How to use opencv on netbeans 1.
detection of a banknote in multiple banknote cluster
tools - c++, opencv
method a. preprocess -> noise reduction using bilateral filter - preserves the edges of the image blobs, which are very valuable structures in the detection of feature points. -> CLAHE (Contrast Limited Adaptive Histogram Equalization) - to increase contrast improves low light recognition
b. reference image in database -> To improve the detection of the relevant parts of the banknotes and to avoid the usage of sections that are similar across several banknotes, the keypoint detection algorithm is only applied inside the masks associated with each banknote. Only relevant parts such as the banknote number and unique textures or patterns are included in the banknotes masks. ref to fig.1
C. recognition(it can recognize any euro banknote)
-> To improve the robustness of the system, 3 levels of detail for each banknote are provided (256, 512 and 1024 pixels wide images).
-> matching technique, the image resolution of both the reference and the target images should be the same.
-> feature detection
-SIFT [2], SURF [24], GFTT [21], FAST [22], ORB [31], BRISK [32], STAR [33] and MSER [23].
-> keypoint extraction
-SIFT [2], SURF [24], FREAK [34], BRIEF [35], ORB [31] and BRISK [32] feature descriptors.
-> The initial matching can be performed using either a brute force or a heuristic approach.
-> In the brute force approach, each descriptor in the image is compared with all descriptors in the reference image to find the best correspondence. In the heuristic approach, such as the FLANN library [36], several optimizations are employed to speed up the computations. These optimizations can be related to the appropriate selection of which keypoints to match, and to the use of efficient data structures to speed up the search (such as k-d trees).
-> ransac
-> ratio test
->(last step is shape analysis) this is a post process for detectimg the top most banknote in multi
ple banknote.