Surveillance Method and System

About

Surveillance Method and System for Discrimination of Target versus Fixed Objects Ref #: 10A0055 Lead Inventor: Ioannis Stamos Background: Laser range scanners have now the ability to acquire millions of 3-D data points of highly detailed and geometrically complex urban sites. This depth of data opens new avenues of exploration in modeling of urban environments and in surveillance generally. However, the collection of high-resolution “point-clouds” is currently a slow process (in the order of hours) requiring range scans to be processed off-line after acquisition. This process is a bottleneck for real-time instantaneous decision making, such as is critical for area surveillance. Thus a system supporting real-time rapid classification of range scans so as detect and announce the spatial location of abrupt changes in the surveillance environment is most desirable. Invention: This invention describes a surveillance system for discrimination of target objects versus fixed objects. More specifically, this provides an automated classification system with highly efficient data acquisition, high-level data recognition and rapid real-time decision-making. The approach uses novel real-time sequential on-line processes to provide (a) the reliable classification between vegetation and non-vegetation data, via a novel hidden Markov model formulation. [Note that range data in the areas of vegetation (trees, etc.) pose significant challenges to current segmentation algorithms], (b) the reliable classification between vertical, and horizontal surfaces in the non-vegetation data regions, including a regiongrowing algorithm that uses the results of the above classifiers and reliably separates the data into connected regions of vertical and horizontal surfaces, and (c) the presentation of such process results based on alarm conditions according to selected criteria. Applications: Border Surveillance System with reliable classification between vegetation and nonvegetation, and between vertical and horizontal surfaces, with real-time display of situational update data. Detection system for abrupt changes. Real-time detection and recognition of objects in a 3D-scene, as well as 3D modeling. Easy extension to an arsenal of sequential algorithms that provide classifications of different types. Advantages: More intelligent sensor Higher-level recognition Real time detection Faster decision process Robust results Developmental Stage Prototype has been built and tested. The results were very accurate, with minimal misclassification. Market Large and growing homeland security market. IP pending. Licensing available.