Machine Vision Engineer

About the Role:

We are looking for a Machine Vision Engineer who designs the imaging chain and develops the vision algorithms that run on it — from lens and lighting selection through to production software. You will combine classical machine vision with AI/ML where it adds value and deliver it as software running inside industrial equipment.

Responsibilities:

Optics and Imaging Design:

• Size the imaging chain for each inspection: field of view (FOV), working distance, spatial resolution (µm/pixel), and the measurement accuracy it must support.
• Select sensors and cameras — resolution, pixel size, sensor format, monochrome vs. colour, global vs. rolling shutter, area scan vs. line scan, dynamic range, and SNR.
• Select lenses — focal length, primary magnification, f-number and aperture, depth of field, mount type (C-mount, F-mount, TFL), telecentric vs. entocentric optics, extension tubes, and spacers.
• Evaluate optical performance before purchase: MTF and resolving power (lp/mm), distortion, vignetting, and chromatic aberration.
• Design illumination geometry — backlight, coaxial, dome, ring, bar, darkfield, and low-angle — and select wavelength, diffusion, and strobe vs. continuous operation.
• Apply bandpass and polarising filters, including cross-polarisation, to control glare, specular reflection, and ambient light.
• Set exposure, gain, and strobe timing against a motion blur budget; configure hardware, encoder, and PLC triggering with defined latency and jitter.
• Build optical test setups, calibration targets, and fixturing that give repeatable imaging on real parts.

Camera and Acquisition Integration:

• Develop real-time image acquisition and vision processing pipelines.
• Integrate industrial cameras and camera SDKs via GenICam/GenTL and SFNC feature access.
• Select interfaces and size bandwidth — GigE Vision, 10 GigE, USB3 Vision, CoaXPress, Camera Link — including PoE, cabling, and frame grabbers.
• Manage acquisition timing, buffering, and frame or line rate against machine cycle time.

Vision Algorithm Development:

• Develop vision algorithms from first principles — sub-pixel edge extraction, geometric and template pattern matching, blob and contour analysis, and image registration.
• Design and tune pre-processing chains: filtering, noise reduction, contrast normalisation, morphology, thresholding, and segmentation.
• Develop measurement, alignment, and defect detection algorithms to defined accuracy, repeatability, and cycle-time budgets.
• Validate algorithm performance on real production parts, not curated sample images.
• Develop camera calibration and lens distortion correction — intrinsics, extrinsics, radial and tangential coefficients, and µm/pixel scale factor.
• Develop hand-eye calibration and coordinate transform chains that convert pixel positions into machine or robot coordinates.

AI/ML:

• Develop, train, and optimise models where classical methods fall short.
• Deploy production-ready inference using PyTorch, TensorFlow, OpenCV, ONNX/ONNX Runtime, or TensorRT.

Application Development:

• Design and maintain robust PC applications integrating vision, AI/ML, and automation.
• Combine hardware SDKs, APIs, vision libraries, and machine interfaces into a well-structured desktop application.
• Build operator interfaces for machine status, live view, results, recipes, calibration, diagnostics, and production data.
• Ensure reliability and performance through multithreading, profiling, testing, logging, and fault recovery.

Machine Integration:

• Integrate with PLCs, robots, motion controllers, sensors, and machine I/O over industrial protocols.
• Support installation, commissioning, and debugging on equipment, in-house and at customer sites.
• Work with mechanical, electrical, controls, systems, and process engineers to validate software on real machines.
• Track developments in machine vision and AI/ML and assess their relevance to the product roadmap.

Mandatory Qualifications:

• Bachelor's or Master's in Computer Science, Electronics, Electrical, Mechatronics, or a related engineering field.
• 3+ years of professional software development in machine vision, computer vision, robotics, or industrial automation.
• Hands-on optical system design — calculating FOV, working distance, focal length, magnification, depth of field, and µm/pixel resolution, and selecting the lens, sensor, and lighting to match.
• Practical experience with industrial cameras and image acquisition: exposure, gain, triggering, illumination geometry, filters, interface, and bandwidth selection.
• Proven ability to develop vision algorithms, not only call library functions.
• Strong grounding in image processing fundamentals: filtering, morphology, thresholding, segmentation, sub-pixel edge detection, feature extraction, pattern matching, and correlation.
• Camera calibration, lens distortion correction, and coordinate transformation for vision-based measurement or alignment.
• Strong C++ or C#, with working Python for vision prototyping and AI/ML.
• Practical experience with OpenCV and/or a commercial vision library such as Halcon, Cognex VisionPro, Matrox, Basler pylon vTools, or NI Vision.
• Deployed at least one AI/ML model for a vision task — classification, detection, or segmentation.
• Integrated third-party camera or hardware SDKs into production desktop applications.
• Built reliable desktop applications using Qt/QML, WPF/.NET, PySide/PyQt, or equivalent.
• Familiar with multithreading, performance optimisation, and real-time constraints.
• Solid engineering practice: Git, debugging, testing, code reviews, and modular design.

Preferred Qualifications:

• Telecentric and bi-telecentric optics, macro imaging, Scheimpflug adapters, or line scan imaging.
• MTF and optical resolution analysis, lens qualification, and acceptance testing.
• CoaXPress, 10 GigE, frame grabbers, or other high-bandwidth acquisition.
• Strobe and overdrive lighting controllers, encoder-synchronised acquisition.
• Commercial vision packages: Halcon, Cognex VisionPro/In-Sight, Keyence, or Matrox Imaging.
• 2D/3D vision, laser profilometry, photometric stereo, metrology, optical inspection, or precision alignment.
• Deep learning for industrial inspection — anomaly detection, defect classification, segmentation — including data strategy for rare defects.
• 1D/2D code reading and OCR/OCV, including DataMatrix and print quality grading to ISO/IEC 15415/15416.
• CUDA, TensorRT, or GPU-accelerated computer vision.
• Industrial automation, PLCs, robots, motion controllers, and machine I/O.
• Industrial protocols: TCP/IP, EtherCAT, OPC UA, Modbus TCP, EtherNet/IP, or PROFINET.
• Measurement system analysis (Gauge R&R), SPC, or validating inspection accuracy against a reference gauge.
• Software for 24/7 production: diagnostics, traceability, image and result logging, and fault recovery.