Smart Inspect 3D

SI3DStereo2B + 3B + AIPLC / Robot

One software for the whole inspection station

Smart Inspect 3D (SI3D) checks in a single application whether a part is dimensionally correct, whether it carries a defect such as a scratch, dent or missing component, and whether its label and code can be read. The verdict is reported instantly to the PLC or robot on the line, so faulty parts are stopped or diverted.

One piece of software drives the camera, the laser, the 3D scanner, the robot and the AI models — all from the same screen.

27ready-made inspection modules

2B+3B+AIthree technologies, one platform

4simultaneous independent lines / stations

µmsub-pixel measurement resolution

At a glance

What the platform brings

27 ready-made inspection modules

Measurement, part locating, code reading, AI and 3D scanning — combined with drag-and-drop logic.

Set up without writing code

A “recipe” is built on screen for each product; no programmer required.

2D + 3D + AI

Use all three technologies on one platform instead of buying three separate products.

Four lines / stations at once

One computer, four independent inspection flows — none waits for another.

Direct line integration

Two-way PLC and robot communication with a ready-made signal set.

Full traceability

Every part's measurements, image and verdict are recorded.

Capabilities

What can it do?

Twenty-seven inspection modules are combined in a single recipe: measurement, 3D scanning, AI, code reading and robot guidance.

01

Measurement and metrology

Diameter, length, angle, distance, concentricity, roundness, flatness… You mark the region to measure on screen; the software finds the edges with sub-pixel accuracy and converts them to real-world units (mm). Every measurement gets lower/upper tolerances and is evaluated automatically as Pass / Fail.

Geometric primitives: point, line, circle, arc, triangle, polygon, plane
Intersection, tangency, distance and angle between two geometries
Alignment and levelling against a reference plane — measurements stay correct even when the part arrives slightly tilted
Deviation-based evaluation against a master part

02

3D measurement and scanning

Three different 3D methods for features a single camera cannot see: height, depth, slope, volume. The resulting point cloud can be rotated, inspected on screen and saved.

3D point measurement with a stereo camera pair — two cameras, like human eyes, give the true X-Y-Z coordinate in space
Structured-light 3D scanning — a projector casts a pattern onto the surface and a point cloud of the whole surface is produced
Laser profile scanning — precise cross-section profiles for weld seams, grooves, channels and height measurement

03

CAD comparison

Overlays the 3D scan of the part on your original CAD model and produces a colour-coded deviation map. Green areas are inside tolerance, red areas outside; the maximum deviation can be tied directly to the Pass/Fail verdict. On complex castings, welded bodies and formed parts it answers “where exactly did it deviate?” at a glance.

04

AI-based defect detection

For defects that measurement cannot catch — surface scratches, stains, paint faults, missing or incorrect assembly, burrs, contamination. The most striking part: you do not need to collect images of defective products to train the system. You show it images of good products only; the software learns what “normal” looks like and flags anything that deviates from it.

Model training runs in the built-in DL Studio screen: upload the images, mask the areas you don't care about, start the training
No external AI specialist required
Large, high-resolution images are automatically tiled and scanned, so even the smallest defects are not missed
Rare defects never seen before are caught as well

05

Part locating and positioning

Parts never arrive in exactly the same place. The software finds the part in the image, calculates its angle and offset, and automatically shifts every measurement region onto the part.

Shape/edge-based model finding — including rotated and rescaled parts
Image pattern matching
Blob/hole/object counting with area and shape filtering (e.g. “there must be 88 holes”)
Coordinates of located parts are sent to the robot for vision guided robotics

06

Code and text reading

Label and code verification runs inside the same recipe as the measurements.

DataMatrix and barcode reading — traceability and part tracking
Optical character recognition (OCR) for labels, serial numbers, lot codes and date prints
The value read can be compared with the expected value and folded into the Pass/Fail verdict

07

Robot and line integration

Inspection results are translated into the language of the line and written back to it.

Two-way PLC communication: part-ready signal, inspection complete, OK/NOK, alarm, step handshake
The robot's live pose is read; results are converted into the robot coordinate system and written back
Hand-eye calibration: with a camera mounted on the robot arm, the relation “which robot coordinate does this camera point correspond to?” is solved automatically
Motorised axis control (scanning motion, positioning) is managed from within the software

On the shop floor

Built for daily production

Everything the operator, the quality engineer and the maintenance team touch every day is designed to need as little attention as possible.

Operators need to learn nothing

The operator screen shows one thing: a large, unmistakable PASS / FAIL, next to the part image and which measurement failed and why. A single key switches to full-screen view, readable from metres away on the large monitor facing the line.

No setup when the product changes

The PLC tells the software which product is coming and the correct recipe is loaded automatically. Operators do not pick from a list, so measuring the wrong product with the wrong recipe becomes impossible.

Daily production counters

Total / good / faulty counts for the day are kept per product and shown on screen. At the end of the shift, “how many parts today, how many faulty?” is already answered.

Evidence of every faulty part is kept

The image of every NOK part is archived automatically. When a customer complaint arrives, being able to say “this is how that part looked as it passed the line” — together with the measurement log — is strong evidence.

Lighting is handled automatically

Each inspection step switches its own light on and off: which LED group or projector colour is used and how long to wait after it fires are all defined in the recipe. Multi-exposure (HDR) imaging handles bright, reflective surfaces — a decisive difference on chrome, glass and machined metal.

Guided, step-by-step calibration

The calibration board is brought into view, the software states how many images to take at which position and reports the quality of the result itself. A verification screen answers “is it really measuring correctly?”, and every calibration is stored with its history.

Fast setup and commissioning

Recipe building is entirely visual: add the module, draw the region, enter the tolerance, save. Recipes can be prepared offline on stored images with no camera connected, and an automatic test loop runs a recipe hundreds of times in a row to prove its stability before commissioning.

The software updates itself

When a new version is released the program announces it at startup and updates with a single confirmation; only changed files are downloaded and each one is integrity-checked. No walking around the plant with a USB stick. Firmware of the 3D scanner hardware is updated from the same interface.

User levels and remote triggering

Operators only run the system and see the verdict; critical actions such as editing recipes, changing tolerances and calibrating are closed to unauthorised users. Higher-level plant systems (MES/ERP) can connect, call an inspection program and collect the result.

Technical advantages

Why it holds up in production

One platform, one point of contact

You don't buy one product for 2D measurement, a second for 3D and a third for AI. They live in the same recipe, the same coordinate system and the same screen. Licensing, training, maintenance and support all come from one place.

Four independent inspection flows at once

Up to four stations or lines run in parallel on a single computer. Each flow has its own trigger signal, its own OK/NOK output and its own result panel — one running a long 3D scan does not hold up the others.

Sub-pixel accuracy

Edge detection does not stop at the pixel boundary; intensity transitions are evaluated to compute the position within the pixel. With the same camera and optics you get markedly finer measurement resolution than an ordinary system.

Measurement independent of part position

Fixture tolerance, conveyor drift or robot repeatability do not corrupt the result. The software first finds the part, seats the coordinate system on it and only then measures — which reduces the need for expensive precision fixtures.

Results are translated into the robot's language

Thanks to hand-eye calibration, the position the camera sees is converted straight into the robot's coordinate system and written to it. The robot picks the located part where it lies — no fixturing needed.

Modular licensing

Only the modules you need are enabled. Start with 2D measurement today and add 3D scanning or the AI module to the same software later — the investment is staged.

Open data output

Results are written to Excel/CSV tables, XML records and PLC data blocks, and together with the image archive they plug into your quality management system. The data stays in your systems instead of being locked in a black box.

Freedom of industrial camera choice

Area scan cameras from different manufacturers, GigE line scan cameras and 3D scanners are all driven from the same interface. The camera brand is chosen per application; the software stays the same.

Communication that never stalls the line

PLC traffic is managed thread-safely, so signal reads and writes never collide while an inspection is running. Alarm and main-alarm signals are defined separately and faults are reported to the line immediately.

Where it is used

Sectors and applications

Automotive supply industry

Dimensional inspection of castings and machined parts, assembly verification, weld seam control.

Metal forming and sheet metal

Hole position and diameter, bend angle, cut dimensions.

Appliances and electronics assembly

Presence and position checks, screw / clip / connector verification.

Plastic injection moulding

Flash, short shots, dimensional deviation.

Paint and coating

Surface defects, runs, stains, coating continuity.

Packaging and labelling

Label presence and position, barcode/DataMatrix reading, expiry date verification.

Robotic feeding cells

Locating randomly presented parts and reporting their position to the robot.

Setup and requirements

An industrial PC running Windows 10/11 (64-bit)
If the AI module is used, a GPU speeds up training and inference significantly
The installer ships as a single package; the AI component is an optional separate package
Camera, lighting, laser and scanner hardware are specified together, according to the application

Commissioning, calibration and the first recipes are carried out on site by the Gökser Vision engineering team, and your team is trained during the same process.

Why Smart Inspect 3D?

It installs as easily as an off-the-shelf smart camera and stays as flexible as an engineering platform.

Solutions on the market usually fall into two groups: closed, limited smart cameras — easy, but you hit a wall the moment your application steps outside the box — or complex platforms that need a specialist and a consultant call for every change.

Smart Inspect 3D closes the gap between the two. Your own team builds and edits the recipes, while advanced capabilities — 3D scanning, CAD comparison, AI and robot guidance — sit there for the day you need them. And the team developing the software is in Turkey, so when you have a special request you are not waiting on a foreign vendor's roadmap.

Request a Demo Other R&D Projects