A comparison of inter-frame feature measures for robust object classification in sector scan sonar image sequences

Abstract

This paper presents an investigation of the robustness of an inter-frame feature measure classifier for underwater sector scan sonar image sequences. In the initial stages the images are of either divers or remotely operated vehicles (ROV's). The inter-frame feature measures are derived from sequences of sonar scans to characterize the behavior of the objects over time. The classifier has been shown to produce error rates of 0%-2% using real nonnoisy images. The investigation looks at the robustness of the classifier with increased noise conditions and changes in the filtering of the images. It also identifies a set of features that are less susceptible to increased noise conditions and changes in the image filters. These features are the mean variance, and the variance of the rate of change in time of the intra-frame feature measures area, perimeter, compactness, maximum dimension and the first and second invariant moments of the objects. It is shown how the performance of the classifier can be improved. Success rates of up to 100% were obtained for a classifier trained under normal noise conditions, signal-to-noise ratio (SNR) around 9.5 dB, and a noisy test sequence of SNR 7.6 dB.

The Application of Sector-Scanning Sonar: Strategy for Efficient and Precise Sector-Scanning Using Freedom of Underwater Walking Robot in Shallow Water

“…If a sector-scanning sonar is usually installed on an underwater robot or vehicle, it is highly useful to detect objects, avoid obstacles, and acquire the necessary information for navigation [6]. By using the precision position sensor installed in a remotely operated vehicle (ROV), it is possible not only to know the position of the acquired image but also to continuously obtain the image by moving to the desired position.…”

“…The SONAR (sound navigation and ranging) system is the most widely adopted solution for remote sensing and is very useful for underwater observation and surveillance in coastal waters with poor visibility due to sediments [2][3][4][5]. Sector-scanning sonars are a device used to capture two-dimensional images and are widely used for vehicle navigation, obstacle avoidance, and general inspectional surveys of the surrounding environment by unmanned underwater vehicles (UUV) [6][7][8].…”

Multi-Feature Extraction and Fusion for the Underwater Moving Targets Classification

“…Basic object shape and intensity features. The basic object shape and intensity fearues are extracted from the segmented target images.These intra-frame features(known as static features) are grey-level and shape descriptors of the targets for a single scan.These features were previously found to give the best classification performance with fixed segmentation threshold and no added noise.There are ten features used for the classification including object area, object perimeter, object compactness, maximum dimension,object mean intensity,object intensity variance, normalization contrast1 [6] ,contrast2 [5] , range elongation [5] , and difference Ratio [5] .…”

“…The diver detection soanr is kind of sector scan sonar, the classification of the underwater targets in sector sonar are mainly based on the image features [1] .The temporal image features after the tracking association are used to improve the classification for these underwater moving tragets [2][3][4] .These features are also used for recognition of the small man made object [5] .The classification results among divers,pier lags,chains,sea bottom have been discussed using these features. [6] compares the classifcation results with or with out tracking stage by using the recurrent structures. Jae-Byung Jung [7] discusses the classification between the divers and the bottles,fishes using the Broadband active sonar systems.The features extracted from the spectral analysis after PCA are used for the classification by artificial neural network.…”

Object detection and tracking method of AUV based on acoustic vision

“…Williams [3] [4] used the object templates to analyze the features variety of objects in acoustic images and track the diver. Considering the noise interference, Chantler [5] and Ruiz [6] selected measured values of features in each time to distinct detected regions, and improved its robustness. Petillot [7] proposed an object detecting and tracking method based on a multi-beam forward looking sonar sensor.…”