In the step of object tracking, we estimated the object poses and object information using CamShift and optical flows. Each of the object pose is evaluated and confirmed by object information, comprising several elements such as feature points, descriptors, and object within the sub-window, the expected object pose, and its histogram. The tracked objects are estimated from the credibility for learning of object information, and is described below.
In this study, to evaluate the credibility of the object, information of the tracked object is compared to the initial information, which is detected by the previous detection module. This information comprises two parts: the feature points and its histogram, indicating a standard information, named …show more content…
Morphological shape of the object can be extracted by the intersection of the positive and negative distribution of the sub-windows, as shown in Figure 8.
Fig. 8. Morphologic characteristics map of the detected object. Green area shows the largest frequency of the object histogram, and red area is extraneous to the object color
Next step of the tracking is to provide a label to the feature points, which belongs within the positive sub-window. Thus, information of the object histogram and feature with the sub-windows is formulated by the following equations.
■(Sub-window(〖Label_0 (k〗_1⊃X_n (x_1,x_2⋯x_n)))@⋮@Sub-window(〖Label_j (k〗_j⊃X_j (x_1,x_2⋯ x_j)))) …show more content…
The proposed algorithm recognizes objects with invariant features and reduces dimensions of the feature descriptor for using on mobile devices. The experiments show that the proposed method is more robust than the traditional methods, especially in the changes of appearance and viewpoint, and this can accurately track objects in various environments. Main contribution of this study was to propose an efficiently suitable method for cultural heritage by applying cultural object detection and tracking algorithms to cultural
In this study, to evaluate the credibility of the object, information of the tracked object is compared to the initial information, which is detected by the previous detection module. This information comprises two parts: the feature points and its histogram, indicating a standard information, named …show more content…
Morphological shape of the object can be extracted by the intersection of the positive and negative distribution of the sub-windows, as shown in Figure 8.
Fig. 8. Morphologic characteristics map of the detected object. Green area shows the largest frequency of the object histogram, and red area is extraneous to the object color
Next step of the tracking is to provide a label to the feature points, which belongs within the positive sub-window. Thus, information of the object histogram and feature with the sub-windows is formulated by the following equations.
■(Sub-window(〖Label_0 (k〗_1⊃X_n (x_1,x_2⋯x_n)))@⋮@Sub-window(〖Label_j (k〗_j⊃X_j (x_1,x_2⋯ x_j)))) …show more content…
The proposed algorithm recognizes objects with invariant features and reduces dimensions of the feature descriptor for using on mobile devices. The experiments show that the proposed method is more robust than the traditional methods, especially in the changes of appearance and viewpoint, and this can accurately track objects in various environments. Main contribution of this study was to propose an efficiently suitable method for cultural heritage by applying cultural object detection and tracking algorithms to cultural