The RDI service "Study evaluating image enhancement algorithms under low-light conditions, for increased accuracy of automated visual inspection systems in industrial applications" provides specialist support for the assessment, development and optimisation of the image processing solutions used in automated visual inspection systems. The service is aimed at industrial applications in which the quality of images acquired under reduced lighting directly influences the accuracy of defect detection, the traceability of processes and the efficiency of quality control.
The activities include the analysis and enhancement of images through classical processing methods, such as contrast adjustment, noise reduction, filtering and illumination correction, as well as through advanced deep learning techniques based on convolutional neural networks, generative models and transformer architectures. The experimental data is obtained using industrial cameras and controlled lighting systems, and is subsequently processed, compared and validated with specialised software tools.
The service allows the performance of algorithms to be assessed through quantitative indicators, such as PSNR and SSIM, and also through the qualitative appraisal of visual improvements relevant to defect identification, object segmentation, surface feature recognition and the increased robustness of inspection systems. The results can support the choice of the most appropriate pre-processing methods, the optimisation of visual analysis workflows and the development of more precise and more stable industrial solutions.
The beneficiaries of the service may include companies in the automotive and electronics industries, organisations specialising in non-destructive testing and visual inspection, developers of computer vision software solutions, research laboratories and industrial operators interested in automating quality control. The service contributes to reducing detection errors, improving the reliability of vision systems and increasing the performance of inspection processes under difficult lighting conditions.
The activities are carried out using high-performance workstations equipped with dedicated graphics processors, such as NVIDIA RTX/A100, software tools for image processing and for training deep learning models, including Python, OpenCV, TensorFlow, PyTorch and MATLAB, storage solutions for large image datasets, external storage and backup devices, software licences, experimental datasets and image acquisition configurations. Through these resources, ICSTM provides RDI support for the development and validation of advanced visual inspection solutions applicable in complex industrial environments.