智能检测与标签系统及方法

Publications (1)

Publication Number Publication Date
WO2022227103A1 true WO2022227103A1 (zh) 2022-11-03

Family

ID=76567895

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/092103 WO2022227103A1 (zh) 2021-04-29 2021-05-07 一种零部件智能检测与标签系统及方法

Country Status (2)

Country Link
CN (1) CN113066087B (zh)
WO (1) WO2022227103A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117218097A (zh) * 2023-09-23 2023-12-12 宁波江北骏欣密封件有限公司 一种轴套类丝网垫圈零件表面缺陷检测方法及装置

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115292540B (zh) * 2022-07-22 2023-06-13 杭州易有料科技有限公司 多模态零件信息识别方法、装置、设备和计算机可读介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105158268A (zh) * 2015-09-21 2015-12-16 武汉理工大学 精冲零部件缺陷智能在线检测方法、系统及装置
CN105718836A (zh) * 2016-01-20 2016-06-29 深圳市黑云信息技术有限公司 一种标签自动喷码、错漏检测的控制方法
CN106767416A (zh) * 2017-01-04 2017-05-31 纵科（武汉）信息技术有限公司 零件检测系统及方法
CN110148106A (zh) * 2019-01-18 2019-08-20 华晨宝马汽车有限公司 一种利用深度学习模型检测物体表面缺陷的系统及方法
CN110239997A (zh) * 2019-06-13 2019-09-17 华中科技大学 一种复卷机自动剔废方法及设备
US20210056681A1 (en) * 2018-05-03 2021-02-25 Inspekto A.M.V. Ltd. System and method for visual production line inspection of different production items

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108257171A (zh) * 2018-01-09 2018-07-06 江苏科技大学 基于光视觉的汽车雷达装配孔径检测方法
CN108465648A (zh) * 2018-04-23 2018-08-31 苏州香农智能科技有限公司 一种基于机器视觉的磁芯自动分拣系统
CN110788024A (zh) * 2019-11-13 2020-02-14 苏州大成有方数据科技有限公司 一种用于智能制造的自动化分拣系统及其工作方法
CN111915604A (zh) * 2020-08-20 2020-11-10 魏小燕 一种互联网人工智能电子零配件识别与检测系统
CN112170233B (zh) * 2020-09-01 2021-08-06 燕山大学 基于深度学习的小型零件分拣方法及其分拣系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105158268A (zh) * 2015-09-21 2015-12-16 武汉理工大学 精冲零部件缺陷智能在线检测方法、系统及装置
CN105718836A (zh) * 2016-01-20 2016-06-29 深圳市黑云信息技术有限公司 一种标签自动喷码、错漏检测的控制方法
CN106767416A (zh) * 2017-01-04 2017-05-31 纵科（武汉）信息技术有限公司 零件检测系统及方法
US20210056681A1 (en) * 2018-05-03 2021-02-25 Inspekto A.M.V. Ltd. System and method for visual production line inspection of different production items
CN110148106A (zh) * 2019-01-18 2019-08-20 华晨宝马汽车有限公司 一种利用深度学习模型检测物体表面缺陷的系统及方法
CN110239997A (zh) * 2019-06-13 2019-09-17 华中科技大学 一种复卷机自动剔废方法及设备

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117218097A (zh) * 2023-09-23 2023-12-12 宁波江北骏欣密封件有限公司 一种轴套类丝网垫圈零件表面缺陷检测方法及装置
CN117218097B (zh) * 2023-09-23 2024-04-12 宁波江北骏欣密封件有限公司 一种轴套类丝网垫圈零件表面缺陷检测方法及装置

Also Published As

Publication number Publication date
CN113066087A (zh) 2021-07-02
CN113066087B (zh) 2022-08-05

Similar Documents

Publication Publication Date Title
CN109785316B (zh) 一种芯片表观缺陷检测方法
CN107545239B (zh) 一种基于车牌识别与车辆特征匹配的套牌检测方法
CN109550712A (zh) 一种化纤丝尾丝外观缺陷检测系统及方法
US10062008B2 (en) Image based object classification
WO2022227103A1 (zh) 一种零部件智能检测与标签系统及方法
US9396404B2 (en) Robust industrial optical character recognition
CN111650220A (zh) 一种基于视觉的图文缺陷检测方法
US9558403B2 (en) Chemical structure recognition tool
WO2020248513A1 (zh) 用于综合性能试验的ocr识别方法
JPH0726836B2 (ja) テキスト方位決定方法
US11657599B2 (en) Method for detecting appearance of six sides of chip multi-layer ceramic capacitor based on artificial intelligence
CN104881665A (zh) 一种芯片字符识别与校验方法及装置
CN108710876A (zh) 一种基于机器视觉的电池表面标识缺陷检测方法及系统
CN110880175A (zh) 一种焊点缺陷检测方法、系统以及设备
Szymanski et al. Automated PCB inspection in small series production based on SIFT algorithm
CN112419260A (zh) 一种pcb文字区域缺陷检测方法
CN114266764A (zh) 一种印刷标签的字符完整性检测方法及其装置
CN112001200A (zh) 识别码识别方法、装置、设备、存储介质和系统
CN115775246A (zh) 一种pcb元器件缺陷检测的方法
CN113436180A (zh) 生产线上喷射码的检测方法、装置、系统、设备和介质
CN115797945A (zh) 一种文本识别对比与图像对比相结合的标签检测方法
CN114332983A (zh) 人脸图像清晰度检测方法、装置、电子设备、及介质
CN110852994B (zh) 一种喷印卡片的视觉检测方法及其系统
CN109060831B (zh) 一种基于底板拟合的自动脏污检测方法
WO2017091060A1 (en) A system and method for detecting objects from image

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21938576

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21938576

Country of ref document: EP

Kind code of ref document: A1