Research And Implementation Of The Segmentation System Of Cultural Elements Of Chinese Painting Based On The Integration Of Spatial Structure Characteristics

Traditional Chinese painting is one of the most important carriers of Chinese culture,which contains plentiful cultural elements.At present,in the field of image segmentation research,there is a lack of data sets related to traditional Chinese painting.The algorithm model trained for open data sets is not applicable to traditional Chinese painting images.Therefore,this thesis analyzes the spatial structure,color and line characteristics of traditional Chinese painting images,proposes a segmentation algorithm based on spatial structure feature fusion,and designs and develops a traditional Chinese painting element segmentation system.The main work of this thesis includes:(1)Build a data set of traditional Chinese meticulous painting.The data set is constructed through the steps of collecting,classifying and labeling the images of traditional Chinese painting meticulous painting,which is used for the training and testing of the case segmentation model in this thesis,and provides data support for other work in the laboratory.(2)Build a network architecture that integrates spatial structure features.The spatial distribution of images is an important feature.In artificial painting such as meticulous painting,the color distribution and overall characteristics formed by painting around the main instance are more prominent.By dividing the image into position classes for feature extraction and fusion,the example mask prediction of the corresponding channel is output with each position module as the center.The mask predicted in the close channel will reflect the similarity of the features in the close position,thus completing the segmentation task using the spatial structure features.(3)A segmentation model based on entropy regression of digital culture is proposed.Meticulous painting is an unnatural image created artificially with aesthetic and technical requirements.It has the characteristics of halo and gradual change in color,and discontinuous and discontinuous in contour lines.In view of these two main characteristics,combining with spatial structure extraction features,color entropy is used to locate the location of the instance area,and contour entropy regression is used to predict the mask to obtain the best segmentation effect on the meticulous painting image.(4)Complete the design and implementation of meticulous painting segmentation system.The core function of the system is to segment fine brushwork with high quality based on the algorithm model proposed in this thesis,including data set display,material warehousing display,other segmentation algorithm integration and other functions.The algorithm proposed in this thesis is tested on the constructed Chinese painting dataset,and the average accuracy rate is 55%.The segmentation accuracy of each semantic category is improved compared with the mainstream model.