Alteria Lasers: Readability Detection

Output display
Output display
Machine Learning model detecting readability
Machine Learning model detecting readability

Role: Product developer and Computer Vision Engineer, Freelancer 

ALTERIA LASER is a technology company engaged in the development of efficient, high performance and affordable Laser marking, coding and materials processing systems.

The company wanted a solution by which they can determine the readability of the letters printed by their Lasers.

I developed an innovative solution for this using Beagle Bone Black, Computer Vision, and Machine Learning to determine the readability of each character and then the quality index of the whole text printed on the products. Whenever quality index drops from a certain value the system generates an alarm indicating that the Lasers are not printing properly. I also developed a customized Operating System using Yocto for the Single Board Computer.

The first image on the left represents the touch screen display of the device. It includes some controls and outputs. It displays the image of the product containing the text area printed by the Lasers and the Quality Index calculated by the algorithm. If the quality index is above a particular threshold the output is 'OK' else it raises a Waring. The device first captures the image of the product using the camera attached to the Single Board Computer and then progress it for the region of interest (which is the text printed by the Laser). Then it iterates over each character to determine it's readability and then the quality index.

The second image demonstrates how the machine learning model was determining the quality of each character. In this example, the readability of characters starts decreasing with every cut in the character.

Sudhir Jain