The invention proposes a rope operating method based on visually self-supervised learning and imitation which comprises main contents as follows: a rope operating mechanical system structure, index liner unit depth convolution learning network, a rope picking and placing module and rope picking and placing simulation. The process is as follows: firstly, transforming a robot mechanical system platform, allowing picking and placing operations, then contrasting manually-controlled images in real world with robot self-controlled images; inputting into the depth neurolearning network to do matching training to generate an inverse model to predict actions; ensuring that the configuration meets manual requirement in the repeated training process. According to the rope operating method based on visually self-supervised learning and imitation, the operation of soft material objects such as ropes and the like can be handled, and the depth neurolearning network is provided to generate the inverse dynamic model, so that the accuracy of operating ropes by robots is improved.