Research And Application Of Image Saliency Object Detection Based On Deep Learning

In computer vision,saliency object detection has become one of the key research areas of researchers.The purpose of image saliency detection is to simulate the human visual attention mechanism through a computer in a complex scene,eliminate redundant information in the image,and improve the computer's ability to understand complex scenes autonomously and intelligently.This kind of computer ability can be applied to real life scenarios and can help other computer vision tasks.With the continuous development of deep learning technology,the introduction of deep learning in image saliency object detection helps to improve the robustness of the model.Therefore,this article mainly studies image saliency object detection and application based on deep learning.The research work is as follows:For image saliency object detection,there are some problems such as blurred foreground and incomplete edge details.To solve this problem,an image saliency object detection model combing local and global information is designed.Firstly,we use the dilated convolution and side fusion structure to extract the rough feature information of the image.Then,we design a hierarchical aware attention mechanism module.By assigning different weights to each channel in the deep and shallow layers to extract useful channel feature information to obtain the correlation between different channel feature information.What's more,the cross-entropy loss function is used to further optimize the detailed information of the feature map to obtain the saliency map,which improves the accuracy of the model.Experiments show that the F-measure of this model on the ECSSD dataset reaches 0.907,and the MAE reaches0.049.Concerning the problems that exist in computer vision,such as saliency areas highlight unevenness and unclear edges,which lead to poor saliency accuracy.To solve this problem,a saliency object detection model based on channel-spatial joint attention mechanism is proposed.Firstly,we improve the channel attention mechanism and add the pixel probability values pixel by pixel in the feature map,so better obtain the correlation of information between the channels.Then,we integrate the spatial attention mechanism in parallel with on the channel attention mechanism,and the saliency areas with prominent object are received by weighting the spatial information of the feature map.In addition,to obtain a more fine-grained saliency map,the two feature maps output by the channel and spatial attention mechanism are weighted fusion,which fed back to the channel-spatial joint attention mechanism.Sufficient experiments on public datasets DUTS-TE and SOD have demonstrated that the proposed method outperforms the others from the value of F-measure and mean absolute error.Since saliency object detection has broad application prospects,meanwhile application is the goal of saliency object detection.We apply the saliency object detection model based on the channel-spatial joint attention mechanism to the image background blur technology,and conduct experimental research and analysis.Experimental research shows that the model combined with Gaussian blur algorithm is applied to the background blur technology to present a better background blur effect.