The task of detecting mine-like objects (MLOs) in side scan sonar imagery has a profound impact on military operations. The current process involves subject matter experts analyzing sonar images searching for MLOs. The automation of this problem has been heavily researched over the years without a definitive solution that outperforms the manual approach in real world scenarios. This paper presents a series of approaches and experiments centered on the use of GentleBoost feature selection classifiers for the detection of MLOs in side scan sonar. In a comparison of semi-synthetic versus real world training data with two different boosted single-feature selection classifiers, we see that semi-synthetic data can provide insight in to potential performance of a classifier. We run experiments training and testing GentleBoost single-feature classifiers on six different feature types, finding that the Haar-like feature classifier performs the best. We propose a GentleBoost multi-feature selection framework that allows for multiple feature types to be in the pool of selectable features, finding that a combination of Haar-like features, speeded up robust features (SURF), and simple shadow features performs better than the Haar-like feature classifier. Experiments with tiered, or cascaded, classifiers show a reduction in false positives for lower true positive rates. The multiple instance learning (MIL) approach shows great potential for future efforts, achieving improved true positive rates at higher false positive rates. A final approach considers the complimentary benefits of computer vision and human vision, introducing two brain- computer interface (BCI) systems. One BCI uses the Haar- like feature classifier as a first stage cascaded in to a human processing second stage. The other adds a third stage that employs a novel support vector machine (SVM) classifier based on the Haar-like feature and human interest scores from multiple subjects. Overall, our GentleBoost feature selection classifier variations result in performance improvement for the detection of MLOs in side scan sonar imagery

In the world of autonomous underwater vehicles (AUV) the prominent form of sensing has been sonar due to cloudy water conditions and dispersion of light. Although underwater conditions are highly suitable for sonar, this does not mean that vision techniques should be completely ignored. There are situations where visibility is high, such is in calm waters, and where light dispersion is not an issue, such as shallow water or near the surface. In addition, even when visibility is low, once a certain proximity to an object exists, visibility can increase. The focus of this project is this gap in capability for AUVs, with an emphasis on computer-aided detection through machine learning and computer vision techniques. All experimentation utilizes the Stingray AUV, a small and unique vehicle designed by San Diego iBotics. The first experiment is detection of an anchored buoy, which mimics the real world application of mine detection for the Navy. The second experiment is detection of a pipe, which mimics pipes in bays and harbors. The current algorithm for this application uses boosting machine learning on hue, saturation, value (HSV) to create a classifier followed by post processing techniques to clean the resulting binary image. There are many further applications for computer- aided detection and classification of objects underwater, from environmental to military

The detection of mine-like objects (MLOs) in sidescan sonar (SSS) imagery continues to be a challenging task. In practice, subject matter experts tediously analyze images searching for MLOs. In the literature, there are many attempts at automated target recognition (ATR) to detect the MLOs. This paper focuses on the classifiers that use computer vision and machine learning approaches. These techniques require large amounts of data, which is often prohibitive. For this reason, the use of synthetic and semisynthetic data sets for training and testing is commonplace. This paper shows how a simple semisynthetic data creation scheme can be used to pretest these data-hungry training algorithms to determine what features are of value. The paper provides real-world testing and training data sets in addition to the semisynthetic training and testing data sets. The paper considers the Haar-like and local binary pattern (LBP) features with boosting, showing improvements in performance with real classifiers over semisynthetic classifiers and improvements in performance as semisynthetic data set size increases.

Detection of mine-like objects (MLOs) in sidescan sonar imagery is a problem that affects our military in terms of safety and cost. The current process involves large amounts of time for subject matter experts to analyze sonar images searching for MLOs. The automation of the detection process has been heavily researched over the years and some of these computer vision approaches have improved dramatically, providing substantial processing speed benefits. However, the human visual system has an unmatched ability to recognize objects of interest. This paper posits a brain-computer interface (BCI) approach, that combines the complementary benefits of computer vision and human vision. The first stage of the BCI, a Haar-like feature classifier, is cascaded in to the second stage, rapid serial visual presentation (RSVP) of images chips. The RSVP paradigm maximizes throughput while allowing an electroencephalography (EEG) interest classifier to determine the human subjects' recognition of objects. In an additional proposed BCI system we add a third stage that uses a trained support vector machine (SVM) based on the Haar-like features of stage one and the EEG interest scores of stage two. We characterize and show performance improvements for subsets of these BCI systems over the computer vision and human vision capabilities alone.