The work aims to design a robust image watermarking technique based on carrier visual content analysis and genetic algorithm optimization by means of DCT (Discrete Cosine Transform), in order to balance the imperceptibility and robustness of the watermark image. Firstly, the carrier image was divided into a series of non-overlapping subblocks, and the singular value decomposition mechanism was introduced to define the visual content analysis method for obtaining the active factor of each subblock, so that the subblock with the larger active factor was viewed as the watermark embedding position. The DCT method was used to process the subblocks with larger active factor, in order to get the corresponding DC coefficients. Then, combined with the SSIM, PSNR and NC, the fitness function was designed based on weight factor. The optimal embedding strength was found by implementation of genetic algorithm. According to the optimized embedding strength, the watermark embedding method was constructed to hide binary watermark into these DC coefficients, and the watermark image was outputted through inverse DCT transform. Finally, the watermark detection method was designed to extract binary watermark from watermark image. Experimental data showed that, compared with the current watermarking technology based on transform domain, the proposed algorithm had higher visual imperceptibility and anti-geometric attack ability. The PSNR and NC values of the proposed algorithm were kept over 45 dB and 0.96 respectively in the face of various geometric attacks. The proposed algorithm can better hide watermark information into the carrier, and has higher visual imperceptibility and anti-geometric attack ability, which has better reference value in copyright protection, information security and other fields.