AI-driven Plastic Surface Defect Detection via UV

Details

As I was reading about the applications of UV (ultraviolet) radiation in industrial operations, especially for anomaly detection, I became fascinated by the possibility of developing a proof-of-concept AI-driven industrial automation mechanism as a research project for detecting plastic surface anomalies. Due to the shorter wavelength of ultraviolet radiation, it can be employed in industrial machine vision systems to detect extremely small cracks, fissures, or gaps, as UV-exposure can reveal imperfections on which visible light bounces off, leading to catching some production line mistakes overlooked by the human eye or visible light-oriented camera sensors.

In the spirit of developing a proof-of-concept research project, I wanted to build an easily accessible, repeatable, and feature-rich AI-based mechanism to showcase as many different experiment parameters as I could. Nonetheless, I quickly realized that high-grade or even semi-professional UV-sensitive camera sensors were too expensive, complicated to implement, or somewhat restrictive for the features I envisioned. Even UV-only high-precision bandpass filters were too complex to utilize since they are specifically designed for a handful of high-end full-spectrum digital camera architectures. Therefore, I started to scrutinize the documentation of various commercially available camera sensors to find a suitable candidate to produce results for my plastic surface anomaly detection mechanism by the direct application of UV (ultraviolet radiation) to plastic object surfaces. After my research, I noticed that the Raspberry Pi camera module 3 was promising as a cost-effective option since it is based on the CMOS 12-megapixel Sony IMX708 image sensor, which provides more than 40% blue responsiveness for 400 nm. Although I knew the camera module 3 could not produce 100% accurate UV-induced photography without heavily modifying the Bayer layer and the integrated camera filters, I decided to purchase one and experiment to see whether I could generate accurate enough image samples by utilizing external camera filters, which exposes a sufficient discrepancy between plastic surfaces with different defect stages under UV lighting.

In this regard, I started to inspect various blocking camera filters to pinpoint the wavelength range I required — 100 - 400 nm — by absorbing visible light spectrums. After my research, I decided to utilize two different filter types separately to increase the breadth of UV-applied plastic surface image samples — a glass UV bandpass filter (ZWB ZB2) and color gel filters (with different light transmission levels - low, medium, high).

Since I did not want to constrain my experiments to only one quality control condition by UV-exposure, I decided to employ three different UV light sources providing different wavelengths of ultraviolet radiation — 275 nm, 365 nm, and 395 nm.

✅ DFRobot UVC Ultraviolet Germicidal Lamp Strip (275 nm)

✅ DARKBEAM UV Flashlight (395 nm)

✅ DARKBEAM UV Flashlight (365 nm)

After conceptualizing my initial prototype with the mentioned components, I needed to find an applicable and repeatable method to produce plastic objects with varying stages of surface defects (none, high, and extreme), composed of different plastic materials. After thinking about different production methods, I decided to design a simple cube on Fusion 360 and alter the slicer settings to engender artificial but controlled surface defects (top layer bonding issues). In this regard, I was able to produce plastic objects (3D-printed) with a great deal of variation thanks to commercially available filament types, including UV-sensitive and reflective ones, resulting in an extensive image dataset of UV-applied plastic surfaces.

✅ Matte White

✅ Matte Khaki

✅ Shiny (Silk) White

✅ UV-reactive White (Fluorescent Blue)

✅ UV-reactive White (Fluorescent Green)

Before proceeding with developing my industrial-grade proof-of-concept device, I needed to ensure that all...

Components

1 × ELECROW Regular PCB (4-layer)

1 × Raspberry Pi 4 Model B

1 × Raspberry Pi 5

1 × Raspberry Pi Camera Module 3 Wide

1 × Raspberry Pi Camera Module 3 NoIR Wide

Build Instructions

Collapse

1

Development process, different prototype versions, design failures, and final results

As I was developing this research project, I encountered lots of problems due to complex mechanical component designs, especially related to the sprocket-chain mechanism, leading me to go through five different iterations. I documented the overall development process for the final mechanism in the following written tutorial thoroughly and showcased the features of the final version in the project demonstration videos.

Every feature of the final version of this proof-of-concept automation mechanism worked as planned and anticipated after my adjustments, except that the stepper motors (Nema 17) around which I designed the primary internal gears could not handle the extra torque applied to my custom-designed ball bearings (with 5 mm steel beads) after I recalibrated the chain tension with additional tension pins. I explained the reasons for the tension recalibration thoroughly in the following steps. In this regard, I needed to record some features related to sprocket movements (affixed to outer gears pivoted on the ball bearings) by removing or loosening the chain for the demonstration videos.
2
Step 2
As I briefly talked about my thought process for deciding the experiment parameters and sourcing components in the introduction, I will thoroughly cover the progress of building the UV-applied plastic surface image sample (data) collection rig in this section.

The simple data collection rig is the first version of this research project, which helped me to ensure that all components, camera filters, UV light sources, and plastic materials (filaments) I chose were compatible and sufficient to produce an extensive UV-applied plastic surface image dataset with enough discrepancy (contrast) to train a visual anomaly detection model.

As mentioned, after meticulously inspecting the documentation of various commercially available camera sensors, I decided to employ the Raspberry Pi camera module 3 Wide (120°) to capture images of plastic surfaces, showcasing different surface defect stages, under varying UV wavelengths. I studied the spectral sensitivity of the CMOS 12-megapixel Sony IMX708 image sensor and other available Raspberry Pi camera modules on the official Raspberry Pi camera documentation.

Since I decided to benefit from external camera filters to capture UV-oriented image samples with enough discrepancy (contrast) in accordance with the inherent surface defects, instead of heavily modifying the Bayer layer and the integrated camera filters, I sourced nearly full-spectrum color gel filters with different light transmission levels for blocking visible light. By stacking up these color gel filters, I managed to capture accurate UV-induced plastic surface images in the dark.
- Godox color gel filters with low light transmission
- Godox color gel filters with medium light transmission
- Godox color gel filters with high light transmission
Of course, only utilizing visible light-blocking color gel filters was not enough, considering the extent of this research study. In this regard, I also sourced a precise glass UV bandpass filter absorbing the visible light spectrum. Although I inspected the glass bandpass filter specifications from a different brand's documentation, I was only able to purchase one from AliExpress.
- UV bandpass filter (25 mm glass ZWB ZB2)
As I did not want to constrain this research project to showcase only one UV light source type while experimenting with quality control conditions by the direct application of UV (ultraviolet radiation) to plastic object surfaces, I decided to purchase three different UV light sources providing different UV wavelength ranges.
- DFRobot UVC Ultraviolet Germicidal Lamp Strip (275 nm)
- DARKBEAM UV Flashlight (395 nm)
- DARKBEAM UV Flashlight (365 nm)
Since I decided to manufacture plastic objects myself to control experiment parameters to develop a valid research project, I needed to find an applicable and repeatable method to produce plastic objects with varying stages of surface defects (none, high, and extreme) and source different plastic materials to produce a wide selection of plastic objects. After mulling over different production methods, I decided to produce my plastic objects with 3D printing and modify slicer settings to inflict artificial but controllable surface defects. Thanks to commercially available filament types, including UV-sensitive and reflective ones, I was able to source a great variety of materials to construct an extensive image dataset of UV-applied plastic surfaces.
- ePLA-Matte Milky White
- ePLA-Matte Light Khaki
- eSilk-PLA White (Shiny)
- PLA+ Luminous Green (UV-reactive - Fluorescent)
- PLA+ Luminous Blue (UV-reactive - Fluorescent)
#️⃣ First, I designed a simple cube on Autodesk Fusion 360 with dimensions of 40.00 mm x 40.00 mm x 40.00 mm.

#️⃣ I exported the cube as an STL file and uploaded the exported STL file to Bambu Studio.

#️⃣ Then, I modified the slicer (Bambu Studio) settings to implement artificial surface defects, in other words, inflicted top-layer bonding issues.

#️⃣ Since I wanted to showcase three different surface defect stages — none, high, and extreme — I copied the cube three times on the slicer.

#️⃣ For all three cubes, I selected the sparse infill density as 10% to outline the inflicted surface defects.

#️⃣ I utilized the standard slicer settings for the first cube, depicting the none surface defect stage.

#️⃣ For the second cube, I reduced the top shell layer number to 0 and selected the top surface pattern as the monotonic line, representing the extreme surface defect stage.

#️⃣ For the third cube, I lowered the top shell layer number to 1 and selected the top surface pattern as the Hilbert curve, representing the high surface defect stage.

#️⃣ However, as shown in the print preview, only reducing the top shell layer number would not lead to a protruding high defect stage, as I had hoped. Thus, I also reduced the top shell thickness to 0 to get the results I anticipated.
#️⃣ Since I decided to add the matte light khaki filament latest, I sliced three khaki cubes with 15% sparse infill density to expand my plastic object sample size.

After meticulously printing the three cubes showcasing different surface defect stages with each filament, I produced all plastic objects (15 in total) required to construct an extensive dataset to train a visual anomaly detection model and develop my industrial-grade proof-of-concept surface defect detection mechanism.
3

Continue reading the full project tutorial

Continue reading the full project tutorial