Hog and sift
- 1. OBJECT DETECTION HOG AND SIFT ALGORITHM
- 2. HOG ALGORITHM ď HOG stands for histogram of oriented gradients. ď The hog descriptor focuses on structure or shape of the object. ď It uses magnitude as well as direction of the gradient to compute the features. ď It generates histogram by using magnitude and direction of the gradient.
- 4. ď Here we calculating gradient magnitude and direction, to calculate pixels intensity we need ď X direction=|40-70|=30 ď Y direction=|20-70|=50 ď By these values we are calculating magnitude and direction of the gradient ď By using magnitude and direction we calculate feature vectors 20 40 70 70
- 7. e
- 8. ď Before getting the hog feature and after concatenating feature vectors we are supposed to do normalize. ď Suppose we have taken 150*300 pixels and multiply with 2 to increase the brightness and divided by 2 to decrease the brightness, then you cant compare two images without normalization becâz the pixels intensity will be changed. ď But if you normalize the feature vectors it is easy to compare
- 10. ⢠For hog features giving human template and giving output for convolving with human model Then it will predict whether it is human or not.
- 11. SIFT ALGORITHM ď SIFT stands for scalar invariant feature transform and was first presented in 2004, by D.Lowe, University of British Columbia. ď It is a way to describe a local area in an image. ď In this whole image is reduced to set of points. ď SIFT is invariance to image scale and rotation. ď This algorithm is patented, so this algorithm is included in the Non-free module in OpenCV.
- 12. ď Major advantages of SIFT are ď Distinctiveness: individual features can be matched to a large database of objects ď Quantity: many features can be generated for even small objects ď Efficiency: close to real-time performance ď Extensibility: can easily be extended to a wide range of different feature types, with each adding robustness.
- 13. ď The scale space of an image is a function L(x,y,Ď) that is produced from the convolution of a Gaussian kernel(Blurring) at different scales with the input image. ď Within an octave, images are progressively blurred using the Gaussian Blur operator. ď Mathematically, âblurringâ is referred to as the convolution of the Gaussian operator and the image. ď Gaussian blur has a particular expression or âoperatorâ that is applied to each pixel. What results is the blurred image. gaussian blur Gaussian blur operator
- 14. ď In difference of gaussian kernel(DOG) we use those blurred images to generate another set of images ď These dog images are used to find interesting points in the image. ď These process is done for all images in the gausian pyramid. ď After that we stack those different images on top of the each other and basically your looking extreme points. ď In these locally distinct which stand out and those are your key points.