This page concerns the Vuforia Engine API version 9.8 and earlier. It has been deprecated and will no longer be actively updated. We recommend migrating to the Vuforia Engine 10 API before this page is removed in February 2022. The equivalent of this page or topic can be found here: Object Targets Supported Objects.
Object Targets are designed to be used in use cases with smaller, well textured objects and are a suitable feature to augment tabletop and indoor objects. Use this article to select, prepare and scan your objects successfully.
Object Targets vs Image-Based Targets
Object Targets are a digital representation of the features and geometry of a physical object. They are distinct from image-based target types, such as Image Targets, Multi Targets and Cylinder Targets that require the use of a planar source image. An Object Target is created by scanning a physical object using the Vuforia Object Scanner.
|planar images||publications, packaging, flat surfaces|
|Object targets||feature scan||toys, products, complex geometries|
In addition, it is worth considering if Model Targets are more suitable for your use case if your object(s) fulfill other criteria and if a precise CAD model of the object is available. Model Targets are also supported by third party object scanning methods. For more information, please see Comparison of Object Scanning Methods.
Object Target Types
Object Recognition are optimized for objects that can fit on a tabletop and used indoors. Generally, if it fits on the object scanning target image that is used for capturing the object, it is suitable as an Object Target. However, very small objects that are smaller than 2cm hould be avoided as it will be difficult to capture enough features for steady tracking that, additionally, only can be performed up close.
In addition to selecting the right size, you will also need to ensure that the object has enough visual detail. These details can be contrast, colors, and complex geometry. An object with little to no visual features will not be tracked consistently. These features also need to be consistent between different instances of the same object. If a toy comes with different texturing variants, a single object target will not reliable detect all of them. Consider using Model Targets instead if you object has little or varying textured features.
Another important factor is to limit the deformability of the object. Many moving parts, such as arms, joints, and detachable items, will break detection of the object. Instead, scan only the rigid parts of the object and cull any deformable parts during the scanning process. See the section on Object Scanning Target Image on how to use culling.
Below, we present some examples on objects suitable for Object Recognition: