The task of object detection is to find interesting objects in images or videos, and at the same time detect their location and size. This is one of the core issues in the field of machine vision.
With the booming development of industrial automation, many manufacturers will always appear in the production process of products
Bad, bad, deformity, fit deviation
Wait.
Sometimes when artificially cannot meet the requirements of the requirements, the positioning/alignment/opposition of the machine vision system can be successfully completed, which can successfully complete the high -precision manufacturing process.
However, there are many uncertain factors during the detection process of objects. Difficulty.
If there is a problem with accuracy, it will lead to the emergence of bad products. Once detection and control are stagnant, it will reduce productivity.
How to achieve fast and accurate positioning has become an indispensable means of efficient production process.
Morning Smart Low Code Platform
Fusion deep learning algorithm, 2D/3D vision, SaaS tool and APAAS modular component. Complete the complete development tool chain of image collection, image labeling, algorithm development, algorithm packaging, and application integration. No programming is required. It supports deep learning model training closed loop, which will increase development efficiency by more than 10 times. Differentiated needs in business scenarios helped the intelligent upgrade of the industry.
Overview of target positioning module
It is mainly used for positioning and classification of different categories and locations in the picture, as well as positioning and detection of different defect characteristics in the picture. It is commonly used to classify, position, count on a variety of workpieces, and position and classify a variety of defects in the same workpiece. It is the most widely used defect detection tool.
Features:
和 Training speed and reasoning speed are fast. After deep learning training, it can solve the effects of translation, rotation, zooming and light.
目 Support multiple targets in the image simultaneous detection, giving all the positions and definitions of all recognition targets
形 The form of rectangular diagram of tools is flexible and convenient.
Application scenario:
标 irregular shape target detection
❇ Small target detection
Case analysis: connecter defect detection
● Detective requirements
To detect the surface defect of the connector, the number of PIN needles is 25 × 2. You need to detect that the PIN needle is not deformed and deformed.
● Difficulties in detection
The number of PIN needles, the variable flawed form, and the background of the detection area is complicated.
● Solution
Using Morning’s Smart’s target positioning function, build a deep -learning connector defect recognition model, provide non -contact detection, accurately present important information such as deformation and partiality of the connector PIN needle, timely find the connector defect, effectively remove the effect, effectively remove it unqualified product.
● Specific steps
1. Create a “target positioning” project
2. Upload pictures
Based on uploading the picture, the defects are induced and classified and labeling, and the corresponding AI processing scheme is given to different defect categories;
3. Mark
Use the labeling tool to mark the type of defect according to the label;
4. Model training
Using [Training] and [Test] function modules in the smart low code platform to enter the deep learning phase to achieve testing needs.
5. Model verification
01
Before using VS
Row No. 1 Fifth to the right is sloping to the right
02
The 11th row 11th is slightly sloping downwards
03
The first row on the right, the fifth to the bottom left is sloping
04
Row 1, 9th, lyph, No. 2, 8th, 9th, and 9th, upward, lyph
05
The second row, the first, the first to the upper left is sloping
06
The 13th row of the 13th to the upper orientation 2nd row, the 13th to the top
● The final effect
Using Morning Smart Low Code Platform, the defect recognition rate was low before, and it was easy to misunderstand.
After testing the test algorithm of the target positioning algorithm of the smart low code platform, the technical requirements are met.
The defect detection rate and accuracy can reach more than 99%.
In the application of connectors, Merguan Intelligent uses AI deep learning algorithm, effectively solves most of the defect detection needs, alleviates various problems existing in artificial quality inspection, improves the quality inspection efficiency of customers, and has received great recognition of customers.
Morning Smart Machine Vision Low Code Platform is a cloud collaborative development platform for machine visual applications. It always adheres to the product concept of 0 costs, 0 code, 0 threshold, and 0 hardware.
The platform takes artificial intelligence technology as the core. In the development of machine visual application development, developers provide image collection, image labeling, algorithm development, algorithm packaging and application integration. cover
Character recognition, defect detection, size measurement, target positioning
Waiting for hundreds of universal functions, it is committed to becoming the most widely used machine visual low -code platform in the world.
More functions are under development, and the latest information will be the first time
Official public account [Morning Smart]
Putting it up, everyone will continue to pay attention!
Before using VS
Before using VS
Before using VS
Before using VS
Before using VS