I am working with person recognition in outdoor surveillance. I ve one doubt that can i use local binary pattern feature for person recognition in outdoor environments .
As far as i know you can use 2D LBP for this particular problem. Because LBP is robust enough to illumination variations and object position and scaling. Yet, you may encounter two significant challenges in your problem. First is related to the extraction of person from the images taken outdoor environment. I mean diffuculties in applying background subtraction. Once you extract the person or can draw bounding box around the person and you have an exact position of the person you should simply crop the ROI and proceed the feature extraction via 2D LBP. Another challenge may be the computational cost of LBP as LBP makes convolution process inherently. This problem may be eased by applying LBP operation on n'th frame image in the video sequence.