Identifying Wrinkles with AI-Driven Vision in Textile and Apparel Manufacturing
How NorrStudio by NorrSpect detects permanent and semi-permanent wrinkles, crease marks, fold lines, and surface deformation in dyed, finished, and technical fabrics at production speed preventing wrinkled fabric from reaching cutting, spreading, and garment assembly where wrinkles cause cutting inaccuracy, panel distortion, and appearance failures.
94%
Reduction in wrinkle-related cutting inaccuracies and garment appearance failures
1.5mm
Minimum wrinkle height detectable via oblique surface topology imaging
98.1%
Detection accuracy for permanent crease marks and process-induced wrinkle zones on woven and knit substrates
Overview
Wrinkles in textile manufacturing exist on a spectrum from cosmetic to catastrophic. A shallow surface wrinkle introduced by uneven batching tension may relax on the spreading table and cause no measurable harm to cut panel accuracy. A deep, heat-set crease created during jet dyeing where the fabric folded against itself under high temperature and pressure is permanently encoded into the fibre structure and will not press out, causing every garment panel cut across that crease to carry a visible fold mark from the finished product. Between these extremes lies a range of wrinkle types rope marks, bagging wrinkles, stenter fold marks, and roll compression creases each with a different severity, permanence, and process origin that determines whether the affected fabric can be recovered or must be written off.
NorrStudio, developed by NorrSpect, uses oblique surface topology imaging to detect wrinkles of all severities across the full fabric width at production speed classifying each wrinkle by type and permanence, and attributing it to the specific process stage responsible, enabling the finishing department to intervene before additional fabric is affected and the cutting room to receive roll maps that identify exactly which zones require avoidance.
About NorrSpect
NorrSpect is a Swedish AI company headquartered in Umeå, Sweden, specialising in industrial visual inspection for precision manufacturing. Its NorrStudio platform is deployed and validated in automotive and industrial sectors including by manufacturers such as Volvo Cars and is now purpose-built for textile and apparel quality inspection. Wrinkle detection thresholds and surface topology models are defined and validated during the pilot phase using real production fabric samples from each client facility.
Industry challenge: why wrinkle severity classification matters as much as detection
The key insight in wrinkle inspection is that not all wrinkles are equal and treating them as equal leads to either excessive false rejections of recoverable fabric or insufficient rejection of genuinely unrecoverable crease-marked cloth. A wrinkle caused by uneven tension in the batching system is typically shallow, non-directional, and fully recoverable under the stenter's smoothing action. A wrinkle caused by fabric folding under its own weight in an overfilled jet dyeing machine is deeper, directional, and partially recoverable with careful re-finishing. A crease set at high temperature during jet dyeing or heat-setting is permanent the crease angle has been thermally fixed into the polymer or fibre structure and cannot be removed by any subsequent finishing process.
A detection system that identifies all three as simply "wrinkle present" forces the quality manager into a binary pass/fail decision with no process intelligence to guide it. NorrStudio's wrinkle severity classification distinguishing surface texture wrinkles from structural fold marks and thermally set creases provides the quality manager with the information needed to make a rational, evidence-based disposition decision for each affected roll zone.
Jet dyeing rope mark
A diagonal or longitudinal crease pattern set into the fabric during rope-form jet dyeing — caused by the fabric twisting and folding against itself under the turbulent liquor flow, with severity determined by temperature, time, and fabric weight
Batching compression crease
A transverse fold mark caused by uneven fabric tension during roll batching — the fabric creasing at the point where it transitions from free span to roll contact under excessive compression, typically appearing as a repeating crease at fixed roll circumference intervals
Heat-set permanent crease
A crease thermally fixed into synthetic fibre fabric — polyester, nylon, or blended fabrics — during heat-setting or calendering at temperatures above the fabric's glass transition temperature, making the fold angle permanent and unrecoverable by subsequent finishing
Stenter entry fold mark
A longitudinal fold mark introduced as the fabric enters the stenter frame in an uncontrolled manner — caused by inadequate fabric spread guidance or a misaligned entry roller — set into the fabric surface by the stenter's drying heat before the fold can be corrected
Wet processing bagging wrinkle
A three-dimensional surface deformation caused by fabric bagging under its own weight during wet processing — the wet fabric stretching and deforming in the gravity direction, creating a bumpy, irregular surface texture that may partially recover on drying
Roll storage crease
A transverse crease introduced during roll storage — caused by the roll resting on a support edge or being stored under excessive pressure — typically shallow and recoverable but requiring identification before the roll enters the cutting room
Solution: NorrStudio AI wrinkle detection and severity classification
NorrStudio uses oblique and raking illumination where the light source strikes the fabric at a low angle to cast shadows from surface height variations, making wrinkles and crease marks visible as light-and-shadow patterns independent of fabric colour or surface texture. Surface topology analysis algorithms measure wrinkle height, width, orientation, and continuity from the shadow profile, classifying each wrinkle by its geometric characteristics and correlating those characteristics to known wrinkle type signatures. Wrinkle permanence is assessed by comparing the geometric severity of the wrinkle against the fabric's finishing history a deep crease on a heat-set polyester fabric is classified as permanent; the same depth crease on a cotton fabric prior to stenter finishing is classified as potentially recoverable.
Detects surface wrinkles as shallow as 1.5mm in height using oblique illumination shadow topology effective across all fabric colours and surface textures
Classifies wrinkle type by geometric signature rope marks (diagonal, continuous), batching creases (transverse, periodic), heat-set creases (sharp-edged, permanent), and bagging wrinkles (irregular, diffuse)
Assesses wrinkle permanence by combining geometric severity with fabric finishing history flagging heat-set creases on synthetic fabrics as unrecoverable and shallow batching wrinkles on natural fibre fabrics as potentially recoverable
Detects stenter entry fold marks within the first metre of stenter exit enabling immediate stenter re-entry for fold correction before the mark propagates through additional roll length
Identifies batching compression creases by their periodic repeat pattern correlating repeat interval to roll circumference to identify the specific batching roll or support causing the crease
Generates roll-level wrinkle severity maps classifying each roll zone as conforming, recoverable, or write-off for cutting room avoidance and re-finishing routing decisions
Provides process stage attribution for each wrinkle type distinguishing wet processing wrinkles from finishing wrinkles and storage wrinkles using production sequence correlation
Solution
NorrStudio AI Inspection Wrinkle Detection Module
Inspection scope
Dyed, finished, and technical fabric rolls at jet dyeing exit, stenter exit, calender exit, and pre-shipment inspection
Hardware
Line-scan cameras, oblique and raking illumination rigs, motion-sync encoder
Output
Real-time wrinkle alerts, severity classification, roll zone disposition maps, process attribution reports, PDF QA archive
Integration
Jet dyeing machine controls, stenter entry guidance systems, ERP / WMS, cutting room CAD and marker planning
Deployment time
Pilot phase calibrated to client fabric fibre content, finishing process, and buyer wrinkle tolerance before full deployment
Use case: polyester dye house jet dyeing rope mark elimination for woven dress fabric
The problem: A polyester woven dress fabric dye house processing 100% polyester crepe and georgette in rope form on high-temperature jet dyeing machines was experiencing a persistent rope mark problem diagonal crease patterns set permanently into the polyester fibre during dyeing at 130°C, where the glass transition temperature of the polyester was exceeded and the fabric's fold angles were thermally fixed. Approximately 9–13% of rolls per dye run contained heat-set rope marks that could not be recovered by subsequent stenter finishing, resulting in full roll write-offs. The dye house's outgoing inspection was identifying only 55–60% of rope-marked rolls the remainder escaping to the garment maker where they were discovered at spreading.
The NorrStudio solution: NorrStudio was installed at the stenter exit after high-temperature jet dyeing. Oblique illumination was configured for the smooth, low-texture surface of polyester crepe and georgette substrates where wrinkle shadow contrast is high and detection sensitivity is greatest. The system classified rope marks by depth and orientation, distinguishing heat-set permanent creases from shallow surface wrinkles recoverable by re-stenting. Rope mark severity correlation with dye machine parameters revealed that marks were significantly more severe on runs where liquor ratio fell below 1:8 insufficient liquor volume to keep the fabric rope moving freely enabling a minimum liquor ratio protocol to be implemented.
Results:
Metric | Before NorrStudio | After NorrStudio |
|---|---|---|
Rope mark detection rate at outgoing inspection | 55–60% (manual visual) | 98.1% (automated oblique topology) |
Roll write-off rate from heat-set rope marks | 9–13% per dye run | <1% per dye run |
Rope marks escaping to garment maker | Frequent — every production run | Zero escapes in 14 months post-deployment |
Recoverable vs permanent wrinkle classification | Not possible — binary pass/fail only | Automatic severity classification — recoverable zones routed to re-stenting |
Root cause process variable identified | Unknown | Liquor ratio below 1:8 — identified in first two weeks |
Roll-level wrinkle severity documentation | None | Full wrinkle severity map per roll with zone disposition — archived |
How does NorrStudio detect wrinkles on smooth, low-texture fabrics like polyester crepe where surface contrast is minimal?
NorrStudio's wrinkle detection uses oblique and raking illumination light sources positioned at a low angle relative to the fabric plane rather than frontal illumination. On smooth, low-texture fabrics like polyester crepe and georgette, this configuration is particularly effective: the smooth surface reflects raking light very cleanly, and even shallow wrinkles of 1.5–2mm height cast clearly defined shadows that the AI model detects as surface height anomalies. Smooth fabrics with no surface texture are actually easier for raking light wrinkle detection than heavily textured fabrics, where the texture itself generates competing shadow patterns.
How does NorrStudio classify a wrinkle as permanent versus recoverable without physically pressing the fabric?
NorrStudio's permanence classification combines wrinkle geometry with fabric process history data. A deep, sharp-edged crease with a high shadow contrast ratio on a synthetic fabric that has been processed above its glass transition temperature is classified as heat-set permanent — the geometric severity combined with the known thermal processing history is sufficient evidence of permanence without physical pressing. A shallow, rounded wrinkle with low shadow contrast on a natural fibre fabric prior to stenter finishing is classified as potentially recoverable the geometry indicates insufficient severity for fibre-level deformation, and the finishing sequence includes a process stage capable of relaxing the wrinkle. The classification logic is calibrated to each client's fabric range and finishing process during the pilot phase.
Can NorrStudio detect batching compression creases caused by a specific roll or support in the batching system?
Yes. Batching compression creases appear at intervals corresponding to the circumference of the roll or support causing the compression. NorrStudio measures the repeat interval of transverse crease patterns and correlates it with the circumference data for each batching roll and support in the system, identifying the specific mechanical component responsible. This targeted maintenance signal is what distinguishes NorrStudio's output from a generic wrinkle report it tells the maintenance team exactly which component to inspect rather than generating a batching system fault with no further guidance.
Does wrinkle detection work on pile fabrics like velvet and velour where surface topology is inherently three-dimensional?
Yes, with fabric-specific model training. On pile fabrics, the pile height creates a natural surface topology that would generate false wrinkle detections under a generic shadow analysis system. NorrStudio's pile fabric wrinkle model is trained to recognise the normal pile topology as the baseline and detect deviations crushed pile zones, pile flatten marks, and structural fold marks — as anomalies against that baseline. The training data includes the specific pile fabric construction from the client's range, enabling accurate wrinkle detection on velour, velvet, and terry pile substrates
At what production stage should NorrStudio wrinkle detection be deployed for maximum value?
The highest-value deployment point depends on the primary wrinkle risk in the client's process. For jet-dyed fabrics where rope marks are the primary risk, deployment at the stenter exit after dyeing catches heat-set creases before they enter the cutting room. For fabrics where batching compression is the primary risk, deployment at the batching station exit is most effective. For facilities where wrinkle risk spans multiple process stages, deploying at both the dye house exit and the final finishing exit provides attribution by process stage and catches wrinkles introduced at any point in the production chain.
Similar Topic
