You are interested in deeper analyses and extracting more useful information from the spectral image data? The perClass Mira extension for Cubert Utils allows you to add a powerful image classification solution to the live data stream, without the need of understanding complex classification methods. Retrieve relevant information from your video stream faster than ever before. Mira is of course available for all Cubert camera models.
Mira is a software interface for analyzing hyperspectral image data. Based on a powerful machine learning engine Mira selects the best statistical model for given labeled samples fully automatically.
Mira helps creating specific solutions for image analysis that can be applied to Cubert’s hyperspectral snapshot cameras, such as the ULTRIS 20, in live mode.
ULTRIS 20 image of a hilly landscape in colored infrared (left) and segmented by Mira (right). The semantic information is displayed in different colors.
Easy-to-use image classification & regression tool
Automated model building
Pixel classification & object segmentation
Applicable to the hyperspectral live data stream
Food quality control
Plant disease detection
Medical and pharmaceutic assessments
Supervision of production processes
Suited for full machine vision integration
Time-critical and non-destructive assessments in any industrial environment
Gathering hyperspectral images is only the first step of solving a problem. Analyzing the data can be challenging and time-consuming. It typically requires software development and a high level of expertise in applied statistics, math and remote sensing.
Mira enables anyone to create an interpretation solution timesaving and without needing this expertise – (1) record spectral data with your Cubert camera, (2) transfer this data to Mira and train as well as evaluate an efficient classifier for specific materials, and (3) apply this classifier to the live data stream as plugin in the Cubert Utils software.
With Mira users can quantify the composition of the hyperspectral image fully automatically. Furthermore, Mira allows you to qualify the samples in the image. Check if your fruits have rot, identify foreign objects during your production process or evaluate the degree of skin burns – within seconds!
Import your hyperspectral image to Mira. Define multiple classes for your samples. Label those classes in the image allowing Mira to learn on this training data. Mira analyzes this data and creates a classification using machine learning.
Mira assigns each pixel to the predefined classes. The composition of your samples in the hyperspectral image is gathered fully automatically. Validate the result of the trained classifier and fine-tune the model interactively.
Mira is also able to count the number of different objects, i.e. connected components, in the image. Set a threshold for the size to adapt to your individual application. Classify the pixels within an object and let Mira identify outliers.