Naiman Banner is located in an arid and semi-arid region and represents a typical agro-pastoral ecotone. Agricultural production in this area has long been constrained by complex natural conditions and the fragmented spatial distribution of cropland, resulting in an unclear understanding of the current state of farmland utilization. This, in turn, limits the region’s capacity for agricultural resource management and targeted policy implementation. Conducting an analysis of farmland utilization in Naiman Banner is therefore crucial for promoting high-quality agricultural development in agro-pastoral transitional zones and supporting the national food security strategy. In this study, Naiman Banner was selected as the research area. Based on Landsat 8 time-series imagery, the spatial distribution of major grain crops from 2020 to 2024 was extracted using the random forest method. Additionally, Jilin-1 high-resolution remote sensing imagery was used in combination with a multi-scale segmentation algorithm and the U-Net model to obtain cropland field parcels and road data for 2020 and 2024, respectively. On this basis, an evaluation index system for assessing the level of farmland infrastructure construction in the region was established. By integrating the extracted spatial distribution of major grain crops, this study focuses on analyzing the spatiotemporal patterns of major grain crops, the level of farmland infrastructure construction, and their interrelationship in Naiman Banner. Results showed a steady increase in the planting area of major grain crops over the five-year period, with a cumulative net growth of 22 673.46 hm², representing a 12.9% increase. Planting areas gradually shifted from general farmland to well-facilitated farmland. Meanwhile, agricultural infrastructure was continuously improved, with significantly higher development levels observed in well-facilitated farmland areas compared to general farmland. Further analysis revealed that regions with higher levels of agricultural infrastructure exhibited higher utilization rates of staple crops, demonstrating clear spatial consistency and a positive correlation. These findings provide provide technical support and decision-making references for regional agricultural resource management, optimization of infrastructure allocation, and food security assurance.