Evaluation on Computer Region Location Algorithm Based on Fully Convolutional Network

The research of computer region location algorithm can obtain a more accurate location. In the field of computer science, target location has always been a very important issue. The traditional target location method usually uses the image as the basis to extract its features, and then carries out the research of regional recognition and location. Through this method, the image can be subdivided into many grouped images with features, and then use the prior knowledge of these images to match. However, there are several problems in doing so. First of all, they would generate a large amount of data in the training process, which is not a good representation of the target. Secondly, these feature extraction algorithms do not consider the structure of the target region itself in the training process, so they cannot effectively extract the target information. Finally, they often use many image features to train the double network structure. In view of the above problems, this article proposed a more excellent and convenient method to extract and recombine a large amount of feature information contained in the image. The algorithm used Fully Convolutional Network (FCN) to extract the feature map of the image. In the training phase, people train the network by representing the original image as a multi-layer FCN. Then the network was connected to the computer screen, and the architecture was optimized by adjusting the Convolutional Neural Network (CNN) to calculate the better results, and feedback to the model of the full CNN. During the experiment, a large number of experiments were carried out on two public data sets (Cityscapes and OpenImage Segmentation, respectively). The results showed that the method proposed in this article has achieved good results: the accuracy of the computer region location algorithm based on FCN studied in this article can reach more than 98%.