Abstract: Spillways are important hydraulic structures. Intelligent detection of spillways in remote sensing images are of great scientific significance and application values. However, due to the small scale and intricate surroundings of spillways, it is difficult to automatically detect spillways in remote sensing images. In this study, the domestically developed high-resolution imagery from the GF-7 satellite is used, which has stereoscopic imaging capability. The Spillway Geometry Stereoscopic Index (SGSI) is constructed and an automatic spillway recognition method is proposed. Firstly, data preprocessing is conducted to obtain fused imagery and high-precision DSM. Next, the Mean-Shift algorithm is employed for segmenting the fused imagery, extracting multi-dimensional features of segmented objects, and determining target objects through multi-feature joint decision-making. Finally, integrated and post-processing of spillway patches are performed based on the spatial context relationship of target objects to output the final recognition results, achieves automatic recognition of spillways. The proposed method is validated on three dam reservoirs containing spillways, and experimental results show good matching between the extracted spillway boundaries and reference boundaries, with an overall accuracy of 89.23%. The proposed method is proved to be efficient and accurate for automatic spillway recognition.

Key words: GF-7 satellite, spillway geometric solid index, automatic recognition, object oriented, multi-feature joint decision-making