pcopenlod

On this page	* Distance Queries * Solid Angle Queries * Aggregation * Example: Proximity Query * Example: Threshold Solid-angle Query * Example: Limited Solid-angle Query

int  pcopenlod(string filename, string Pchannel, vector P, int min_pts, ...)

This function opens a point cloud file (.pc) and queues up access to the points contained in it. You can then iterate over the points with pcunshaded or pciterate and add new data to the point cloud using pcexport.

While this function is similar to pcopen, the major difference is the points that it queues up may be aggregates of entire groups of points. In other words, a single point may represent many points. This allows you to perform queries at any desired level of detail without ignoring points in the point cloud. For example, you can perform a query in which points near the query origin are queued up as usual, but points far from the origin are averaged. This can lead to dramatic performance increases because entire groups of points can be processed as if they are a single point.

As in pcopen, P specifies the query origin and Pchannel specifies the position channel. During construction, the tree structure starts out as a single bounding box that encompasses all the points in a point cloud, and is recursively subdivided until there are fewer than min_pts points in a node - at which point subdivision stops and a leaf node is created. A good default for min_pts is 8.

Queries are performed by descending the tree structure from the root node until some condition is met. Conceptually, you start with a coarse query and refine it until you decide that it is detailed enough. You use a measure to decide when the query has the desired level of detail. Two measure values are supported: distance and solidangle. Distance Queries

distance-queries

distance mode is provided for compatibility with pcopen and does not queue up aggregate points. Distance queries take a threshold parameter that indicates the radius within which to accept points.

The threshold argument specifies the radius within which points are accepted - identical to the radius passed to pcopen. For example, calling pcopenlod(…, "measure", "distance", "threshold", radius, …) queues up points that lie within the specified radius of the query origin.

Solid Angle Queries

solid-angle-queries

Solid angle queries prioritize points by how close they are to the query point and also by the area of the point, so points that are close to the query point and that have a large area are given a greater weight. The query process will tend to split points with a larger contribution by queueing their children.

The exact equation used to compute point contribution is the following:

Ai / ||Pi - P||^2,

where Ai is an aggregate area value, Pi is the closest point to P in the aggregate box, and P is the query origin. Calling pcopenlod(…, "measure", "solidangle", "area", "A", …) performs a solid-angle query in which the A channel is assumed to hold area values.

There are two different ways to use the solid angle query - an unlimited (threshold) query which returns a different number of points depending on how many points meet the given threshold, and a limited (samples) query which always returns the same number of points. If a samples argument is present, a limited query is assumed.

Limited queries work by prioritizing rather than thresholding samples - so that regardless of the total weight of the points being considered, the same number of points are returned. The algorithm works by iteratively picking the point that has the greatest contribution and splitting that point until enough points have been split to meet the desired sample count. Limited queries are useful when you need a fixed performance or minimum quality level for the query.

Threshold queries work by comparing the point contribution to a fixed threshold - and accepting or rejecting the point based on this comparison. Since different query points lead to different point contributions, a variable number of points will be queued up for threshold queries. Threshold queries are useful when it is acceptable to use a lower number of points for query positions that are far from the point cloud.

Aggregation

aggregation

Additional string parameters indicate how point values are aggregated. Each channel can have a different aggregation mode: mean, sum, or weighted. Calling pcopenlod(…, aggregate:P, sum) will aggregate the values in channel P by summing them. Calling pcopenlod(…, aggregate:A, weighted, weight, W) will aggregate the values in channel A using a weighted mean with weights from channel W.

Example: Proximity Query

example-proximity-query

int handle = pcopenlod(texturename, "P", P, 8,
"measure", "distance", "threshold", 2.0,
"aggregate:P", "mean",
"aggregate:value", "sum");
Cf = 0;
while (pciterate(handle))
{
pcimport(handle, "value", valueSum);
Cf += valueSum;
}
pcclose(handle);

Example: Threshold Solid-angle Query

example-threshold-solid-angle-query

handle = pcopenlod(texturename, "P", P, 8,
"measure", "solidangle", "area", "A", "threshold", 0.01,
"aggregate:A", "sum",
"aggregate:irradiance", "weighted", "weight", "A",
"aggregate:P", "mean");
Cf = 0;
while (pciterate(handle))
{
pcimport(handle, "irradiance", irradiance);
Cf += irradiance;
}
pcclose(handle);

Example: Limited Solid-angle Query

example-limited-solid-angle-query

handle = pcopenlod(texturename, "P", P, 8,
"measure", "solidangle", "area", "A", "samples", 4,
"aggregate:A", "sum",
"aggregate:irradiance", "weighted", "weight", "A",
"aggregate:P", "mean");
Cf = 0;
while (pciterate(handle))
{
pcimport(handle, "irradiance", irradiance);
Cf += irradiance;
}
pcclose(handle);