Using the Histogram of Oriented Gradients to track different objects we could distinguish the target and the background with HOG visualization and the technique counts occurrences of gradient orientation in localized portions of an image
ball database : https://1drv.ms/u/s!Ap2v7aRcA-T9cUWZQ0yk4wFwR6E?e=xjsvcf (201 frames) car database : https://1drv.ms/u/s!Ap2v7aRcA-T9cvgY8bLyN1_Rb-I?e=PZXDbP (254 frames)
Part of ball tracking result:
Part of car tracking result:
