Abstract
The use of attributes in person re-identification and video surveillance applications has grabbed attentions of many researchers in recent times. Attributes are suitable tools for mid-level representation of a part or a region in an image as it is more similar to human perception as compared to the quantitative nature of the normal visual features description of those parts. Hence, in this paper, the preliminary experimental results to evaluate the robustness of attribute detectors against pose and light variations in contrast to the use of local appearance features is discussed. Results attained proven that the attribute-based detectors are capable to overcome the negative impact of pose and light variation towards person re-identification activities. In addition, the degree of importance of different attributes in re-identification is evaluated and compared with other previous works in this field.
Highlights
Several challenges related to person re-identification (PRI) includes lighting and view point variations that could cause severe effects on the re-identification output quality
Low level features performance is acceptable on images with some variety of illumination and view point, but in the severe cases, low level features are unable to perform well
The main problem of pose and illumination differences can be solved in attributebased methods by using the samples in training set with different illumination and pose conditions, which can be more robust than the descriptors with low level features
Summary
Several challenges related to person re-identification (PRI) includes lighting and view point variations that could cause severe effects on the re-identification output quality. In appearance-based re-identification, the main purpose is representing the subject’s image based on the visual features which are robust enough against these variations. Low level features performance is acceptable on images with some variety of illumination and view point, but in the severe cases, low level features are unable to perform well In such cases, segmented grids are more preferable since this approach could contribute too much better results [4] along with local parts of the silhouettes [2] rather than on individual pixels. The use of mid-level representation of the scenes brings us one step closer to the way that human really understand what is happening around It eliminates the difficulties caused by illumination and pose variations. Based on these different options to define attributes for re-identification, it would be of high interest to do PRI purely based on attributes
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: International Journal of Engineering & Technology
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
