In order to detect weak targets on sea surface in the background of strong sea cluttera detection algorithm based on optimized ESMD-ICA is proposed.The wavelet modulus maxima denoising algorithm is used to denoise the signalthen the ESMD-ICA denoising algorithm is optimizedand finally the ELM model improved by Gray Wolf Optimization (GWO) is used to predict the reconstructed denoised signal.IPIX radar measurement data is used to verify the new algorithm.The experimental results show that the new algorithm does not destroy the chaotic property of the signal under the premise of suppressing sea clutterand the detection performance of the new prediction model is much better than that of the traditional prediction model (BP neural network).The new algorithm performs well in denoising effectmodel training and detecting timedetector stability and so onwhich is conducive to its wide application in practical engineering.