A component quantity identification and mixed batch detection system based on image recognition

By using image recognition technology and machine learning, combined with vibration table and other equipment, automatic detection of component quantity and batch information is achieved, and the problems of low efficiency and poor reliability in component supplementary screening are solved, and detection efficiency and reliability are improved sex.

CN114973223BActive Publication Date: 2024-11-05Shanghai Institute of Basic Aerospace Technology
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210204850.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-03
Publication Date
2024-11-05
Estimated Expiration
2042-03-03

AI Technical Summary

Technical Problem

During the component replenishment screening process of component reliability laboratory, the confirmation of component quantity, type, and batch information needs to be done manually, resulting in low sorting efficiency and poor reliability, and it is difficult to quickly and accurately observe the surface batch information, which is costly. A lot of time and energy.

Method used

Using image recognition-based component quantity recognition and mixed batch detection system, using machine learning, OCR recognition, image preprocessing and image feature extraction technology, component outline features are obtained through cameras, real-time quantity inventory and batch information recognition, combined with Vibrating table, transmission belt, glass turntable and other equipment realizes automatic inspection.

Benefits of technology

It greatly reduces the participation in manual sorting, improves detection efficiency and reliability, realizes the rapid and accurate identification of component quantity and batch information, and solves the problems of large workload, long cycle and low reliability of manual sorting.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114973223B_ABST
    Figure CN114973223B_ABST
Patent Text Reader

Abstract

The present invention discloses a component quantity recognition and mixed batch detection system based on image recognition, including: a vibration table, a conveyor belt, a height limiter, a glass turntable, a camera, a material box, a feeding push rod, a microphone, a power switch, and an industrial computer; the user places the components into the vibration table, and the vibration table gradually vibrates the components into the conveyor belt through vibration, and the conveyor belt and the glass turntable step synchronously; the components are transmitted to the glass turntable through transmission movement; a height limiter is installed on the conveyor belt, and the height limiter can restrict the components from being stacked or placed vertically for a second time; after the components enter the glass turntable, the upper and lower cameras shoot the surface of the components and obtain the component information; when it is found that the batch information of the components under the same frame is inconsistent, the industrial computer marks the wrong components; after the detection is completed, the components on the glass plate are scanned into the material box by the feeding push rod. The present invention solves the problems of large workload, long sorting cycle, and reduced reliability in the current manual sorting of basic information of components.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the application field of component detection engineering, and in particular to a component quantity recognition and mixed batch detection system based on image recognition, which can automatically detect the number of indefinite components and whether they are mixed batches when the number, type and batch of components are uncertain. Background Art

[0002] The quality assurance of components is the most basic and core quality assurance link in the field of quality assurance of electronic components for military equipment. This quality assurance link is the ability of components to complete the specified functions under specified conditions and within a specified time. Its reliability consists of its inherent reliability and operational reliability. The former is mainly determined by the quality control of component production units in the process of component design, process, and selection of raw materials; the latter is mainly determined by the quality control of component selection, procurement, and use by component users. Among them, military components all need to be supplemented by the component reliability laboratory. The purpose of supplementary screening is to eliminate products that fail early and ensure the reliability of components after installation.

[0003] In the process of supplementary screening of components, the component reliability laboratory first needs to establish screening tasks to ensure that the quantity, component type, batch and screening task information of each task are consistent. Since the components used in different equipment are very different, the components present the characteristics of multiple varieties and small batches. Confirming these basic information is done manually. Due to the small size of components, the batch information printed on the surface is difficult to observe quickly and accurately with the naked eye. Some information can only be accurately read under a magnifying glass. Manual visual sorting requires not only counting the quantity, but also counting whether the components in the task are from the same batch. Different batches need to be eliminated. Surface batch identification consumes a lot of time and energy. Component reliability laboratories generally need to test tens of thousands of batches of different types of components every year. The efficiency of component task quantity sorting and surface batch information mixed batch elimination is low, and the reliability is poor. Summary of the invention

[0004] The purpose of the present invention is to provide a component quantity recognition and mixed batch detection system based on image recognition, which takes machine learning, OCR recognition, image preprocessing, and image feature extraction technology as core technologies, and can be applied to most types of component quantity detection and batch information recognition equipment to solve the problems of low efficiency and poor reliability of manual sorting.

[0005] The image feature extraction technology is mainly used to obtain the contour features of multiple components under test within the camera frame. The image is processed by image preprocessing technology, mainly including the removal of redundant background features and the binarization of component surface information. Only the surface and appearance contour features are retained. The machine learning method of component samples can distinguish different types of components in real time through machine learning technology, and count the number of recognized appearance contours in real time. The OCR technology is used to extract the features of component surface information again, identify the surface information text, and obtain batch information in real time.

[0006] In order to achieve the above-mentioned purpose, the present invention provides a component quantity recognition and mixed batch detection system based on image recognition, including: a vibration table, a conveyor belt, a height limiter, a glass turntable, a camera, a material box, a feeding push rod, a microphone, a power switch, and an industrial computer;

[0007] The industrial computer is used to control the start, stop and movement of the system, and to check the detection status of components;

[0008] The microphone is used by the user to perform voice command control to control the start, stop and movement of the system;

[0009] The power switch is used to control the start, stop and movement of the system;

[0010] The user places the components onto the vibration table, which gradually vibrates the components into the conveyor belt through vibration, and the conveyor belt and the glass turntable step synchronously; the components are transferred to the glass turntable through transmission movement; a height limiter is installed on the conveyor belt, which can restrict the components from being stacked or placed vertically for a second time, ensuring that the components passing through the conveyor belt are laid flat and the surface can be completely presented on the glass turntable; after the components enter the glass turntable, the upper and lower cameras shoot the surface of the components and obtain the component information; when it is found that the batch information of the components in the same frame is inconsistent, the industrial computer marks the wrong component with a red outline frame; when the inspection is completed, the components on the glass plate are swept into the material box by the feeding push rod.

[0011] The above-mentioned component quantity recognition and mixed batch detection system based on image recognition also includes a photoelectric sensor for identifying whether there are any residual components on the glass turntable.

[0012] In the above-mentioned component quantity identification and mixed batch detection system based on image recognition, the camera has a zoom function.

[0013] The above-mentioned component quantity identification and mixed batch detection system based on image recognition, wherein the component information includes: surface information, quantity information, and batch information.

[0014] The machine learning method of component quantity recognition and mixed batch detection system based on image recognition includes the following steps:

[0015] Step 1: Place components: Place the components to be inspected or pre-learned on the glass table. The components can be laid flat at will, but not stacked.

[0016] Step 2: Image contour recognition: Use the camera to capture the components placed on the glass table, and perform minimum circumscribed rectangle segmentation on the components, segment the component image into minimum circumscribed rectangles, and calculate the perimeter of the circumscribed rectangle;

[0017] Step 3: Parameter configuration: manually configure the zoom range of the corresponding camera through the perimeter of the outer rectangle, and record the relevant parameters so that the camera can clearly capture the text on the surface of the component;

[0018] Step 4: Feature annotation: Annotate the classification and specifications of components as materials for machine learning;

[0019] Step 5: Machine learning: All the annotated images and annotation information are subjected to machine learning, and the corresponding image recognition model is output through machine learning. The model can output the type and specification information of components;

[0020] Step 6. Parameter application: According to the rectangular perimeter of the component, the corresponding camera focal length, exposure, and parameters are automatically configured so that the camera can automatically identify components of different sizes.

[0021] The detection method using component quantity recognition and mixed batch detection system based on image recognition includes the following steps:

[0022] Step 1: Place components: Place the components to be inspected or pre-learned on the glass table. The components can be laid flat at will, but not stacked.

[0023] Step 2: Image contour recognition: Use the camera to capture the components placed on the glass table, and perform minimum circumscribed rectangle segmentation on the components, segment the component image into minimum circumscribed rectangles, and calculate the perimeter of the circumscribed rectangle;

[0024] Step 3: Machine vision recognition: Use machine learning models to identify the types and specifications of all components within the camera frame.

[0025] Step 4: Component OCR recognition: Use OCR technology to identify the batch text information on the surface of the components, and compare whether the batch information of the components within the frame is consistent;

[0026] Step 5: Label the recognition results: Use rectangular boxes of different colors to represent the recognition results.

[0027] The above-mentioned component quantity identification and mixed batch detection method based on image recognition, wherein, in the step five, the identification results are represented by rectangular boxes in three colors: green, red and blue. Blue means that no valid surface information is collected, green is the surface information of most components, and red is the surface information of inconsistent components.

[0028] Compared with the prior art, the technical benefits of the present invention are:

[0029] This patent proposes a component quantity identification and mixed batch detection system with low modification cost, high reliability and strong versatility, which greatly reduces the participation of sorting personnel in the sorting process, realizes automatic sorting and mixed batch prompts. It solves the current problems of large workload, long sorting cycle and reduced reliability in manual sorting of basic information of components. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] A component quantity recognition and mixed batch detection system based on image recognition of the present invention is provided by the following embodiments and drawings.

[0031] Figure 1 It is a component quantity identification and mixed batch detection system based on image recognition. DETAILED DESCRIPTION

[0032] The following will further describe in detail a component quantity recognition and mixed batch detection system based on image recognition of the present invention in conjunction with the accompanying drawings.

[0033] Step 1: Open the detection software and self-test

[0034] Turn on the power supply and device terminal, power on the device, and open the system control software to check whether the device hardware is normal. If there is no abnormality, the system enters standby mode.

[0035] Step 2: Machine learning to create a sample template

[0036] The pre-inspection personnel places a sample to be tested on the template making table and clicks the "Make Template" button. The camera above the template making table starts to detect the sample to obtain the appearance size, component type, and surface batch information. The pre-inspection personnel clicks "Confirm Template" to obtain the template information as the basis for this inspection. A machine learning data set is constructed and machine learning is performed to finally obtain a usable machine learning model for sample recognition.

[0037] Step 3: Component Inspection

[0038] The pre-inspection personnel clicks the "Start Inspection" button, and the equipment vibration plate and the transmission mechanism glass inspection plate start to move. The vibration plate shakes the components apart and spreads them on the height-limited conveyor belt. According to the height of the components of the template, the height-limited conveyor belt adjusts the corresponding height railing to limit the height of the components passing through the conveyor belt, further ensuring that the components are laid flat, and the components passing through the height-limited conveyor belt are transmitted to the glass turntable. After the software-limited movement step distance, the three transmission mechanisms pause.

[0039] Step 4: Batch detection and identification

[0040] There is a set of cameras above and below the glass turntable. When the three transmission mechanisms stop, the cameras start shooting. The component object recognition module recognizes the number of components in the frame and the surface information of each component in real time. If the information is consistent with the template (either the front or the back), a green frame is marked, and if it is inconsistent, a red frame is marked. The pre-inspection personnel make manual judgments through the software detection interface and give the final result. After completion, click the "Continue" button. The equipment pushes all components on the glass tray into the material box through the feeder. And transfer the next batch of components to the glass tray for identification. The component object recognition module will accumulate the recognition results and the number of recognitions until the detection of this batch of components is completed.

[0041] Step 5: Detection completed

[0042] The pre-inspection personnel clicks the "Finish" button to end the inspection task, and the software uploads the inspection result information to the management system.

[0043] Step 6: Turn off the device

[0044] The pre-inspection personnel can shut down the system using the shutdown button in the software or the physical button on the device. In either case, the device will automatically power off after a hardware reset.

Claims

1. A component quantity identification and mixed batch detection system based on image recognition, characterized in that: include: Vibration table, conveyor belt, height limiter, glass turntable, camera, material box, feeding push rod, microphone, power switch, industrial computer; The industrial computer is used to control the start, stop and movement of the system, and to check the detection status of components; The microphone is used by the user to perform voice command control to control the start, stop and movement of the system; The power switch is used to control the start, stop and movement of the system; The user places the components on the vibration table, which gradually vibrates the components into the conveyor belt through vibration. The conveyor belt and the glass turntable step synchronously; the components are transferred to the glass turntable through transmission movement; a height limiter is installed on the conveyor belt, which can restrict the components from being stacked or placed vertically, ensuring that the components passing through the conveyor belt are laid flat and the surface can be completely presented on the glass turntable; after the components enter the glass turntable, the upper and lower cameras shoot the surface of the components and obtain the component information; when the batch information of the components in the same frame is found to be inconsistent, the industrial computer marks the wrong components; after the inspection is completed, the components on the glass plate are swept into the material box by the feeding push rod; The component quantity recognition and mixed batch detection system based on image recognition performs the following machine learning method: Step 1: Place components: Place the components to be inspected or pre-learned on the glass table. The components can be laid flat at will, but not stacked. Step 2: Image contour recognition: Use the camera to capture the components placed on the glass table, and perform minimum circumscribed rectangle segmentation on the components, segment the component image into minimum circumscribed rectangles, and calculate the perimeter of the circumscribed rectangle; Step 3: Parameter configuration: manually configure the zoom range of the corresponding camera through the perimeter of the circumscribed rectangle, and record the relevant parameters so that the camera can clearly capture the text on the surface of the component; Step 4: Feature annotation: Annotate the classification and specifications of components as materials for machine learning; Step 5: Machine learning: All the annotated images and annotation information are subjected to machine learning, and the corresponding image recognition model is output through machine learning. The model can output the type and specification information of components; Step 6. Parameter application: According to the rectangular perimeter of the component, the corresponding camera focal length, exposure, and parameters are automatically configured so that the camera can automatically identify components of different sizes.

2. A component quantity identification and mixed batch detection system based on image recognition as claimed in claim 1, characterized in that: It also includes a photoelectric sensor for identifying whether there are any remaining components on the glass turntable.

3. A component quantity identification and mixed batch detection system based on image recognition as claimed in claim 1, characterized in that: The camera has a zoom function.

4. A component quantity identification and mixed batch detection system based on image recognition as claimed in claim 1, characterized in that: The component information includes: surface information, quantity information, and batch information.

5. A detection method for component quantity recognition and mixed batch detection system based on image recognition as described in any one of claims 1 to 4, characterized in that: The steps include: 1) Place components: Place the components to be inspected or pre-learned on the glass table. The components can be laid out flatly but not stacked. 2) Image contour recognition: Use the camera to capture the components placed on the glass table, and perform minimum circumscribed rectangle segmentation on the components. Segment the component image into minimum circumscribed rectangles and calculate the perimeter of the circumscribed rectangle. 3) Machine vision recognition: Through machine learning models, the corresponding types and specifications of all components within the camera frame are identified; 4) Component OCR recognition: Use OCR technology to identify the batch text information on the surface of components and compare whether the batch information of components within the frame is consistent; 5) Recognition result labeling: The recognition results are represented by rectangular boxes of different colors.

6. A detection method for component quantity recognition and mixed batch detection system based on image recognition as claimed in claim 5, characterized in that: In the step 5), the results of the recognition are represented by rectangular boxes in green, red and blue colors. Blue means that no valid surface information is collected, green is the surface information of most components, and red is the surface information of inconsistent components.