Contact Us
  • This field is for validation purposes and should be left unchanged.

Contextual Reasoning for Object Identification

TECHNOLOGY AREA(S): Information Systems, Sensors OBJECTIVE: Develop a technique to incorporate variable contextual information to aid object identification and target designation. DESCRIPTION: When dealing with well-understood threats in a clean environment, a simple formula using a previously defined set of sensor features may be adequate to identify the threat object. However, when encountering novel threats or complex scenarios, a greater capacity to reason with the scene and its environment may be needed. For example, using a broad range of marginal information and behaviors to guide classification logic for the system may be needed. Additionally, expanding individual object identification to include information about all objects in the scene and reasoning on the whole could help resolve the true classification.The purpose of this topic is to develop a method to utilize all available, relevant information obtained by multiple sensors to aid decision making for object selection. This analysis should involve all tracked objects and their respective features, as well as environmental information, or any type of information which could influence belief in the value of a target. The focus should be on the underlying logic, or calculus, that supports reliable generalization from possibly limited data. The developed technique should be robust to sensor or feature drop-outs and able to provide a system for real time decision making with variable information. In particular, the developed technique should enable reasoning as to which tracked object(s) in a missile complex should be targeted. This approach could utilize Bayesian statistics, probabilistic generative models, probabilistic programming or any reasonable approach which considers the entire engagement. This effort should be able to analyze existing data to learn patterns and structures, as well as to provide a system for real time decision making with variable information.Recent research into cognitive science has produced representational systems and computational formalisms that may enable the BMDS to more effectively make decisions in novel situations. Static decision paradigms that classify an object with respect to a fixed set of features from a given set of sensors may not be adequate in real time for a highly uncertain engagement where sensors may be unavailable and features may be corrupted.An innovative method to reason with the battlespace scene as described by multiple sensors is sought. It should be assumed that there are two sensors reporting for the baseline, either two radars, or one radar and one space-based EO/IR sensor. The designed system should demonstrate functionality in the case where one sensor drops out, or various types of corruption or confusion are introduced. PHASE I: Develop and demonstrate, through proof-of-principle tests, a technique to combine information from multiple sources to identify the target of interest. The system should demonstrate robustness when the scene is degraded, a sensor is lost, and/or features are corrupted. The technique should demonstrate the ability to reason with the scene and use auxiliary information for target determination. PHASE II: Refine and update concept(s) based on Phase I results and demonstrate the technology in a realistic environment using agency provided engagements. Demonstrate the technology?s ability in a stressed environment; with few sensors and many targets with countermeasures. PHASE III DUAL USE APPLICATIONS: Demonstrate the new technologies via operation as part of a complete system or operation in a system-level test bed to allow for testing and evaluation in realistic scenarios. Market technologies developed under this solicitation to relevant missile defense elements directly, or transition them through vendors.Potential commercial and military uses include areas such as intelligence gathering and analysis, supply chain distribution logistics, automated processing, robotics, and manufacturing processes. KEYWORDS: Contextual Reasoning, Cognitive Reasoning, Machine Learning, Target Identification, Target Characterization, Computer Vision POINT OF CONTACT: Karla Spriestersbach, Phone: 256-450-3289, Email: karla.spriestersbach@mda.mil

  • Agency: Missile Defense Agency,Department of Defense,Department of Defense
  • Program: STTR
  • Phase: Phase I
  • Release Date: August 27, 2015
  • Open Date: September 28, 2015
  • Close Date: October 28, 2015
  • URL: https://sbir.defensebusiness.org/topics
Contact Us
  • This field is for validation purposes and should be left unchanged.