In the active SONAR system, the detection of a target is performed by Matched-Filter processing proceeded by a Constant False Alarm Rate (CFAR) thresholding method. CFAR alone is a not a competent method, to annihilate the invalid echoes. This method on sole does not attain the best pursuance under the reverberation scenario in the non-homogeneous condition of the acoustic environment. In this, the paper Fusion algorithm is recommended for active sonar application, where the CA-CFAR and Support Vector Machine (SVM) based classification method are together used to annihilate the invalid echoes and simultaneously improve the detection of valid echoes. In this paper, a comparison of linear and non-linear SVM with the CA-CFAR method is proposed for elevating the performance of detecting the valid echoes and annihilates the invalid echoes. The interpretation of the algorithm is accomplished using measured data.