A new training approach based on ECOC-SVM for SAR image retrieval
G. Siva Krishna, N. Prakash · International Journal of Intelligent Enterprise · 2021
Synthetic aperture radar (SAR) is a high-resolution remote sensing imagery which is used in environment monitoring, earth-resource mapping and military systems. The objective of the paper is to perform the image retrieval of the SAR by using the error correcting code based support vector machine (ECOC-SVM) and denoising of SAR images. The features from the SAR images as texture, colour and shape are extracted by using different techniques such as texture spectrum (TS), grey level difference method (GLDM), scale-invariant feature transform (SIFT) and hue, saturation, value (HSV) model. The bi orthogonal wavelet transform (BWT) with particle swarm optimisation (PSO) is used for optimising the soft threshold for denoising the SAR images. The result is analysed with one existing method named IIRM in terms of average accuracy of proposed the SAR image retrieval and precision 98.438%, 70.31% is high for ocean image compared with the average precision of IIRM method that is 63.5%.