This project proposes a fully automatic object detection technique for marine images, which enable the non-invasive monitoring of species overcoming the problems associated with the traditional tow-net based approaches. The proposed approach introduces a GMM based colour modelling technique to represent colour distributions of image background. The GMM background modelling method successfully partitions an image into foreground and background groups. With the segmentation, blob analysis is applied to foregrounds for individual recognition, where blob features bounding box, compactness and circularity are calculated for each blob, and feature selection criteria are presented.
• Automatic target detection and recognition system that uses data from an airborne three dimensional imaging laser radar (LADAR) sensor. The automatic target recognition system uses size signatures from target models to detect and recognize targets under heavy camouflage cover in extended terrain scenes. But this technique requires predefined target dimensions and characteristics (target library) which is impractical to be applied in seabed features.
• A threshold technique on morphological gradients is applied combining with a background difference thresholding scheme
• It needs predefined objects to check its accuracy which is difficult to apply on seabed objects.
In our proposed system, after the noises are removed from the image, it is segmented by using Gaussian Mixture Model(GMM) and features are extracted by Blobs analysis. Then performance of the system is evaluated.
• It quickly compute features, and progressively reduces the computational load.
• Accuracy is improved compared to other methods.
Processor Type : Pentium -IV
Speed : 2.4 GHZ
Ram : 128 MB RAM
Hard disk : 20 GB HD
Operating System : Windows 7
Software Programming Package : Matlab R2014a
B.H. Robinson, The coevolution of undersea vehicles and deep-sea research. Marine Echnology Society Journal, Vol. 33, pp. 69-73, 2000.