Mobile Phone Manufacturing

2.5D White Glass Inspection

Detect accuracy : 0.78um (size larger than 8200*12800 pixels).

Detect speed : <= 2S, for more than 5 channels

False negative rate: <=5%; missed detection rate: <=3%

Defect types: dent, scratches, hair, dust spots, indentations, chipping, dirt, water drops, water ripples, hole deformation, etc.

Dent

Dent detection

Scratch

Scratch detection

Impress

Impress detection

FPC board appearance defect detection

Detect accuracy : 4 pixel levels (can be adjusted according to different resolution)

Detect speed : <=430ms

Killing rate : <=5%; Missing rate : <=1%

Defect types: scratches, crushes, different colors, oil spots, etc.

Scratch

Scratch detection

Crush

Crush detection

Blush

Blush detection

Tobacco Production

Tobacco leaf grade classification

Detection speed: <=60ms/1 piece (resolution > 2048*1200 pixels)

Classification accuracy: leaf front / back classification accuracy >=99%; leaf parts classification accuracy >=85%; Good / Bad / Mixed leaf classification accuracy >=95%; Four-grade classification accuracy >=75%

Classification categories: leaf front and back, leaf parts (top, middle, and bottom), leaf grade (four levels).

Grenn

Green detection

Mix

Mix detection

Front

Front detection

Government

AI visual inspection and analysis for highway pavement


  • Highway pavement

  • Sample picture

  • Detection picture
Food Production

High-speed detection of food impurities

Detection accuracy: <=50um (1600*1400 pixels with color)

Detect speed : <=600ms with 5 channels

False negative rate: <=5%; miss detection rate: <=5%

Defect types: hair, black slag, fruit worms, iron filings, paint, air bubbles, scale (as an example of jelly impurity detection).

hair

hair detection

black slag

black slag detection

fruit worm

fruit worm detection

Agricultural Products Production

Apple classification inspection


  • Apple picture

  • Apple color classification

  • Apple size classification
Green New Energy

Garbage classification inspection


  • Construction garbage

  • Life garbage

  • Medical garbage
Deep Learning OCR

Deep learning OCR

Detection speed: <=50ms/each, may vary with the number of detected characters.

Classification precision: multi-row characters >= 99.9% per product.

Classification: numbers, letters, special symbols, other...

print

print detection

print

print detection

chip coding recognition

parts steel stamping

Small and Micro Component Inspection

Hexahedral part surface defect detection

Detection precision: 4 pixels (1920*1200Pix), adjustable according to different resolutions

Detection speed: <=80ms/each, combined processing of 13 pictures

False negative rate: <=6%; miss detection rate: <=2%

Defects: resin, scratches, dirt, breakage, inductor winding defects, pad defects, grease defects...

resin

resin detection

dirty

dirty detection

damage

damage detection

3C metal parts surface defect detection

Detect precision : 4 pixels(39.6875μm)

Detection speed: <=50ms / map (for size of 2560 * 2048)

Strip light spot leak detection <=1.11%, false negative rate <=2.5%; bowl light scratch detection <=0.28%, false negative rate <=1.11%; bright print dirty leak detection <=0.56%, false negative rate < =0.56%

Defects: spot, scratch, dirt, bright print, stamp, gap...

spot damage

spot damage detection

scratch

scratch detection

dirty

dirty detection

Silver plated bracket defect detection

Image size: 4096*15000 pixels

Detect speed : <=790ms

Accuracy : >=96%

Typical defects: scratches, silver, dirt, deformation, black spots...

scraping

scraping detection

silver coarse

silver coarse detection

out of shape

out of shape detection

3C components solder joint defect detection

Detect speed : <=350ms

False negative rate: <=5%; miss detection rate: <=2%

Defects: deep pool, breakdown, splash (take the mobile phone vibration motor solder joint as an example).

puncture

puncture detection

splash

splash detection

swelling

swelling detection