Abstract:

It is found that generic objects with well-defined closed boundary can be discriminated by looking at the norm of gradients, with a suitable resizing of their corresponding image windows into a small fixes size,which can save a lot of time. Inspired by the high quality of ObjectNess binarized normed gradients (BING), it is used for the multi-ship target detection on the sea. Considering the characteristics of the ship targets and the artificial objects, a method of predicting the object candidate windows based on corner points and ObjectNess BING is proposed, which can also generates a small set of high quality ship target windows, yielding 96.2% object detection rate (DR) just like the former ObjectNess BING dose for the test of images downloaded from the internet, but with only 900+proposals. It reduces the time cost of ship targets detection and makes the ship detection more efficient than the former ObjectNess BING.