Measurement and quality control systems with industrial machine vision

We treat dimensional measurement, burr and deformation as separate requirements

In measurements such as diameter, width, hole position or contour deviation, the element to be controlled, reference and tolerance are defined. In burr control, it is determined where and in what view the area protruding from the edge is sought. Just finding a defect and measuring its physical size are different outcomes. Which plane or surface the measurement belongs to is written in the technical requirement.

We determine whether deformation appears as an in-plane contour change, height variation or bending. If a two-dimensional view is insufficient, alternative viewing directions or 3D profile measurement are assessed. Visibility, acceptance limits and reference verification are defined for each defect type so different quality decisions are not reduced to one undifferentiated pass/fail label.

• Measurement, reference, tolerance and control zone
• Separate definition of edge burr and surface defect
• Distinguishing 2D contour requirements from height and deformation requirements

We choose camera, lens and lighting by real part

Backlighting can highlight the outer contour; A mark or burr on the surface may require different light directions. Shiny metal, dark plastic and textured surface do not produce the same contrast. Field of view, working distance, optics and light pattern are evaluated based on the target feature. Changes in ambient light or contamination should not make the good image in the initial sample unsustainable in production.

Exposure and sensor readout matter when imaging moving parts. Area-scan and line-scan cameras are compared according to part geometry, feed conditions and inspection area. Camera resolution is not presented as measurement accuracy. On the prototype setup, we verify that the target defect can actually be distinguished and critical dimensions evaluated under defined conditions.

• Light suitable for part surface and defect direction
• Joint evaluation of field of view, optics and exposure
• First imaging attempt with real samples

We develop mechanical fixture, trigger and electronic interfaces together

The height, orientation and vibration of the part relative to the camera are part of the measurement conditions. The camera bracket, light housing, parts guide, and fixture, if necessary, are designed with the existing line layout. Cleanliness, adjustment and service access are also taken into account. If it is desired to control the part without stopping it, it is examined with what position stability the feed can provide this.

Image capture is synchronised with trigger signals from the photoelectric sensor, encoder or line controller. Where custom electronics are needed, we develop the electrical interface between the sensor, lighting and PLC. Identity and position tracking link each result to the correct part. Changes in speed or part spacing before the sorting point are addressed rather than simply assuming a fixed time delay.

• Brackets, fixtures, guides and mechanical guards
• Sensor, encoder and lighting synchronization
• Part identification and position tracking on the line

The inspection task determines the choice of image processing or AI

Comparing the location of a significant edge, number of holes, or defined contour can be resolved with rule-based tools. Artificial intelligence-based model development can be considered for surface defects or classification tasks with a large variety of appearances. It is not assumed that every camera project requires AI or that adding a model automatically corrects poor image.

If AI is used, sample collection, labelling, training and independent test data form part of the scope. Mixing very similar images of the same part between training and testing can overstate performance. Model version, decision threshold and operating conditions are recorded. Geometric measurement may be combined with a learned defect decision where suitable, but model confidence does not replace physical measurement accuracy.

• Measurement and rule tools for defined geometry
• Data and model needs according to appearance diversity
• Model, recipe and decision threshold version tracking

Integration with the PLC, sorting mechanism and recording system

The image processing result, along with the measurement value and error code, can be transferred to PLC or the corresponding control software. The interface is defined for situations such as the part has been received, the examination has been completed and the result is valid. The separation mechanism receiving the command and the actual separation of the part are different events; The required verification sensor and mechanical layout are planned accordingly.

Missing images, lost triggers, communication interruptions and late results are distinguished from valid acceptance results. The inspection station complements the existing control system; machine safety functions remain within the safety controls and architecture established by risk assessment. Product recipes, user permissions, sample images and event logging are scoped around the customer's maintenance and traceability needs.

• Result, part ID and PLC handshake
• Separating the sorting command from verification of the physical outcome
• Invalid inspection and connection loss scenarios

We attribute the precision claim to actual measurement and verification data

Converting length in pixels to physical units requires calibration appropriate to the application. Optics, camera position and measurement plane are part of this relationship. The measurement is compared to the reference method; Behavior in different locations and repeated parts is evaluated. When the position of the camera or lens changes, the relevant controls are refreshed. Resolution, repeatability and accuracy are treated as separate metrics.

In defect control, unnecessary rejection of the healthy part and acceptance of the defective part are recorded separately. Boundary samples, different production batches and actual line conditions are included in the verification plan. Performance; It is described by the data set, fault description, line condition, and software version. We do not promise a specific speed, tolerance or defect detection rate without testing. Measurable acceptance conditions are completed based on pilot findings.

• Reference measurement and calibration plan
• Evaluating false rejects and accepted defective parts separately
• Acceptance result explained by data, condition and version

Guide to camera-based measurement and defect analysis on a conveyor belt →

From prototype to pilot line, from confidentiality to source files

Imaging and critical functions are validated first, then electronics, mechanics and software are combined in the pilot station. Maintenance steps cover lighting and lens cleaning, recipe changes and revalidation. Prototype PCB fabrication and assembly support for the required custom board is offered only for customer projects developed by VOMRO. Scope, acceptance criteria and technical deliverables are clarified at the outset.

Before technical details, we will sign a confidentiality agreement (NDA) upon your request. Project source codes, electronic and PCB design resources, mechanical CAD files and technical documents are delivered at the end of the project. Model outputs developed under AI, training/operation information, and limits of data delivery are also listed; The license terms of third-party tools, models and libraries are explained separately. You can start the first meeting with the part drawing, samples, control purpose and existing line information.

The quote separately explains the scope of the imaging trial, camera and optical hardware, mechanical setup, software or model development, production line integration and commissioning. Equipment to be reused and third-party licences are also specified; the price of a single camera does not represent the cost of a complete inspection station.

• Sample trial, pilot station and acceptance plan
• NDA and customer-specific development on request
• Source, design and technical file delivery at the end of the project