Post-harvest quality assurance is a crucial link between grain production and end users. It is essential to ensure that grain does not deteriorate due to heating during storage. To visualize the temperature distribution of a grain pile, the present study proposed a three-dimensional (3D) temperature field visualization method based on an adaptive neighborhood clustering algorithm (ANCA). The ANCA-based visualization method contains four calculation modules. First, discrete grain temperature data, obtained by sensors, are collected and interpolated using back propagation (BP) neural networks to model the temperature field. Then a new adaptive neighborhood clustering algorithm is applied to divide interpolation data into different categories by combining spatial characteristics and spatiotemporal information. Next, the Quickhull algorithm is used to compute the boundary points of each cluster. Finally, the polyhedrons determined by boundary points are rendered into different colors and are constructed in a 3D temperature model of the grain pile. The experimental results show that ANCA is much better than the DBSCAN and MeanShift algorithms on compactness (around 95.7% of tested cases) and separation (approximately 91.3% of tested cases). Moreover, the ANCA-based visualization method for grain pile temperatures has a shorter rendering time and better visual effects. This research provides an efficient 3D visualization method that allows managers of grain depots to obtain temperature field information for bulk grain visually in real time to help them protect grain quality during storage. © 2023 Society of Chemical Industry.