本项目以复杂场景下的运动目标跟踪为研究对象，旨在研究构建高精度的实时运动目标跟踪算法，用于为智能视频安防系统提供高效的技术支持。如何提高跟踪方法对运动目标的判别和定位精度，是现有研究面临的核心问题。本项目的主要内容是研究双重Siamese网络框架下基于多层深度特征融合的运动目标跟踪方法，为提升现有跟踪算法精度，提高算法效率提供新的思路。为此将研究三个主要问题：1、研究基于通道选择机制和亲密度聚类的深度特征优选模型，有效抑制复杂场景中的背景、噪声干扰；2、研究基于残差注意机制的目标判别模型，利用视觉机制提升深度语义特征的目标判别能力，为目标精确定位创造条件；3、研究双重Siamese网络框架下基于多层深度特征融合的目标定位模型，实现目标的精确定位。解决了这三个问题，将保证算法能够在高精度跟踪的前提下兼顾算法实时性。研究成果将提高智能视频安防系统的实时目标跟踪精度，辅助智能视频安防系统建设。

This project takes the moving target in complex scenes as the research object, and aim to construct a high-precision real-time moving target tracking algorithm, which is used to provide efficient technical support for intelligent video security systems. How to improve the discrimination and positioning accuracy of the tracking method is the core problem of existing research. The main content of this project is to study the moving target tracking method based on deep feature and residual attention mechanism under the dual Siamese network framework. In order to improve the tracking accuracy of existing tracking algorithms and improve the efficiency of algorithms. Three main issues should be studied: 1. To effectively suppress noise interference and improve feature representation in complex scenes, we intend to develop a deep feature optimization mechanism based on channel selection mechanism and intimacy propagation. 2. To create conditions for precise targeting, the target discriminant model based on residual attention mechanism will be studied. In our programs, the visual mechanism is used to improve the target discriminating ability of deep semantic features. 3. To achieve accurate target location, we plan to research the target location model based on multi-deep feature discrimination under the dual Siamese network framework. Solving these three problems, the algorithm calculation speed will be balanced and ensure the real-time performance of the algorithm at the same time satisfying the tracking accuracy. The related achievements of the project will improve the real-time target tracking ability of the intelligent video security system, and assist intelligent video security system construction.