Image Search Technology: Mimicking Human Sight to Mine Image Content

By Hal Reid

_A few weeks ago I attended a piXlogic webinar presented by CEO Joe Santucci and moderated by Mike Shrader from the piXlogic Intelligence team. piXlogic is a software company that develops products to recognize content in imagery and video. The software allows you to see and understand data that reside within imagery and video through intelligent searches and comparisons.

Most of us look at data as attributes in a geographic or tabular representation. With piXlogic’s products, you look at data within an image. For example, text within imagery is stored as patterns, making it language independent.

(Click for larger image)

The system also automatically indexes imagery and video stored in multiple databases using the piXserve database for metadata. The software can also accommodate databases that use keywords for locating imagery and video.

(Click for larger image)

A continuation of this representation of the system is shown in the image below.

(Click for larger image)

The process works like this. The target image is analyzed; a search is formulated; the database is searched, and the results are shown as they appear in the right-hand portion of the image above.

The system, after converting the imagery to vectors, stores the data in a manner similar to storing spatial data in modern databases. It “sees” the imagery itself as geography within the extent of the image, in a similar manner to a parcel or other geographic file.

piXlogic includes face recognition, as well. The software lets you look at a digital image, select a part of it and search for a match. Matching usually returns a series of images, with the best match first. The reasons for this are not just to increase the number of possible choices, but also to help identify data adjacencies that may contain relevant information.

Adjacent results were demonstrated with a search using a single image of a person’s face. The images returned showed the same person with a patch over one eye and a big mustache. This demonstrated the ability to see past the obvious and use the patterns from the original image as the basis for the search.

Searches can also be performed by selecting just part of an image (drawing a box around it) and then searching for imagery matching the selection. You can search using the foreground of the image or even ask for similar backgrounds. For example, you could look for similar rock walls, grass or trees. This lets you look for similar locations or different images of the same location.

All this works with video as well, as selections can be made frame-by-frame or by selecting part of the image from within a video frame. Text in images can be searched, too. Text is stored as a pattern, so it is not necessary to translate the text to initiate a search.

This product is obviously used in the intelligence field – but it is also used in other industries as well. The webinar covered just some of the software’s applications. In business, you could use piXlogic for facility access applications that are better, according to the company, than what can be offered using identification badges. Another application might be to search for similarities in packaging, facilities or product presentations.

Published Thursday, November 9th, 2006

Written by Hal Reid

If you liked this article subscribe to our bimonthly newsletter...stay informed on the latest geospatial technology

Sign up

© 2017 Directions Media. All Rights Reserved.