Some Application Cases

2.5D / 3D Phone Cover Glass

Surface detection

Speed:≤ 2.3s/pc

Result:Loss ≤ 2.5% | Fault ≤ 10%

Type:Smudge, hair, dust spot, scratch, edge collapse, bump, deep scratch......

Defect Type:deep scratch
  • master map

  • result

Metallized Ceramics

Surface detection

Speed:≤ 100ms/pc

Result:Loss ≤ 0.3% | Fault ≤ 1%

Type:Scratches, chipped edges, smudges, cracks, scratches, missing......

Defect Type:scratch
  • master map

  • result

Metallic Part

Surface detection

Speed:≤ 500ms/pc

Result:Loss ≤ 0.1% | Fault ≤ 5%

Type:Smudge, bright mark, scratch, spot, mold, knife grain, notch......

Type:Burr、scratch
  • master map

  • result

Steel plate detection

Surface detection

Speed:≥120pcs/min

Result:Loss ≤ 0.1% | Fault ≤ 5%

Type:Burr, scratch, bruise......

Type:Burr、scratch
  • master map

  • result

Industrial character recognition

Surface detection

Speed:≤ 120ms/pc

Result:Precision rate ≥ 99.99%

Defect Type:Capital
  • master map

  • result

The finished jelly

Surface detection

Speed:≥100pcs/min

Result:Loss ≤ 0.2% | Fault ≤ 5%

Type:Hair, pulp insects, black residue, air bubbles, iron filings......

Defect Type:Hair
  • master map

  • result

Face detection

Surface detection

Speed:≥235pcs/min

Result:Loss ≤ 1% | Fault ≤ 1%

Type:Singe, foreign body, hair, dirty......

Defect Type:Singe
  • master map

  • result

Medicine plate detection

Surface detection

Speed:≥500pcs/min

Result:Loss ≤ 0.001% | Fault ≤ 0.001%

Type:Deformation、foreign body、line error、blow-by......

Defect Type:Deformation
  • master map

  • result

Medicine boxes

Surface detection

Speed:≥500pcs/min

Result:Loss ≤ 0.001% | Fault ≤ 0.001%

Type:Damaged,smudged,medicine box opening,deformation,label curling......

Defect Type:deformation
  • master map

  • result

OCR test of medicine plate and box

Surface detection

Speed:≥500pcs/min

Result:Precision rate ≥ 99.99%

Defect Type:character
  • master map

  • result

Electronic Components

Surface detection

Speed:≤ 2ms/pc

Result:Loss ≤ 0.01% | Fault ≤ 1%

Type:Magnetic leakage,bubbles,cracking,deformation,black,crack,crescent,trachoma,indentation......

Defect Type:End of bad
  • master map

  • result

Pouch power cell

Surface detection

Speed:≤ 5s/pc

Result:Loss ≤ 0.1%(serious defects is 0) | Fault ≤ 5%

Type:Polar piece folding, breakage, leakage, edge sealing foreign body,convex, pinhole......

Defect Type:Sealing foreign body
  • master map

  • result

Wet magnetic tile

Surface detection

Speed:≥80pcs/min

Result:Loss ≤ 0.1% | Fault ≤ 15%

Type:Crack, collapse, air hole, bad grinding, bad chamfering......

Type:Falling blocks, leakage of wear
  • master map

  • result

Dry magnetic tile

Surface detection

Speed:≥120pcs/min

Result:Loss ≤ 0.1% | Fault ≤ 10%

Type:Crack, collapse, trachoma......

Defect Type:Crack
  • master map

  • result

Ferrite bead

Surface detection

Speed:≥120pcs/min

Result:Loss ≤ 0.1% | Fault ≤ 10%

Type:Crack, collapse, plug hole, burr......

Defect Type:Crack
  • master map

  • result

Sintered magnet

Surface detection

Speed:≥100pcs/min

Result:Loss ≤ 0.1% | Fault ≤ 10%

Type:Crack, corner drop, pitting, knife mark, hole, inclusions, corrosion.....

Defect Type:Crack
  • master map

  • result

Soft magnetic core

Surface detection

Speed:≥200pcs/min

Result:Loss ≤ 0.1% | Fault ≤ 10%

Type:Crack, corner drop, speckle, mucous membrane, slurry, abrasion......

Defect Type:Crack
  • master map

  • result

Bonded magnetic ring

Surface detection

Speed:≥100pcs/min

Result:Loss ≤ 0.05% | Fault ≤ 10%

Type:Cracks, defects, pitting, magnetic flux leakage, pits, foreign bodies......

Type:Pitting, bump
  • master map

  • result

 FAQ 
How many image data need to be learned for each defect category?
The number of images to be learned will vary according to the complexity of the images, but in the initial stage, about 40-100 images for different defect categories can be provided。
How long does it take to build the neural network model?
The general image size of 2048*2048 pixels, 500 for the standard, about 30 minutes。
Industrial AI vision kit installation, need to change production line?
Can be directly installed to the original production line, generally do not need to extend the additional major production line。
What is the recognition speed that can be achieved in the actual production line?
For example, the inspection project of medicine box packaging can detect about 500 products in a minute through full inspection,the specific situation of each project is different, so please take the actual project situation as the criterion。
Can multiple different defects be identified?Is there an upper limit on the number of identifications?
Supports multiple different categories of defects in the same project, with no upper limit on the number of categories detected。
Whether the software is a standardized product?
Products are standardized, with a strong versatility.At the same time, according to the special needs of users in different industries, technical personnel can customize the program according to the needs on the basis of the product。
Is the image sample formatted?
Image format is not limited, basically covering all image formats.Such as JPEG, BMP, PNG and so on。