Convolutional neural networks(CNNs) have become powerful segmentation models, but are usually designed manually, which requires extensive time and can result in large and complex networks. There is a growing enthusiasm for automatically designing efficient architectures that can accurately segment domain-specific images. However, most approaches either do not consider building more flexible network architectures or do not consider multiple objectives to optimize model. To cope with these issues, we propose an adaptive multi-objective evolutionary convolutional neural architecture search algorithm called AdaMo-ECNAS for domain-specific image segmentation, which considers multiple performance metrics during the evolutionary process and adapts to a specific dataset by optimizing multiple objectives. AdaMo-ECNAS can build flexible and versatile predictive segmentation models whose architecture and hyperparameters are found by a multi-objective evolutionary algorithm that is adaptively PBI-based tri-objective evolutionary to improve the $F_1$-score of predictive segmentation, minimize the computational cost and maximize the additional training potential. The main contribution of this work is a fully exploited feature information of segmentation and automatic search for high-performance and efficient architectural models. AdaMo-ECNAS is evaluated on two real world datasets, and the proposed method is competitive and even superior to some advanced algorithms.