Similar part recognition system, medium and method based on deep learning and model simulation
A technology of model simulation and deep learning, applied in neural learning methods, design optimization/simulation, biological neural network models, etc., can solve problems such as difficulty in meeting actual needs, reduced accuracy of model recognition, time-consuming and labor-intensive problems, and achieve reduction Effect of labor costs and process response times
Pending Publication Date: 2022-03-22
JIANGSU UNIV OF SCI & TECH
View PDF1 Cites 3 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] However, the training set required by the deep learning model often contains tens of thousands of sample image information. Although the manual labeling method has the advantage of high accuracy, it is time-consuming and labor-intensive. At the same time, in industrial production, customization is often required. Multi-variety parts classification and recognition tasks, making a training set of all possible parts at one time will lead to a reduction in the accuracy of model recognition, which is difficult to meet the actual needs of industrial production
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0058] The technical solution of the present invention will be further described below in conjunction with the accompanying drawings.
[0059] like figure 1 As shown, the present invention provides a similar part recognition based on deep learning and three-dimensional model simulation:
[0060] Step 1: Use the Bullet physics engine to simulate the free fall process of rigid body parts, so as to obtain the static pose of the 3D model of the part to be recognized in the virtual space;
[0061] Step 2: Use the OpenGL library to simulate the static posture of parts observed from different viewing angles and light, and save the simulation images of all parts under different viewing angles;
[0062] Step 3: Using the color distinction between the part and the surrounding environment in the simulation image, use the OpenCV library to obtain the minimum bounding box coordinates of the part in the simulation image;
[0063] Step 4: Use the simulation image, part name and bounding bo...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a similar part recognition system, medium and method based on deep learning and model simulation. The method comprises the steps of obtaining a static pose of a part three-dimensional model, generating a simulation image, obtaining a minimum bounding box coordinate of a part image, generating a similar part recognition network training set, carrying out YOLO4 neural network training, constructing an optimal recognition view angle set according to a test set result, and controlling a camera to collect a field part image. And judging whether the pose of the similar part under the camera visual angle is adjusted through the motion turntable according to the confidence of the recognition result, and then performing recognition again. According to the method, the training set of the YOLO4 algorithm is generated by using the simulation image of the three-dimensional model, self-generation of the training set is realized, so that the time for manufacturing a training sample is shortened, and the problem that distinguishable features of parts are easy to block under a single view angle is effectively solved through a motion turntable mode; experimental results show that the method has a high-precision recognition result and has a feasible practical value.
Description
technical field [0001] The present invention relates to the recognition of similar parts, in particular to a similar parts recognition system, medium and method based on deep learning and model simulation. Background technique [0002] In industrial production, there are often classification and identification requirements for multi-variety, multi-batch, and highly similar parts. For a long time, machine vision technology has attracted much attention. As an important part of machine vision, image classification and recognition technology has been continuously researched and developed. Not only has there been continuous improvement and innovation in theory, but it has also achieved significant promotion and breakthroughs in practical applications. For example, the patent document CN112132783A discloses a part recognition method based on digital image processing technology. At present, the application of machine vision technology for classification and sorting of target prod...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & AuthorityApplications(China)
IPC IPC(8): G06F30/17G06F30/27G06N3/04G06N3/08
CPCG06F30/17G06F30/27G06N3/08G06N3/045
Inventor张辉官震杨育陈瑶朱成顺方喜峰
OwnerJIANGSU UNIV OF SCI & TECH