Abstract |
: |
Human and face detection plays an important role in the surveillance system, Intelligent user interface. Conventional human and face detection methods usually take the pixel color directly as information cues which are sensitive to noise and changes in illumination. This is difficult to classify the images. To classify the images by hybrid features that combines Local binary pattern (LBP), Local gradient pattern (LGP), Binary histogram of oriented gradients (BHOG). Local binary pattern makes the changes in global illumination. Local gradient pattern makes the intensity variation of edge components robust. Binary histogram of oriented gradient computes the gradient magnitude and orientation in all the pixels in block and it changes the local pose. To select the features that needs to combine local features in each stage by using Adaboost method. Introducing nearest neighbor classification this work focuses on images that is classified correctly with feature selection and minimizes the false negative per image. |