In the power information platform,the extraction of required multi-media elements faces problems such as poor image filtering and denoising effect,rough image quality,and significant influence of skewness indicators on the extraction effect.Therefore,a multi-media element automatic extraction model based on visual semantic segmentation algorithm is proposed for the power information platform.This model adopts an image filtering denoising enhancement algorithm,which detects multi-media image noise points through pulse detection of extreme values,and combines adaptive median filtering to complete multi-media image filtering denoising processing in the power information platform.This paper constructs a visual semantic segmen-tation network consisting of a fully convolutional segmentation network and a regional suggestion network.The processed multi-media images are used as inputs to the visual semantic segmentation network.The preprocessed multi-media images are segmented through the fully convolutional segmentation network,and multi-media elements are extracted.Combined with the regional suggestion box containing category marker information obtained by the regional suggestion network,the multi-media element extraction effect of the fully convolutional segmentation network is optimized.The experimental results show that the pre-processed multi-media images of this model have higher clarity and can effectively extract the required multi-media ele-ments.The extraction time of multi-media elements under different degrees of skewness is between 2.2 s and 2.4 s,which in-dicates higher extraction efficiency.
维普
万方数据