부산대학교 도서관

Although image detection technology is widely used in various fields, there is still a large gap in the similarity detection of science fiction painting. Therefore, a similarity detection method for science fiction painting is proposed. Firstly, a k-layer pyramid image is created for the source image to be detected and the template image. Then, multi-strategy improved sparrow algorithm (MISSA) is used to perform rough matching in the top subgraph of the source image to obtain the coordinates of the initial matching target, the location transferred from the layer above each layer is the starting point of search, and pixel by pixel matching is carried out within the set window range. Finally, pHash is used as the similarity measure to calculate the similarity of matching results. The hybrid search strategy based on step function, multi-stage dynamic control of safety threshold, and food search strategy based on Logistic model are used to improve sparrow search algorithm, thereby forming MISSA to improve the accuracy and real-time performance in the matching process. In terms of performance verification of MISSA, the rationality and effectiveness of the three improved strategies are verified by ablation experiment, and the experimental results on CEC2017 benchmark function show that the optimization performance and convergence performance of MISSA are better than that of peer algorithms. The comparison results in the similarity detection experiment of science fiction painting fully verify that the proposed detection method has strong robustness in meeting the requirements of real-time and accuracy of matching.