Small defects are missed
Scratches, cuts, pinholes, stains, folds and marking defects are easy to miss at high line speed.
Result: defects move downstream and become more expensive at later stages.
The machine vision system detects defects on rolled metal, fabrics, films, conveyor belts, panels and markings: scratches, cuts, folds, coating flaws, contamination, print shifts and visible deviations.
On a moving belt, web material or panel, a defect can be small, short-lived and visually similar to normal texture. The system detects the deviation, stores the frame, calculates statistics and sends a signal to the operator or enterprise systems.
Scratches, cuts, pinholes, stains, folds and marking defects are easy to miss at high line speed.
Result: defects move downstream and become more expensive at later stages.Metal, film, fabric, rubber and panels have glare, patterns, pile, noise and natural variation.
Result: stable lighting, proper optics and models adapted to the material are needed.A deviation is often noticed after a roll, batch or shift, when a large amount of defective material has already been produced.
Result: more scrap, returns and harder root-cause analysis.Without coordinates, size and photo evidence, it is difficult to analyze defect recurrence by line, batch and material.
Result: quality remains reactive rather than managed.The solution is based on an industrial camera, properly selected optics, stable lighting, defect detection and segmentation algorithms, an operator interface and integration with production systems.
The system can work with a simple pass/fail logic or go deeper: define defect type, coordinates, area, length, width, severity class, recurrence and statistics by rolls, batches and shifts.
Functionality can be launched step by step: from one camera and one defect class to a network of lines with a unified analytics center.
Basic automatic rejection based on a visually apparent surface deviation.
Detection of scratches, dents, folds, pressure marks, rust, coating flaws, edge damage and surface defects.
Inspection of holes, cuts, folds, pile, uneven dyeing, contamination, foreign inclusions and texture deviations.
Checking print quality, marking shift, missing elements, unreadable symbols and application errors on panels, packages and parts.
Inspection of edge, dimensions, shape, position, skew, folds, tears and visually measurable deviations.
Accumulation of real defect examples and adding new classes without rebuilding the entire system.
The operator sees current events and confirmation frames, the process engineer sees defect maps and recurrence, and management sees KPIs by line, shift, material and batch.
A selection of STATANLY industrial cases: assembly inspection, completeness, rolled metal, fabric, web, print and production-line defect control.
The system controls the continuity of sealant application on the cylinder block and correct traverse installation on valves on the main assembly conveyor.
The system controls engine assembly quality: missing components, correct element installation and sealant application defects.
The system recognizes critical surface defects, web damage and poor-quality flexographic printing on the top surface of products.
The system recognizes fabric defects in real time: oil stains, holes, broken-needle traces, uneven dyeing and other deviations.
The solution includes a camera, optics, lighting, protective design, computing contour, web interface and integration with enterprise systems.
Camera selection depends on defect size, line speed, web width and working distance.
Reliable defect detection requires stable light, repeatable scene geometry and stable material position.
The system can run on the customer's site: edge, local server or hybrid architecture.
Events are transferred to quality systems, HMI, SCADA, MES, ERP and rejection systems.
The system is compared with the customer's reference during the pilot: manual inspection, engineering labeling or an approved acceptance test plan. Target metrics are fixed by agreed defect classes, capture conditions, materials and product types.
The project starts step by step: pilot at one point, metric confirmation, process integration and scaling.
We define the material, surface, defect types, accuracy requirements, line speed and reaction to defects.
We select the camera, optics, lighting, working distance, protection design and computing contour.
We label data, train the model, agree on defect classes, configure the interface and events.
We fix accuracy, exclusions, interpretation logic, operator interface and operation rules.
We send events to HMI, SCADA, MES, ERP, a rejection system or quality analytics.
We expand the system to new lines, materials, defect classes and a unified reporting center.
Name: System for detection and recognition of characteristics of processes and objects.
Brief description: an open computer-vision algorithm library for determining characteristics of objects and processes: size and dimensions, type, speed, color, geometric parameters, fonts, numbers, QR codes and other features.
Links to the registry, documentation, installation and operation guides, repository, model weights, accreditation and the right holder's charter.
We start from one inspection point: material, defect types, acceptance reference, pilot KPIs, line integration and impact calculation before scaling.
