Skip to content

Technology

The Technology Behind DOKTEK FoodTech

Machine vision, motion, process and engineering. Four disciplines, one company — and the reason a product that arrives in random orientation on a moving belt can still be cut, filled, rolled or sorted correctly.

Four pillars

One company, four disciplines.

A machine that adapts to a natural product needs all four at once. Buying them from four suppliers is how food automation projects stall.

  1. 01

    Vision

    Industrial cameras and deep-learning models establish what each product is and where its features sit before anything touches it.

    • Object detection
    • Keypoint detection
    • Position and orientation detection
    • Defect detection
    • Classification
  2. 02

    Motion

    Detected coordinates become machine coordinates, and the tool arrives where the product actually is while the belt keeps moving.

    • Servo motion
    • Conveyor synchronization
    • Robot coordination
    • Real-time positioning
  3. 03

    Process

    The food operation itself, engineered as a repeatable industrial step rather than a skilled manual gesture.

    • Cutting
    • Trimming
    • Rolling
    • Filling
    • Sorting
    • Inspection
  4. 04

    Engineering

    Stainless-steel machinery designed for wash-down production areas and for the line it joins.

    • Stainless-steel structures
    • Food-grade design principles
    • Cleanability
    • Modular machinery
    • Industrial controls
    • Line integration

Machine vision

The Machine Doesn't Just See the Eggplant.It Understands It.

Computer vision detects product geometry, analyses stem location and sends cutting coordinates to the automation system in real time.

  • OBJECT

    Product isolated from the belt

  • GEOMETRY

    Length, width and curvature

  • ORIENTATION

    Long axis measured in frame

  • KEYPOINTS

    Calyx and shoulder located

  • CUT LINE

    Computed for this product alone

  • TRANSFORM

    Image space into robot space

Coordinate transform
Image spaceRobot spaceu612 pxX428.4 mmv344 pxY216.7 mmθ+13.0°C82.0 mm

Interface illustration — values shown are the ones in the animation at the top of this page.

DOKTEK stem-detection interface showing an aubergine with its detection box, calyx keypoints and computed cut line
The real stem-detection interface during a cutting trial — detection box, keypoints and computed cut line.

Detect → Calculate → Cut

Every eggplant is different. The cut should be too.

Doktek FoodTech analyses each product individually instead of relying on a single fixed mechanical cutting position. Three things happen to every product that enters the machine.

  1. 01

    Detect

    CAMERA FRAME

    Machine vision identifies the product and its stem geometry as the belt carries it past.

  2. 02

    Calculate

    GEOMETRY · CUT LINE

    Software determines the cutting location for that product alone, in real time.

  3. 03

    Cut

    ROBOT FRAME

    The automation system executes the trim on the moving belt.

  • Camera frame
  • Geometry
  • Keypoints
  • Cut line
  • Robot target
  • Cut

Deep dives

Read the engineering in detail.

Send Your ProcessView machines