Detection and matching of object using proposed signature - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 499

In this figure, the input object in the left is mismatched with its corresponding object in
the original image. Additionally, this mismatch repeats many times with the test objects us-
ing SURF. From the previous results, although SURF method and proposed algorithm have
presented to detect and matching objects in an image; however, the presented algorithm is
more effective and accurate in objects matching process than SURF and simply in use by some
humble statistical equations without any constraints as in the other methods. Next section
shows the conclusion of this work.
In Figure 5 , another example for mismatching of objects using SURF method whereas exact
matching has illustrated in Figure 6 .
FIGURE 5 The SURF example with the mismatch an object. (a) The original image; (b) the
detected object; (c) the exact matching with the exact object in the image.
FIGURE 6 The SURF example with the exact match an object. (a) The original image; (b) the
detected object; (c) the exact matching with the exact object in the image.
6 Conclusions
This chapter has presented proposed algorithm for object detection and matching based on its
own signature using morphological segmentation tools. The algorithm has divided into three
parts; first is segmentation process, second is the construction of object signatures, and the last
part is the matching based on them to classify and define the object. Actually, this signature
has a singularity, simply in use, saving it on a small memory, and working in a variety of light
levels. Moreover, the proposed algorithm matched the objects neither among object's blobs
nor regions of interest; but among the constructed signatures. On the other hand, SURF meth-
od has been presented for comparison with the proposed method on the experimental results.
Many difficulties are appeared in matching process, such as object in unusual intensity lumin-
ance, shape, orientation, position, different sizes, or unclear image of objects; but the proposed
algorithm has overcome on them, while SURF has not done. The performance has been tested
120 from a wide variety of different objects; it has been achieved 100% in the case of construct-