The millimeter wave (MMW) radar has good compromise characteristics of both microwave radar and optical sensors. It has better angular and range resolving abilities than microwave radar, and a longer penetrating range than optical sensors. We used the MMW radar to detect targets located in the sea and among sea ice clutter based on fractals, wavelets, and neural networks. The wavelets were used as feature extractors to decompose the MMW radar images and to extract the feature vectors from approximation signals at different resolution levels. Unsupervised neural classifiers with parallel computational architecture were used to classify sea ice, sea water and targets based on the competitive learning algorithm. The fractal dimensions could provide a quantitative description of the roughness of the radar image. Using these techniques, we can detect targets quickly and clearly discriminate between sea ice, sea water, and targets.

Fractals provide a good description of natural scenes and objects based on their statistically self-similar property. They are also used to discriminate natural or man-made objects because natural objects have a better fitting to the fractional Brownian motion (fBm) model than artificial objects. Sea clutter as natural phenomena well fit to the fBm to induce little error. On the other hand, targets as man-made objects induce much more error because they frequently deviate from the fBm model. Therefore, the fractal error has a good characteristic to detect targets buried in clutter. We modified the fractal error defined by Cooper to be suitable for radar image processing. For the X-band radar image, the performance of our proposed method is comparable to that of the Cooper's method. For the millimeter wave (MMW) radar images, our method is better than the Cooper's one.

A novel algorithm associated with fractal preprocessors, wavelet feature extractors and unsupervised neural classifiers is proposed for detecting radar targets embedded in sea ice and sea clutter. Utilizing the advantages of fractals, wavelets and neural networks, the algorithm is suitable for real-time and automatic applications. Fractal preprocessor can increase 10 dB signal-to-clutter ratios (S/C) for radar images by using fractal error. Fractal error will make easy to detect radar targets embedded in high clutter environments. Wavelet feature extractors with a high speed computing architecture, can extract enough information for classifying radar targets and clutter, and improve signal-to-clutter ratios. Wavelet feature extractors can also provide flexible combinations for feature vectors at different clutter environments. The unsupervised neural classifier has a parallel operation architecture easily applied to hardware, and a low computational load algorithm without manual interventions during learning stage. We modified the unsupervised competitive learning algorithm to be applicable for detecting small radar targets by introducing an asymmetry neighborhood factor. The asymmetry neighborhood factor can provide a protective learning to prevent interference from clutter and improve the learning effects of radar targets. The small radar targets in Millimeter wave (MMW) and X-band radar images have been successfully discriminated by our proposed algorithm. The effective, efficient, high noise immunity characteristics for our proposed algorithm have been demonstrated to be suitable for automatic and real time applications.

A new algorithm based on wavelets and neural networks is proposed for discriminating oil leaks using synthetic aperture radar (SAR) images. Utilizing the advantages of wavelets and neural networks, the algorithm is speedy and effective to distinguish oil embedded in both sea clutter and land clutter. The iterative algorithm uses a wavelet feature extractor and two unsupervised neural classifiers. The first stage classifier can divide the pixels in the SAR image into sea water, land and oil clusters. In the second stage, the classifier extracts oil pixels from previous oil cluster until matching the characteristics of the oil template. Using our proposed algorithm, the oil cluster will be formed automatically, provided the desired oil template is defined in advance.