Aiming at the problem that there is no applicable flexible wire harness recognition method in the current wiring automation production process, a flexible wire harness recognition method for wiring robots is proposed. Firstly, based on the UNet network structure, ResNet-34 is adopted as the feature extraction network in the encoder, and the SE attention module is introduced in the decoder part to construct a new image segmentation model RS-UNet for the segmentation of wire harness images. Then, the Zhang-Suen image refinement algorithm is used to refine the wire harness mask image, and finally the geometric center position information of the wire harness is obtained to give guidance to the wiring robot to operate the wire harness. Through experiments, it is proved that the RS-UNet network improves the IoU value by 4.95% and the F1 value by 0.029 in the wire harness segmentation accuracy compared to the when -UNet network, and the selected Zhang-Suen refinement algorithm has an average processing time of 0.38 s, an average refinement sensitivity of the image refinement result of 0.39, and an average thickness reference value of 0.87, the proposed method can accurately identify the geometric center of the flexible wire harness.