As the primary goal of the 17 Sustainable Development Goals (SDGs), poverty eradication is still one of the major challenges faced by countries around the world, and relative poverty is a comprehensive poverty pattern triggered by the superposition of economic, social, and environmental dimensions. Therefore, Therefore, this paper introduces the perspective of coupled coordination to consider the formation of relative poverty, constructs indicators in three major dimensions: economic, social, and environmental, proposes a fast and more accurate method of identifying relative poverty in a region by using machine learning, measures the degree of coupled coordination of China’s relatively poor provinces using a coupled coordination model and analyzes the relationship with the level of relative poverty, and puts forward suggestions for poverty management on this basis using typology classification. The results of the study show that: 1) the fusion of data crawlers, remote sensing space, and other multi-source data to construct the dataset and propose a fast and efficient regional relative poverty identification method based on big data with low comprehensive cost and high identification accuracy of 0.914. 2) Currently, 70.83% of the economic-social-environmental systems of the relatively poor regions are in the dysfunctional type and are in a state of disordered development and malignant constraints. The regions showing coupling disorders are mainly clustered in the three southern prefectures of Xinjiang, Qinghai, Gansu, Yunnan, and Sichuan, and their spatial distribution is relatively concentrated. 3) The types of poverty and their coupled and coordinated development in each region show large spatial variability, requiring differentiated poverty eradication countermeasures tailored to local conditions to achieve sustainable regional economic-social-environmental development.