Search results

(1 - 2 of 2)
AN IMPROVED BACKGROUND SUBTRACTION ALGORITHM FOR INTELLIGENT VIDEO SURVEILLANCE
Title
AN IMPROVED BACKGROUND SUBTRACTION ALGORITHM FOR INTELLIGENT VIDEO SURVEILLANCE
Creator
Xia, Zhengyu
Date
2013, 2013-07
Description
This thesis briefly introduces background subtraction which is one of the current advanced techniques for intelligent video surveillance. The...
Show moreThis thesis briefly introduces background subtraction which is one of the current advanced techniques for intelligent video surveillance. The paper also presents an improved background subtraction technique for motion detection in the area of intelligent video surveillance. The proposed technique is difference from the state-of-the-art background subtraction algorithms such as Gaussian Mixture Models or Behavior Subtraction, etc. The idea of this enhanced background subtraction is inspired by the ViBe algorithm. In the first step, which is the initialization of background subtraction, it takes samples only in its current location and 3x3 neighborhoods for each pixel. Then compare each pixel with the sampling set to determine whether this pixel belongs to background models or foreground objects. The next step is to update the sampling sets so that the algorithm can adapt to different scenarios. The final procedure is the post-processing method, which has better performance to overcome difficulties. This approach differs from those based upon the classical belief that the samples set of current pixels should be collected according to certain mathematic models. Experimental results of the proposed algorithm are better than earlier techniques that use the ViBe algorithm. Keywords: Background subtraction, computer vision, motion detection, real-time systems, intelligent video surveillance.
M.S. in Electrical Engineering, July 2013
Show less
IMPROVING DEEP LEARNING BASED SEMANTIC SEGMENTATION USING CONTEXT INFORMATION
Title
IMPROVING DEEP LEARNING BASED SEMANTIC SEGMENTATION USING CONTEXT INFORMATION
Creator
Xia, Zhengyu
Date
2021
Description
Semantic segmentation is an important but challenging task in computer vision because it aims to assign each pixel a category label accurately...
Show moreSemantic segmentation is an important but challenging task in computer vision because it aims to assign each pixel a category label accurately. Nowadays, applications such as autonomous driving, path navigation, image search engine, or augmented reality require accurate semantic analysis and efficient segmentation mechanisms. In this thesis, we propose multiple models to improve the performance of semantic segmentation. In the first part, we focus on the single-task network, which aims to improve the performance of semantic segmentation. Our research includes exploiting context information using mixed spatial pyramid pooling to extract dense context-embedded features in FCN-based semantic segmentation. We also propose a GAF module to generate a global context-based attention map to guide the shallow-layer feature maps for better pixel localization. In the second part, we focus on a multi-task network that incorporates semantic segmentation to improve other computer vision tasks such as object detection. Specifically, a multi-task network, along with a learning strategy is designed to let semantic segmentation and object detection assist each other since they are highly correlated. Also, we include weakly-supervised multi-label semantic segmentation learning to deal with the shortage of high-quality training examples and to improve the performance of cross-domain object detection. In the third part, we focus on improving the performance of video panoptic segmentation, which is a unified network that incorporates semantic segmentation and instance segmentation using video streams. We design a new ConvLSTM pyramid to transmit spatio-temporal contextual information in our video panoptic segmentation network. Specifically, we propose a modified ConvLSTM to generate temporal contextual information. Also, we design an MSTPP module to obtain mixed spatio-temporal context-embedded feature maps. Experimental results on different datasets show that our proposed method achieves better performance compared with the state-of-the-art methods.
Show less