Objectives Spatiotemporal fusion technology generates high spatiotemporal resolution images by leveraging the complementary spatiotemporal information of multisource optical images, offering cost-effective and flexible solutions, garnering widespread attention. Among these methods, the reliable and adaptive spatiotemporal data fusion method (RASDF) is recognized for its robustness in handling discrepancies between heterogeneous images and its ability to reconstruct surface information under drastic change scenarios. However, RASDF is hindered by inefficiencies and blurred boundary information in fusion results. To address these limitations, this study introduces an enhanced RASDF method based on an object-oriented strategy, termed object level RASDF (OL-RASDF).

Methods By incorporating multi-scale imagesegmentation techniques and assuming consistent spatiotemporal variation patterns within segmented objects, OL-RASDF replaces the original pixel-level fusion steps, which rely on a moving window strategy, with object-level fusion steps that use image segments as the basic processing units. This approach streamlines the fusion process and enhances the preservation of boundary information.

Results Experimental results from two study areas demonstrate that OL-RASDF achieves the highest fusion accuracy, the sharpest boundary details, and a significant improvement in efficiency. The average processing time in the two study areas was reduced by 62.1%.

Conclusions The results indicate that OL-RASDF is highly applicable to long-term and large-scale image fusion tasks, offering a superior balance of accuracy and efficiency.