public double Granularity { get; set; }

public double Granularity { get; set; }

Public Property Granularity As Double 
	Get 
	Set

Public Property Granularity As Double 
	Get 
	Set

public:
property double Granularity {
	double get ();
	void set (double value);
}

public:
property double Granularity {
	double get ();
	void set (double value);
}

Property Value

The amount of coarse subsampling is controlled by the Granularity. This property specifies the length of one side of a cube in 3D space. The space occupied by your input data is divided into tightly-packed cubes of this size, and your data is subsampled by replacing all of the points within each cube by a single new point that best represents the original points.

The amount of fine subsampling is controlled by the product of GranularityFineFraction and Granularity. In other words, the fine subsampling cube size is a specified fraction of the coarse subsampling cube size.

These cube sizes allow you to control the tradeoff between alignment speed and accuracy. As you make the size of each cube larger the number of trained points gets smaller. This makes the trained pattern smaller and faster, but less accurate.

If your time budget allows, you should set the coarse subsampling cube size to be about half the size of the smallest salient feature in your input data. If you need more speed, try increasing the size. Be sure to verify that you are still getting the accuracy and robustness you need.

The 3D input data that you provide at runtime is also subsampled. The cubes used at runtime are the same size as the ones used to train the pattern. These cubes are aligned to the 3D selected space of the input data.

If GranularityAutoSelect is "true" the coarse subsampling cube size is recomputed, and overwritten, whenever the pattern is trained.

The default value of GranularityFineFraction works well on a wide variety of patterns and run-time data. Typically you would leave it alone. However, in certain circumstances you might want to adjust it to fit particular needs. For example, when RunMode is RefineStartPose, or when you are working with high-quality input data and need to get higher accuracy in the result pose.

You have direct control over GranularityFineFraction. The tool does not attempt to adjust this value even if GranularityAutoSelect is "true". Also note that although specifying a very small fraction can result in a nominal fine subsampling cube size smaller than the vision data's voxel size, the tool will internally clamp the effective fine subsampling cube size to the vision data's voxel size.

If you change this value on a trained pattern object, the pattern will become untrained.