Segmentation of Cast Shadow in Surveillance Video Sequence: A Mean-Shift Filtering Based Approach
The accuracy of the segmented motion objects tracking in surveillance videos will decline when shadows are detected as moving objects. To address this, a new spatial based method for the segmentation of cast shadow regions from the motion segmented video
- PDF / 451,062 Bytes
- 13 Pages / 439.37 x 666.142 pts Page_size
- 98 Downloads / 162 Views
Abstract The accuracy of the segmented motion objects tracking in surveillance videos will decline when shadows are detected as moving objects. To address this, a new spatial based method for the segmentation of cast shadow regions from the motion segmented video sequence is proposed. The motion segmented frame is processed using Mean-Shift filter for smoothening and then the cast shadow pixels are segmented by interval value based representation of RGB color channels. The proposed model overcomes the restrictions on direction of light source and surface orientation which are generally considered for cast shadow segmentation. Experiments have been conducted on challenging indoor and outdoor video sequences of IEEE Change Detection (CD) 2014 and ATON datasets. Further, comparative evaluation with contemporary methods using standard evaluation metrics has been carried out to corroborate the efficacy of proposed method.
Keywords Cast shadow segmentation Video surveillance Mean-shift Interval value
M. Chandrajit (&) R. Girisha T. Vasudev Maharaja Research Foundation, Maharaja Institute of Technology Mysore, Mandya, Karnataka, India e-mail: [email protected] R. Girisha e-mail: [email protected] T. Vasudev e-mail: [email protected] M. Chandrajit R. Girisha T. Vasudev PET Research Foundation, PES College of Engineering, Mandya, Karnataka, India © Springer Nature Singapore Pte Ltd. 2018 D.S. Guru et al. (eds.), Proceedings of International Conference on Cognition and Recognition, Lecture Notes in Networks and Systems 14, DOI 10.1007/978-981-10-5146-3_28
291
292
M. Chandrajit et al.
1 Introduction Smart video surveillance is currently one of the most active fields of research. The smart video surveillance detects motion objects, tracks their actions and subsequently analyzes its behavior to prevent any untoward incidents. While segmenting the motion objects from the video sequence, cast shadows which are part of motion object are also segmented. The presence of shadow causes object merging and object shape distortion, and thereby affecting the object tracking and classification tasks. Figure 1 shows an example of object tracking in the presence of shadow. The shadow which is segmented during the motion segmentation is also tracked along with the foreground object. Therefore, the elimination of moving shadow in frames of video sequence is an important task. The shadow segmentation in the frame of a video sequence is a challenging task because the shadow pixel will be having similar temporal characteristics as the motion object pixel. Further, the shadow region will be often camouflaging the foreground object blob region which makes the shadow segmentation a complex task [1–4]. In this paper, a new method modeled using Mean-Shift filter and interval value based representation of RGB color channels for segmenting the cast shadow pixel region from the video sequence captured in a complex environment is proposed. The rest of the paper is organized as follows: Sect. 2 presents the related works, Sec
Data Loading...
