Every week there are new technological advancements being made and new applications leading to new opportunities. Machine learning and deep learning are at the forefront as two of the most exciting areas of technology in AI today. Vision-based detection and tracking algorithms have evolved significantly over the past ten years. Where features used to be engineered, they are now automatically determined with deep learning. Where tracking algorithms worked well with a few objects, they now perform well with many objects. This means we have an increased ability to process image and/or video data in a more time-efficient and accurate way. Mining and Manufacturing companies have an opportunity to capture such data with the aim of tracking personnel, threat detection for safety, product inspection, defect detection etc.
In this article we look at computer vision and how it can be used to automate a lot of these processes. Image processing (IP) and Computer Vision (CV) are strongly related and often used interchangeably. IP mainly deals with how to manipulate a pixel in an image or video, whereas CV looks at using the data to detect, classify and track objects or events in order to interpret a real-world scene. Figure 1 illustrates some of the key properties of the two techniques.
Inspection and defect detection are critical for high-throughput quality control in production or process systems, however, they can be labour-intensive which makes them very good candidates for automation. Automated systems have been widely adopted to identify flaws on manufactured surfaces such as metallic rails, semiconductor wafers and so on. Recent developments in deep learning have not only significantly improved our ability to detect defects but to also improve safety and security, process heaps of text/handwritten data and so much more. Figure 2 illustrates just some of the applications of CV in these industries.
In this next series of posts, we will show how to use traditional image processing techniques as well as deep learning-based approaches to detect and localise different types of anomalies. MATLAB provides an easy and extensible framework for CV, from data access to deployment. You will learn about:
- Data access and preprocessing techniques including denoising, registration and intensity adjustment
- Semantic segmentation and labeling of defects and abnormalities
- Defect detection using various deep learning techniques
- Deploying to multiple hardware platforms such as CPUs and GPUs
What Can I Do Next?
- Request a trial.
- Find out more from the team.
- Visit our Smart Mining & Manufacturing Focus Area page to learn more.