Description: The TrackDialog tool assumes that there is a temporal sequence of images with objects moving in it. The temporal sequence can reside either in a memory (RAM), for example, by loading images to one ImageObject structure, or in a file system (on a hard drive). The input type is set in the choice list of the TrackDialog tool.
The tracking is executed in two steps. During the first step, called training, a model of the tracked object is created. The object is selected by dragging a mouse in the image frame around an object and clicking the "Train" button in the TrackDialog tool (see Figure 1). If the area around the object is inaccurate then it can be adjusted in the SubArea dialog that appears before any training is performed. The object model is always extracted from the first band of the image shown in the frame (an image at one time instance), for example, in Figure 2. It is possible to save a model to a file and load a model from a file using the buttons "SaveModel" and LoadModel".
Figure 1: The dialog for setting up and executing the TrackDialog tool.
Figure 2: An illustration of a mouse-selected training area overlaid onto the first frame of a sequence of images.
During the second step, called tracking, three parameters should be set before launching the tracking algorithm. The three parameters are a model type, a size of the search neighborhood and an acceptance level for the model match. The model type can be either dynamic with model updates after processing each image frame (band) or static with no model updates. The size of the search neighborhood refers to the area where we track the model. The size can be set by entering positive numbers into the "RowSearch" and "ColSearch" edit boxes and should be large enough to cover the model motion in pixels between two consecutive images. The last parameter, an acceptance level, is a value that determines when the match is declared as valid. The acceptance level can take values between 0 and 100 with the value equal to 100 for a perfect match. Once the parameters are set, the tracking algorithm is launched by clicking the "Find" button.
Run: In order to process a sequence of images, train the tool on any image sub-region and run the tool according to the "Description" section. The tracking results can be viewed in a text format and in a graphical format by clicking the "ShowRes" button. The text format contains the information about the row, column and score of the best match in each frame. Statistics of row and column coordinates are provided at the bottom of the text output area. The graphical output displays rows and columns as a function of an image index in a new frame.
Release notes:
The current release does not save out any text of the track plots in the "Save As Image" mode.