Target recognition software

The application forms a hypothesis based on multi-factor inputs and compares this to range data for a higher confidence result

Software & Information Technology

There is an ongoing technology race to accurately identify objects of interest at further and further distances. Applications range from the obvious military threat reduction uses to private sector security, and many forms of remote sensing which span industries. This invention from the Navy compares synthetic data with observed data to yield a more accurate identification of a target and the range to a target.  The software model is built using input from 14 discrete sources including:

  • Sensor range image – computed from a LADAR sensor, uses location and orientation to derive GPS coordinates
  • Target information – the rendered image compared to a database of images utilizes feature extraction, recognition, and localization
  • Initial target search parameter extents – equivalent to the breadth of search settings such that if it is known that the target is from a specific class such as automobiles, the model will restrict its search to cars
  • Target model – a wiregrid, 3D model of a potential target match
  • Other parameters include the range of noise in the sensor and object modeling errors

Initially, LADAR had been rightfully questioned as a commercial tool due to its slow processing speed, high costs, and varying accuracy.  Hardware and software advances have addressed the first two of these but accuracy has remained an issue.  This system proposed by the Navy directly addresses that third concern by creation of the synthetic range image. Iterative comparison of the synthetic model to the LADAR data increases the accuracy of the result.

Those interested in this technology should also refer to US patent 8,810,779. This patent, ‘888 and the ‘779 patent share the same title and abstract.  ‘779 includes range syncing in the specification (with the mathematical definition) but not in the claims.  ‘888 has the range syncing in the claims with a mathematical definition.  ‘888 further specifies whether the matching score “C” exceeds a pre-determined threshold score and issues alerts that the target model is a match or not based on the C score.  ‘888 also includes hard values for h and w in the claims for the calculation of Rsync.

Do you have questions or need more information on a specific technology? Let's talk.

Contact Us