Example result.
Algorithms for fast detection of objects, using a commercial range imager. A sequential search of selected sparse 3D points is made to find a chain of points along the jump boundary and compute a coded description of shape.

TSLS is a project motivated by the need to provide real time protection of moving convoys from attack by ambush with small arms. Our approach is to develop a 3D machine vision system which can detect the weapons by shape. Rocket propelled grenades (RPGs) and similar weapons are a great threat to our forces, in part because of their low cost and easy availability to our asymmetrical adversaries. We seek to provide an effective means to detect visible weapons of this kind before they are fired. Also, this technology is expected to substantially advance the state of the art in the motion tolerant high resolution 3D imaging of large spaces, and in the efficient detection of objects based on surface shape. Potential applications include the identification of unexploded munitions, face recognition and robot guidance. The technology is intended to be applicable very broadly to the automated perception of 3D objects.

This work has two parts; a system for generating 3D images of the surrounding scene, and a software system for analyzing these images to detect and locate objects of interest.

Concept of fielded TSLS system. Surroundings are systematically searched for the characteristic shapes of ambush weapons, even in dark. Affine invariants of small clusters of spots facilitate live calibration as cameras are steered.
Concept of fielded TSLS system. Surroundings are systematically searched for the characteristic shapes of ambush weapons, even in dark. Affine invariants of small clusters of spots facilitate live calibration as cameras are steered.

The imaging system consists of three steerable cameras and a special flashed projector emitting a fine irregular pattern of bright spots. It uses a xenon flashlamp and a chrome/glass photomask bearing the spot pattern. Each spot is detected by the cameras and processed to produce a 3D point in space. Fig. 1 illustrates the deployment of these imaging system elements in the field. It is expected to function at a distance greater than 100 meters and scan most of the visible scene area as it flows by.

We are exploring several methods for detecting shapes of interest in the 3D image data. One (see Fig. 2) is called “Range Scaled Chain Codes”. This is useful for interpreting 3D images when the depth coordinate is measured less accurately than the two transverse coordinates. The data is probed sequentially in such a way that deep-shallow pairs of points are discovered, indicating a jump boundary. Then the boundary is followed, forming a chain of fixed length segments. The angles between these segments contain information about the shape of the object, and can be used to match them quickly against known models.

Contact:
Dr. Frank Pipitone
Perceptual Systems, Code 5516
Naval Research Laboratory
Washington DC 20375
Email: w5514 "at" itd.nrl.navy.mil

Key publications

F. Pipitone, Gilbreath, C., and Bonanno, D., “Tripod Operators for Efficient Search of Point Cloud Data for Known Surface Shapes”, SPIE: Active and Passive Signatures, vol. 8382. 2012.
F. Pipitone, “Efficient RPG detection in noisy 3D image data”, SPIE: Active and Passive Signatures, vol. 8040. 2011.
F. Pipitone, “Manifold Learning for Compression and Generalization of Euclidean Invariant Signatures of Surface Shapes”, SPIE: Active and Passive Signatures, vol. 7687. SPIE, 2010.
F. Pipitone, Hartley, R., and Creamer, G., “A Structured Light Range Imaging System Using a Moving Correlation Code”, Proceedings of the Third International Symposium on 3D Data Processing, Visualization and Transmission. IEEE, Chapel Hill, pp. 931-937, 2006.PDF icon Download PDF (395.47 KB)