Target Recognition Network Based on Improved Convolutional Regression Network
Shuang Zhang, Jinyu Song, Weiwei Li · Journal of Physics Conference Series · 2023
Abstract For refine the regression precision and speed of convolutional neural recognition network, propose improved detection and identification algorithm on the basis of deep learning algorithms, mainly aiming at improving the accuracy and speed of regression prediction of target recognition. Firstly, the acquired images are preprocessed in batches to obtain image datasets with different sizes of detection objects, that is, the enhancement of small sample images. Secondly, the one-stage target detection network SSD algorithm is improved. ResNet50 network is used to extract features, while FPN is fused with the main algorithm to ameliorate the detection network, which purpose is to improve the target detection accuracy. The test results on remote sensing datasets show that the actual accuracy was elevated to about 0.9% compared to the conventional algorithm. The speed in the actual measurement is also increased by 0.6 frame/s.