Machine Vision Intelligent System Suite
Visual Mesurement and Inspection

On a lot of lines, dimensions and defects are still judged by people. A gauge on a few sampled parts, a visual pass over the rest, and a stop to investigate when something looks wrong. That holds up at low volume. As volume rises it stops working: sampling misses what it does not cover, two people judge the same mark differently, and by the time a problem is confirmed a batch has already been made.
Fitting cameras and sensors is not the hard part. What follows is: how images get aligned once they arrive, how noise is handled, what gets measured and against which limit, how the result comes out, where the data goes, and how a changeover avoids starting from nothing. Stacked together, that is why so many vision projects end with cameras installed and nothing actually running.
RUIYI Visual Measurement and Inspection covers that stretch. It is vision measurement and inspection software built for in-line use: acquisition, calibration and alignment, processing, measurement and inspection, judgement and data output configured as one flow — so dimension and defect decisions on a line move from someone looked at it to the system judged it and kept the record.
Measure it or look for it, in one place. from outer dimensions, height differences, flatness and hole position, to scratches, deformation, missing parts, wrong assembly and label checks — measurement and inspection configured in one software, one judgement output, and data back into the quality system.
What it is
It is vision measurement and inspection software that runs on the line. Cameras or sensors capture the image, the software aligns and processes it, runs the configured measurement and inspection steps, produces item-level and overall judgements, and sends the result to line control, the quality system or a dashboard.
Measurement and inspection in one configuration. No need for one software to measure dimensions and another to look for defects. Both kinds of task are arranged in the same project and tuned in the same interface.
Built for in-line use, not offline analysis. It follows the line cycle and returns a judgement per part, feeding back to the line and the quality system as it runs.
Configured rather than rewritten. Measurement items, inspection items, limits and output are all set by configuration and saved as a template — reused on the next variant or the next line instead of rebuilt from scratch.
A result that is more than pass or fail. Each judgement carries the measured value, the basis for the decision, the time and the station, so it can be traced per part and aggregated into yield, distribution and trend.
2D and 3D are both supported in software; what the hardware allows is decided on site. The software carries 2D and 3D measurement and inspection algorithms alike. Which items are actually achievable, and to what accuracy and cycle time, is determined by the sensors fitted, the calibration and the available compute. Section 3 covers this in detail.
What it measures and inspects
Measurement
Dimensions. Length, width, height, gap, diameter and angle on the outer form of the part.
Height and step. Surface-to-surface height difference, step height and small variations in level.
Flatness. Points sampled across a surface and fitted to judge whether it is even and within tolerance.
Position and holes. Location, diameter and offset of features such as holes, slots, locating pins and studs.
Cross-section and profile. Width, height and cross-sectional area of a bead or coating, and comparison against a reference profile.
Counting. A running count of parts or features as they pass.
Inspection
Surface condition. Scratches, dents, bulges, contamination, discoloration and similar surface anomalies.
Shape and deformation. Distortion, warping, skew, burrs, chipping and broken edges.
Assembly state. Missing, wrong, reversed or misaligned parts, and out-of-tolerance gaps.
Presence and count. Whether something is there, how many there are, and where they sit.
Bead and coating. Breaks in the bead, overflow, bead width or height outside limit, and missing coating.
Marking and mistake-proofing. Character recognition, barcode and 2D code reading, and checking a label against the part it is on.
Software and hardware: what each decides
This part is worth stating plainly, because it determines how far a project can actually go.
What the software provides:2D and 3D measurement and inspection algorithms configured in one interface; the full flow of acquisition, alignment, pre-processing, measurement, judgement and output; template projects reused across changeovers and copied lines; and result records, statistics and communication outward.
What the hardware decides:
Whether 3D is possible at all. Height, flatness, cross-section and volume need depth information, so they only work when a 3D sensor — a 3D camera or a laser profiler — is fitted. With 2D cameras the software handles what a flat image can support: dimensions, position, surface condition, presence and marking.
The accuracy ceiling. Achievable repeatability and resolution depend on sensor resolution, calibration quality, lighting and site conditions. Software cannot manufacture accuracy that the sensor does not deliver.
The cycle-time ceiling. Time per part depends on acquisition speed, data volume and the compute available.
Coverage. How large a field of view is, and whether the whole part fits in it, comes down to sensor choice and mounting — extended where necessary by combining more than one sensor.
Put simply: the software organises the capability and keeps it from being tied to one piece of hardware; the hardware decides how far that capability reaches on a given line. What gets measured, the target accuracy and the cycle time are settled together with the sensor choice during selection — not by choosing software first and leaving the outcome to chance.
Where it is used
Application | What is measured or checked | Typical industries |
Flatness and coplanarity | Points sampled across a surface, judged against a flatness tolerance | Automotive parts, electronics and semiconductors, EV battery |
Height difference and step | Small height differences, weld height, stepped surfaces | Electronics and semiconductors, automotive parts |
Dimensions and spacing | Outer length and width, hole pitch, assembly gap, seam width | Mechanical equipment, automotive parts |
Hole and feature position | Position and diameter of holes, slots, locating pins and studs | Mechanical equipment, automotive parts |
Bead and coating | Bead width, height and cross-sectional area; breaks and overflow | Automotive parts, EV battery |
Assembly completeness | Missing, wrong, reversed or misaligned parts; out-of-tolerance gaps | Automotive parts, mechanical equipment, appliances |
Surface defects | Scratches, dents, bulges, contamination, discoloration | Across industries |
Presence and counting | Whether a part is there, how many, and where | Across industries |
Marking and mistake-proofing | Characters, barcodes, 2D codes, label checked against the part | Food and beverage, medical, cross-border logistics |
Outer dimensions and volume | Length, width, height and calculated volume | Cross-border logistics, warehousing |
How it works
Acquire. Images are taken from cameras or sensors on a trigger — an arrival signal, an encoder, a timer or a manual call.
Calibrate and align. A relationship between image and real dimension is established, and each arriving part is position-corrected, so a part sitting slightly off does not move where the measurement is taken.
Pre-process. Denoising, filtering, enhancement and region cropping suppress interference before measurement begins.
Measure and inspect. The configured measurement and inspection items run; items needing depth information draw on the 3D data.
Judge. Each item is judged against its limit, then an overall judgement is applied by rule — for example, release only when every critical item passes.
Output and record. The result goes to line control or the sorting mechanism in real time, while the record captures measured values, judgement, time and station — available per part or aggregated by batch.
What you get
Software features
Feature | What it covers |
2D and 3D in one | 2D images and 3D data handled in the same software, with measurement and inspection items configured together |
Graphical project setup | Acquisition, alignment, processing, measurement, judgement and output configured through the interface, with typical project templates included |
Calibration and alignment | Sensor calibration, coordinate setup and per-part position correction, keeping the measurement location consistent |
Pre-processing and denoising | Filtering, enhancement and region cropping to deal with site lighting and material variation |
Measurement toolset | Tools for dimensions, height, flatness, position, cross-section and profile |
Inspection toolset | Tools for surface defects, shape, assembly, presence, bead and marking |
Judgement rules | Per-item limits and multi-condition overall judgement, configured per product and operation |
Multiple sensors combined | More than one sensor combined for a wider field of view or multiple angles on a complex part |
Records and statistics | Measured values and judgements stored per part, with query, aggregation, trend and yield reporting |
Industrial communication | Data and signals exchanged with PLCs, line control and quality systems over standard industrial protocols |
Permissions and logs | Parameter changes logged and permissions graded, so judgement criteria cannot be altered casually on the floor |
Integration & deployment
Line control. Linked to PLCs, sorting mechanisms and rejection devices, with judgement results driving the action directly.
Quality and manufacturing systems. Measured values and judgements written to QMS or MES against batch, operation and equipment, supporting traceability and statistical analysis.
Sensors and cameras. Selected against the measurement items and accuracy targets; 2D cameras and 3D sensors can be used together.
Capability layer. Runs on RUIYI Visual CAST, sharing model and device management, compute scheduling and lifecycle governance.
Deployment model — on-premise, edge or cloud — and the integration scope are agreed during scoping, usually starting with one line or one inspection station before being copied to similar lines.

