Development of image processing softwar

Image processing software is not only used to automate inspections, but also serves as an important foundation for stabilizing on-site quality, reducing variations in judgment, and easing the burden on workers. At Katotech, we carefully organize inspection processes and judgment conditions, and propose image processing systems that can be used in actual situations.

Recently, in addition to traditional rule-based processing, there has been increasing interest in inspection, character recognition, classification, and anomaly detection that combine AI and deep learning. However, it is not just a matter of replacing everything with AI; it is important to choose a method that takes into account stability, ease of explanation, and operational costs. At Katotech, we value the idea of ​​using the appropriate technology for the necessary areas.

Inspection process issues may be solved with just cameras and lighting, or may require jigs, transportation, PLC coordination, history management, and even operation screens. At Katotech, we proceed with systemization by looking at the overall picture that fits into on-site operations, rather than focusing on individual software.

Flow from PoC to mass production introduction

First, check the target workpiece, inspection purpose, required accuracy, takt time, and site conditions, and decide which method is realistic. Then, through prototype evaluation and PoC, we will flesh out image acquisition conditions, judgment logic, necessary UI, and equipment coordination methods. In order to create a system that can be used for a long time, it is important to not only focus on accuracy at the initial stage, but also to create a design that can withstand operation.

Katotech doesn’t just develop software, we also handle fine-tuning and reviewing inspection conditions after installation. Since products and conditions change at the production site, creating a configuration that is easy to improve even after introduction will lead to results.

Proposals based on recent trends

In recent years, while expectations are high for generative AI and deep learning, there is also a focus on reproducibility, explainability, and maintenance costs in the field. Katotech proposes a balance that is reasonable for the field, such as keeping the configuration simple for areas that can be adequately covered by conventional image processing, and using AI to assist in difficult judgments.

In addition to simply displaying inspection results on the screen, we also place emphasis on a design that allows for log accumulation, image storage, linkage with higher-level systems, and analysis for quality improvement. As a result, the effects of introduction extend beyond the automation of inspection processes to continuous quality improvement.