基于时空双流全卷积网络的视频目标分割算法研究及改进

Research and Improvement on Video Target Segmentation Algorithm Based on Spatio-temporal Dual-stream Full Convolutional Network

运动目标检测算法在视频监控等领域应用广泛,现实场景背景复杂多变,传统的运动目标检测算法往往效果不佳,目标轮廓常常难以识别。采用了一种结合空间时间双结构网络用于视频目标分割,空间网络的输入是RGB图像,相当于外观模型,时间网络的输入是光流图像,相当于运动模型,两者有互补的作用。该方法既能对目标的外观进行建模,也添加了运动信息,同时采用多帧光流信息的分割结果进行融合的方法改进空间流网络输出,使分割结果更加精准。

This paper uses a spatial time dual structure network for video target segmentation.The input of the spatial network is an RGB image,the input of the time network is the optical flow image,they can complement each other.In this way,this method can not only model the appearance of the target,but also add motion information.At the same time,the method of fusion using the multi-frame optical flow information segmentation results can improve the output of the spatial stream network and makes the segmentation result more accurate.