Abstract: Dongba pictograph is a kind of pictograph used in early China. The research on the recognition and digital protection of Dongba pictograph is of far-reaching significance to the inheritance of Chinese culture. In view of the complex structure of the pictographs extracted from ancient Dongba books, the presence of heterogeneous characters, and the interference of texture features of the special Dongba paper recording the characters, an improved image recognition algorithm based on CBAM-YOLOV4 is proposed. The attention mechanism module CBAM and feature fusion module were added to the algorithm. The channel and spatial attention sub-modules in CBAM were used to deduce the attention diagram of the images in turn, and the feature fusion module was used to further extract the features of the Dongba pictograph images, so as to realize the optimization of YOLOv4 image detection and recognition algorithm. The improved CBAM-YOLOV4 algorithm was applied to Dongba pictograph recognition. Compared with the YOLOv4 algorithm, the mAP value of the improved algorithm is improved by 4.42 percentage points, indicating that the algorithm has better Dongba character recognition performance.