The brightness and contrast of the image are usually low, which makes the target information blurred and increases the difficulty of recognition. Aiming at the problem that the traditional recognition methods are not accurate in the recognition of multiple salient objects, a new recognition method of salient objects in low-light images based on double-branch convolutional neural network is proposed. Image gray processing and denoising processing are implemented for low-light images. The low light image is enhanced, the low light problem is adjusted, and the salient object features of low light image are extracted by using double branch convolutional neural network. The experimental results show that: under the application of the proposed method, no matter how many significant objects exist in the image, the Kappa value is above 0.8, with high accuracy.