Computer vision and sensor fusion for detecting buried objects

Abstract

Given multiple images of the earth's surface f r o m dual-band infrared sensors, our system fuses information from the sensors t o reduce the eflects of clutter and improve the ability to detect buried or surface target sites. Supervised learning pattern classifiers (including neural networks) are used. W e present resuNs of ezpen'ments t o detect buried land mines from real data, and evaluate the usefulness of fusing information from multiple sensor types. The novelty of the work lies mostly i n the combination of the algorithms and their application t o the very important and currently unsolved problem of detecting buried land mines from an airborne standoff platform.

Sensor fusion for automated landmine detection on a mobile robot

“…An example of pixel-level sensor fusion used for IR cameras with different bands can be found in Clark et al (1993) and Clark et al (1992). The decision-level fusion is the simplest approach in terms of implementation.…”

Data fusion for vehicle-borne mine detection

“…where the model B is defined by the set of objects fO j g n j=1 which themselves contain parameters, at least one defining the object's type t j and possibly others defining other properties of the object; the latter are grouped for each object into a set j . The posterior of the object set B given a fixed segmentation S is pBjX;S = PXjB;S pBjS R PXjB;S pBjS dB ; (4) where pBjS is the prior probability density of the object set B given a fixed segmentation S. This term contains both subjective prior knowledge about the objects' parameters and about relationships between objects. The segmentation itself may be marginalised out, giving the posterior of the object set conditional only on the data, pBjX = Z pBjX;S pS dS:…”

“…Infrared imagery has proved invaluable in detecting the tell-tale thermal signatures of buried land mines [1,2]. Dual-band approaches appear to offer the prospect of more reliable operation [3,4,5]. In this regard, DelGrande has shown, in an analysis based on heat transfer theory, that dual-band infrared systems can extract accurate temperature maps by compensating for the spatial variation of emissivity.…”

A novel detection and navigation approach based on OWA fusion method

“…In contrast, electromagnetic waves (Gao and Collins, 2000; Gao et al , 1999a), X‐ray (Zhong et al , 2000), radar waves (Gader et al , 2001), infra red and thermal waves (Miller et al , 1999) can be categorized as contactless methods for mine detection. By applying sensor data fusion methods to the obtained data from the mentioned sensors, reliability and efficiency of mine detection process can be raised (Clark, 1992; Del Grande, 1990). In Metz (1992), the results of a research about an automatic mine detection vehicle without any driver are presented.…”