A novel wing icing area recognition method based on morphological processing enhanced U-Net network is presented in this paper so as to solve the problem of complex boundary recognition. The wing icing image sample sets are obtained by self-fabricated icing wind tunnel equipment under the low temperature surrounding in Jilin city, China. The closed operation of morphological processing is used to dilate and erode the pixel value of the image and added into the U- Net network to improve the anti-interference ability under the complex boundary condition. The novel morphological processing enhanced U-Net network, and the full convolution network (FCN), are established respectively, and compared output results with each other under the same input wing icing image samples. The experimental results show that, the recognition accuracy of morphological processing enhanced U- Net network for training sets and testing sets are 97.78% and 92.06%, and are higher than that 94.42% and 90.73% of the FCN, respectively. The recognition accuracy of the morphological processing enhanced U-Net network is 85.61%, and higher than that 75.23% of the FCN under the condition of small wing icing image sample set. The recognition recall rate of the morphological processing enhanced U-Net network is 85.05%, higher than that 80.13% of the traditional U-Net network. So, the recognition accuracy and recall rate of the morphological processing enhanced U-Net network are superior to those of the FCN and the traditional U-Net network.