By signing into the Stream Analytics site, everyone has access to use or create Analytics Apps to detect items that have value to their specific business.
Stream utilizes a simple framework of Capture/Analyze/Visualize to best understand how it works.
There are two basic ways to capture spectral data:
- Take a picture using a multispectral camera
- Take a reading/scan using a spectrometer
- Any Ocean Optic spectrometer (It has to be Seabreeze API supported)
- Any spectrometer that uses Open Source "Spectral Web Service"
- ColorFlow™ Camera
- LabFlow Soil tester
Cloud Connect is the link from the capture device to the Stream Analytics Engine. With Cloud Connect, all spectral data is automatically uploaded to the Stream Analytics Engine to your own account for analysis to begin.
For more on Cloud Connect click here
Once the supported hardware is linked to your account, you are ready to capture the spectral data.
A key component of the process is the Stream Analytics Engine. This engine combines machine learning techniques and neural nets designed specifically to show the test results from spectral images and spectrometer scans.
Stream Analytics Engine uses 'tests' or Analytics Apps to detect the particles or elements needed to be identified. These Analytics Applications are created by anyone who has taken pictures or scans of their physical samples as training data. Analytics Apps will be listed in the online Analytics App store and are made available for purchase.
For more information on Analytics Apps click here
This process is quick and simple. You snap a picture, or touch a sample and the results are available on your phone in your Stream account. Once the sample is uploaded (through Cloud Connect), the Stream Analytics Engine will run the Analytics App and within minutes you will be able to see false colored images or scans in your Stream Account on any browser.
Results are typically false colored images or accuracy levels of the detected element. If you are building a new Analytics App, your results show up the same way along with other meta data, all designed to create accurate detection.
Example of Detecting Organic vs Polyethylene Leaves
The first question you have to ask yourself, is what item or object do you need to detect?
In this example of classification, we are looking to decipher the difference between Organic and Polyethylene leaves.
We begin by selecting the objects that we would like to identify. On the right, there is a picture of 3 different leaves with 2 of them being Polyethylene. We are looking to identify the Polyethylene leaves.
With a spectral camera, we are able to take a picture, or we can use a spectrometer to take a probe of a sample simply by touching it.
The image or data is then sent to the Stream Analytics Engine, where you should be signed in to your free account. The Stream Analytic Engine will use an Analytics App (Click here for more info) to identify the elements in your image or data.
In this case we will be using the Organic vs Polyethylene Leaves Analytics App.
If a multispectral camera is used to capture the object, the detected pixels are false colored in the photograph. With spectrometer data the test results are represented with a percentage or graph.
In this example the Polyethylene leaves have been colored red, while the Organic leaf has been colored blue. The background has been colored green.
Ready to find out more?
Contact us today!