Tracking¶
Two tracking technologies are currently supported in EV Toolbox:
- Marker-based tracking technology recognizes only square images in a distinct frame (see. more about Marker Recognition Technology), however it is more reliable and less demanding on computing resources.
- Markerless tracking technology tracks images with arbitrary aspect ratio without a frame (unlike marker-based tracking). This tracking technology has certain requirements to the image itself (see section Markerless technology)
Marker-based tracking technology¶
To create a marker-based marker an image of the pattern (the inner part inside the frame) is required. The figure below shows the proportions of the printed marker and the corresponding pattern.
Marker proportions (left) and the contents of the corresponding file with the pattern (right). The dotted frame is here for reference only and should not be included into the marker.
The file with the pattern must contain only the image for the internal part to be put inside the frame including the white margins around the pattern, but excluding the dash-dot frame itself. The image should be black or greyscale on a white background. Make sure the pattern and the background are in sufficient contrast to each other. The pattern should not contain transparent or translucent pixels. It is recommended to use the PNG format, however TIFF, JPG and BMP are supported as well. It is better to use only black and white pixels, and the aspect ratio 1x1. The image size will be automatically resized to 128x128 pixels, and the colors of the image will be converted to black and white (not grayscale), which can affect the quality and contrast parameters of the image.
Markers should preferably be printed on a hard surface to avoid deformation, which may lead to poor recognition. For quality tracking the pattern must not be symmetrical on vertical, horizontal and diagonal axes. The figure below shows examples of wrong and right patters in terms of symmetry.
- Wrong pattern - vertically symmetrical.
- Wrong pattern - horizontally symmetrical.
- Wrong pattern - diagonally symmetrical.
- Correct pattern - not symmetrical.
If multiple marker-based markers are used in a project, then their patterns should visually differ one from another. If the differences are too small then these markers could be interpreted as identical when moved further away from the camera.
How to create a marker¶
To create your own marker, download vector image template and open it in any graphic editor. Place the image of the pattern inside the frame so that it does not intersect with the dashes and dots of the template.There also must be enough space (margins) in between the patternm and the frame. Print the marker in a size that is comfortable to manipulate with without accidentally blocking the frame with your fingers. When using multiple markers in the same project we recommend to keep the same size for all markers.
Color images are also allowed. The system does not recognize colors. Make sure there is enough contrast between the pattern and its background.
Create a pattern by cutting out pattern lines from the marker. Save the file in the format described above. EV Toolbox will convert the image to the desired format, provided that the image does not contain pattern lines. Conversion is done automatically when you add a marker into a project.
You can also use the online marker generator. Drag the pattern file into the frame below, or click on the frame and select a file to create a marker-based marker. Then save the image by right-clicking → on the ‘save image’ option.
Markerless tracking technology¶
Markerless markers are images without borders. The recommended image format is PNG without transparency, however JPG, TIFF and BMP are also supported. The image can be both black and white, as well as color. It can also have an arbitrary ratio sides. Color model is RGB 8 bit.
One of the requirements for the markerlass images is the asymmetry with respect to vertical, horizontal and diagonal axes.
The required image size must be greater than 256 pixels in width and height.
Feature points¶
In order to provide the stable tracking an image must have the so called ‘feature points’ which should meet the following requirements:
Corner element¶
A feature point is a contrasting corner element (corner dot).
| original image | image for tracking with feature points |
|---|---|
 |
 |
The feature points of the image are highlighted in orange (however, such a number of key points is not enough for stable tracking). There are no special corner elements on the image to the right.
| original image | image for tracking with feature points |
|---|---|
 |
 |
There are no corner elements on this image.
Number of the feature points¶
The more feature points the image has - the better is its tracking. Having around 100 points is usually quite enough for stable tracking. If some of the points are not visible during the tracking process, other points will be recognized.
| original image | image for tracking with feature points |
|---|---|
 |
 |
Sample image that has enough feature points
Contrast of the feature points¶
Feature points should be contrasting in tone. Colors of the feature points are not recognized by the tracking system.
| original image | image for tracking with feature points |
|---|---|
 |
 |
Sample image that has insufficient contrast.
| original image | image for tracking with feature points |
|---|---|
 |
 |
In order to create more feature points you can make the background of the image more contrasting.
Uniqueness of the feature points¶
Feature points must be unique. Images with regularly repeating patters and elements are not good for tracking, even if the number of corner elements is sufficient.
| original image | image for tracking with feature points |
|---|---|
 |
 |
This image has a lot of corner elements however they are repeated in a pattern, hence the tracking of such marker will be unstable.
Spreading of the feature points¶
For high-quality tracking it is necessary to have feature points evenly distributed on the image.
| original image | image for tracking with feature points |
|---|---|
 |
 |
On this image the majority of the feature points is grouped in the upper left corner. The remaining part of this image will not be seen by the tracking system which will have a negative effect on the quality of tracking.
| original image | image for tracking with feature points |
|---|---|
 |
 |
To make the tracking of the same image more stable you can crop the image and use only the part with a cluster of feature points.
Scale of the feature points¶
Feature points must be present on images of different scales. Text is not good for tracking.
| original image | image for tracking with feature points |
|---|---|
 |
 |
At a given image scale, the number of points is sufficient for a good tracking.
| original image | image for tracking with feature points |
|---|---|
 |
 |
When the size of the same image is reduced, the number of the feature points is much smaller, which leads to a very poor tracking.
Physical size¶
For quality tracking it is important to specify the exact size of the printed marker in meters in the properties of the “Marker” object.
- For markerless technology the physical size is the width of the image.
- For marker-based technology the physical size is the distance between centers of extreme points or edges of lines of the frame (as shown on the image).
