Software technologies

Application examples of program technology in industry

Table of Contents

Surface inspection

Various types of vision tests can be performed at high speed on the surface of the product or on the surface of the package. For example, it automatically detects the presence of holes, other materials, scratches, wrinkles, cracks, and dirt on the surface of the coating.

Figure) Recognizing defects in pills packaging materials
Detect metal surface scratches

The biggest difficulty of this task is the uneven background and faint scrat

Picture a) Original image
Figure b) extracted scratch
Figure c) extracted scratch
Figure a) Very fuzzy defect image
b) Detected defect
Figure) Detection of protruding faulty condition and boundary line indication

Blob Analysis

Recognize blob images with very difficult conditions

Figure a) original image | b) detected small particles

Figure) Automatic high-speed detection of hexagonal crystals only once at a time

A) Original image (arrows show hexagonal crystals to look for)
B) Result of applying blob analysis algorithm
C) By adding additional image processing, only hexagonal crystals are detected

Figure a) original image | b) detected small particles

Shape recognition

Inspection of good products and defective products through matching|

Even if the location of the inspection product is changed, multiple products are mixed in a single image, some are invisible, some are changed in shape, the subpixel accuracy is extremely fast and overwhelming. And to measure a specific location.
For vision inspection, you can use your own CAD model to automatically identify the shape of the product to be inspected, find a specific position within the product, and display the measured value. This unique vision inspection technology opens up various possibilities such as moving the robot to a recognized position.

Figure a) Set the area of interest
Figure b) Mixed products of the same kind
Figure c) Automatic detection and measurement of even partially visible objects by pattern matching algorithm
Coin distinction

This task is to distinguish between different types of coins depending on the country of origin.

Picture) Different coins with different nationalities are classified by country of origin & display nationality
PCB component location recognition

This is to find the various parts on the printed circuit board in one step. There are usually several different parts mounted on a printed circuit board, and the location of these parts is different due to the error in the mounting process. Component-based matching is used to locate all parts in a fast and robust manner.

Figure A) Training of parts (registration of interest area setting)
Picture b) Find the set parts located in different images
Figure: Detecting even poorly focused parts
Figure) Detecting lids in uneven lighting conditions
Picture a) Normal mesh
b) Detecting fiber foreign matter

Classification of inspection type

Each specimen has color, texture, size or specific shape.
If you register each type of various inspection items in advance, the program can classify the same kind.

Picture) Training images
Picture) Result: Original image and three kinds of metal parts classified
Picture) Separate orange and lemon
Picture) Classification of good, bad, no halogen bulb
Figure: Detecting the color of fuses with different colors
Figure: Detecting the color of fuses with different colors under poor lighting
Figure) Detecting colors of different colored game pieces placed on a red background
Figure) Multiple switches are detected, and the on / off state of the switch is identified.

3D vision inspection

3D Vision Inspection is a 3D image processing technology that uses 3D information with the help of machine vision to approach challenges that traditional 2D technology could not solve.

The precise 3D calibration of the camera determines the horizontal, vertical and height information in 3D space, and the 3D object model obtained in various ways changes the location and direction of the inspection product, Whether mixed, partially visible, or partly shaped, you can use the extremely fast and overwhelming shape recognition technology with subpixel accuracy to spot test products and measure specific locations.

Figure) 3D object model obtained from CAD model

For 3D vision inspection, you can use your own CAD model to automatically identify the 3D shape of the product to be inspected, find a specific position in the product, and display the measured value. This unique 3D vision inspection technology opens up various possibilities such as moving the robot to a known position.

Figure a) Original image for testing b) 3D model image in different directions and positions
Figure a) 3D object model obtained through 3D image sensor (X, Y, Z)
Figure b) 3D individual parts model and measurement results obtained by image processing to remove the background

The task below is to find the 3D clamp shown in the figure and to determine the pose of the clamp to determine its position and orientation in space

Picture a) Original image with clamp
Picture b) Display the 3D model of the clamp found in the image and its pose
Figure: Detecting pipe joints that are scrambled and partially invisible
Figure) Detect specific parts of engine parts placed in different directions and recognize the angle
Figure: Find and sort related parts

Bar code, data code recognition

Our outstanding character recognition technology automatically checks multiple barcodes at once and reads them at amazing speed, no matter where the barcode is located.
Especially when the barcode is extremely thin or less than 1.5 pixels, the barcode can be recognized even if only 5% of the barcode is visible, or only 95% of the barcode is exposed.
ECC 200, QR, Micro QR, Aztec, and PDF417 codes, as well as data codes in distorted form can be read. You can also read “Direct Part Mark” (DPM) codes printed on different surfaces in different lighting situations.

The picture below shows an image of a CD, the barcode on the top of this CD is printed in a circular shape, and this operation is to read this circular barcode. Since bar code readers can not directly read this kind of print, the image must first be transformed into the same elements as the parallel bar code.

Figure a) Circular shaped bar code engraved on CD
Picture b) Horizontal alignment with decoded bar code
Picture) Many bar codes are recognized at once
Picture) Many bar codes are recognized at once

2D data code recognition

We provide a way to read 2D data codes such as portable data format 417 (PDF417), matrix ECC 200, QR code, micro QR code, and Aztec code.
2D data codes (2D barcodes) are used in various fields. Similar to one-dimensional barcodes, characters are composed of graphics in black, white, bar or dot.

Figure) Recognize 2D data codes that look fuzzy under very poor lighting
Figure) Recognize obliquely placed data codes
Figure) Recognize noisy data codes
Figure) Two of the six data codes are recognized on crumpled paper
Figure) PDF417 Code Recognition

Optical character reading device (OCR)

Our exceptionally powerful classification technology can categorize and prove your company’s fonts.

Read form

This is to read and extract the symbols in the form. A typical problem is that the symbols are not printed in the correct location, as shown in the figure.

Figure) Recognize forms that are not printed at the correct location
Figure) Recognize the characters printed on the bottom of the can
Figure) Recognize non-formal characters
Figure) Recognize non-formal characters
Figure: Reading a dot print on a complex background
Figure) Recognize character forms commonly used in the pharmaceutical industry
Figure) recognizes printed characters on semiconductor chips
Figure) Simultaneously recognize the character and switch status of two overlapping switches

Print Inspection

Figure: Checking the printed logo
Figure) Identification of printed characters on pill packaging
Figure) Recognize printed characters in a crumpled sweets bag

Package inspection (2D & 3D)

It is possible to classify the condition of the whole package, the abnormality and the product by the excellent vision inspection method.

Figure) Recognizing the glass bottle in the box
Picture) Recognize printed letters in crumpled confection envelopes

2D measurement

Machine vision algorithms provided by NextAOI provide automatic detection measurements with very stable and high precision of subpixels.

Figure a) Pixel Precision Edge Detection Figure
b) Subpixel Precision Edge Detection
Figure) Interval measurement with subpixel accuracy
Figure) Automatic measurement of switch pin width and pin-to-pin distance
Figure) Automatic measurement of the distance between IC leads and lead
Figure) Length of lead on both sides
Figure a) Original image Figure
b) Detect all existing radius
Picture) Detecting three circles in the gear wheel
Figure) Minimum and maximum distance measurement of screw threads
Figure) Automatically extracts contour lines even when position changes, automatically measuring lines and circles
Figure) Locate the drill hole and measure the diameter
Figure) Automatically extracts the outline of an object with sub-pixel accuracy and then separates lines and ellipses
Figure) Extract only the outline of the desired part with sub-pixel accuracy
Picture) Inspect all razor blades to detect and display missing parts
Figure) Measure width and spacing of inspection object at various positions at once
Figure) Measure the circles and lengths of the objects in various positions at once

www.nextaoi.com
sales@nextaoi.com

Copyright©NEXTAOI Inc. All rights reserved.