In order to recognize the recovery of floating objects on water surface by unmanned sweeper, a recogition model is proposed based on an improved residual neural network. The model improves the network performance by adjusting the structure of the origional residual neural network model, increasing the convolution output appropriately, and reducing residual unit, so as to improve the ability of the residual neural network to extract objection features. Experimental results show that the proposed model has a better recognition performance than the exist residual network both in the self-made MyFloats dataset and CIFAR-10 dataset.