Degree of familiarity ART2 in knowledge-based landmine detection

Abstract

The self-organizing network ART2 is extended to provide a fuzzy output value, which indicates the degree of familiarity of a new analog input pattern to previously stored patterns in the long-term memory of the network. The outputs of the multilayer perceptron and this modified ART2 provide an analog value to a fuzzy rule-based fusion technique which also uses a processed polarization resolved image as its third input. In real-time situations these two classifier outputs indicate the likelihood of a surface landmine target when presented with a number of multispectral and textural bands. Due to the modifications in ART2 this updated alternative architecture (to that of a previous network in [3]) has improved real-time landmine detection capabilities although the registration of all bands is more critical to the accuracy of results in this case. The real-time fuzzy rule-based system in preliminary tests has detected two of the three landmines and the landmine surrogate with two false alarms. Advanced tests on 30 images using the fuzzy rule-based system further confirmed the distinct advantages of fusion and improved detection rates.

Sensor fusion for automated landmine detection on a mobile robot

“…Then, the landmine detection is usually considered as a classification with two classes: landmines and background. Different classifiers and feature extraction methods were used in previous works: Bayesian classification is used in Cremer, Jong, Schutte, Yarovoy and Kovalenko (2003), Zhang et al (2003), Collins et al (2002), Collins et al (1999), Gao et al (1999); support vector machines in Torrione and Collins (2004), Zhang et al (2003); Dempster-Shafer classification in Perrin et al (2004), Milisavljevi et al (2003), Milisavljevi et al (2000); neural networks in Stanley et al (2002), Filippidis et al (1999) and Sheedvash and Azimi-Sadjadi (1997), etc. In some works the different methods are compared, however, a definite conclusion about the most suitable approach is not made.…”

Image indexing and retrieval using an ART‐2A neural network architecture

“…This architecture has the following features: noise filtering, good computing and classification performance. The ART neural network family has been used in several domains, such as to recognize Chinese characters (Gan and Luan, 1992), interpretation of data from nuclear reactor sensors (Keyvan and Rabelo, 1992;Davis, 1993, 1996), image processing (Vlajic and Card, 2001), detection of earth mines (Filippidis et al, 1999), treatment of satellite images (Carpenter et al, 1997) and robot control (Bachelder et al, 1993).…”

Automatic Text Classification Using an Artificial Neural Network

“…Out of the different versions of networks ART, the architecture ART-2A may be highlighted as it allows the quick learning of the input patterns represented by continuous values. Because of its attractive features, such as noise filtering and good computing and classification performance, the neural ART network family has been used in several domains, such as to recognize Chinese characters Gan and Lua, 1992, interpretation of data originated on nuclear reactor sensors Whiteley and Davis, 1996;Whiteley and Davis, 1993;Keyvan andRabelo, 1992, image processing Vlajic andCard, 2001, detection of earth mines Filippidis et al, 1999, treatment of satellite images Carpenter et al, 1997 and robots sensorial control Bachelder et al, 1993.…”