Learn and Recognize Patterns
The CURE Pattern ID Toolkit helps you quickly create automated inspection, process monitoring, anomaly detection, knowledge capture, sensor fusion, machine vision, and other complex pattern identification applications. CURE (Concurrent Universal Recognition Engine) is a scalable pattern identification engine that can Learn-Recognize-Act® in real- time.
The target user of the CURE Pattern ID Toolkit is the automation or manufacturing engineer who is frustrated by the lack of tools that can learn by example. You often know what your waveforms and assemblies should look like, so why can’t you teach that to your automated inspection system? You don’t need to know what your sensor outputs should look like; CURE can figure it out and tell you when something isn’t right.
The CURE Pattern ID Toolkit provides NI LabVIEW users with the following benefits:
- Data-agnostic development (all data/sensor types: waveforms, images, and so on)
- On-the-fly system training by example
- Real-time learning and recognition
- Pattern classification
- Fuzzy or exact pattern matching
- Anomaly detection
- Process modeling by observation
- Algorithm parallelism and scalability
Figure 1. Example of Teaching Your Computer to Recognize Waveforms
VI Library for Pattern Recognition
The CURE Pattern ID Toolkit provides easy-to-use virtual instruments (VIs) so you can acquire data in ways you are used to, preprocess it, extract features, and then forward these features to CURE for pattern learning and identification. You can use CURE identification results to classify features, detect anomalies, and take appropriate action.
For example, imagine using sensors to produce a set of waveforms. You want to capture patterns from these waveforms (for example, take subsamples) and learn them. CURE builds an empirical data model formed from the patterns and notifies you when anomalies occur. If you leave the learning feature on all the time, CURE adjusts to normal baseline shifts (for example, from tool wear). You can virtually eliminate false alarms and the need to frequently recalibrate or adjust thresholds and process limits, which saves you time and money.
Imagine you have images of your assembly and you need quality assurance. When you give CURE examples of what your assembly is supposed to look like and teach CURE what the assembly is, CURE tells you what it sees. You no longer need to develop a machine vision algorithm based on image processing and measurements in a highly controlled lighting environment, which saves you time and reduces development costs. CURE captures subjective knowledge and automates the expert.
Figure 2. Example Application – Inspecting Automobile Taillights