Many people may wonder, why does the image recognition Why consider used tracker. First explain here, just under real-time conditions of the object recognition process discussed here, if only simple identification, please move.
Straight to the point, the real-time image of the identification process, we often encounter such a problem that the movement we object to be identified is continuous, and our recognition area is not continuous. In simple terms, we assume that a frame to identify the object in the next frame, if we re-identification of the whole picture, which will result in an inevitable waste of resources. That every time we have to recognize the next recognition is independent. In order to make better use of resources, there is an idea that is directly on the original based on a recognition of the target track, suspected to get the ROI, then the next recognition As for the current ROI can be operated without the need for original operation. The idea wonders in the process of identifying and tracking single goals.
For example, FIG:
(Upper picture shows the entire FIG identified)
(I.e., using the FIG identified ROI)
Recognition efficiency can be imagined.