Title: Feature vector sharing and scale comprehensive optimisation for targets detection in smart neighbourhood governance and monitoring
Authors: Liu Jianmin
Addresses: Guangxi Key Laboratory of Cross-border E-commerce Intelligent Information Processing, Guangxi University of Finance and Economics, Nanning, 530007, Guangxi, China
Abstract: This article proposes feature vector sharing and scale comprehensive optimisation strategy of image target detection and recognition method of complex street maximum suppression based on the calculation of the corresponding feature area corresponding to the feature map and completely complete eigenvector. Based on this, this article also combines a fine-tuning method based on transfer learning generalisation, which is suitable for non-convex optimisation and high-dimensional space. First, the method described above implements the optimal rectangular selection box competition based on the scale comprehensive optimisation strategy, and selects the selection box that can reflect the core essence of the target in each classification set. Then, this article realises the model of detecting image target in complex neighbourhood, which improves the accuracy and robustness. Furthermore, we experimentally demonstrate that the accuracy and robustness of our proposed method are superior to those of conventional methods.
Keywords: feature vector sharing; scale comprehensive optimisation; neighbourhood images; smart neighbourhood monitoring.
International Journal of Computer Applications in Technology, 2020 Vol.64 No.3, pp.260 - 268
Received: 05 Jun 2020
Accepted: 17 Jun 2020
Published online: 16 Dec 2020 *