Intend to adopt the method of optical flow for moving target tracking, using Lucas kanade algorithm, using the algorithm to calculate the two pictures (size is M pixels X N pixels) between each pixel of the Dx, Dy (size is M * N 2 d matrix, personal understanding is for each pixel in the image in the X direction and Y direction movement speed), ask, Dx, Dy, respectively what does that mean?

Or calculate the Dx and Dy how to judge the moving target in the image of the X and Y?

I thought all the pixels of knocking=Dy Dx + Dy Dx * *, and then take the knocking from big to small order, and then corresponding pixels is considered to be against a larger moving point target, but I put all the knocking value amplification (such as magnified 40 times, become DXY40), and then to all DXY40 as grey value, back to a picture with the original image is the size of the image (the size is M pixels x N pixels), found that the image is very strange,

My understanding is that if the two parts of the original image difference is larger, the corresponding DXY40 is larger,