I’ve been trying to solve this problem for a couple days now, but I’ve run into several roadblocks that are hindering my progress. My goal is to render a point cloud in Cesium using the new 3d tiles format, without showing a globe or any object other than the point cloud. Visually, the point cloud will hover on a black background, with no other object visible, and the user can pan, tilt, and zoom around the point cloud. So far I’ve successfully rendered the point cloud in Cesium, so my remaining challenges are removing the globe and tuning the camera controls. While it’s easy enough to remove the globe when initializing the Cesium viewer, the resulting camera controls are less than satisfactory. If the globe is present, the camera acts how I expect, and rotates around a point picked from the point cloud or the globe, depending on what the cursor has picked. Removing the globe changes this behavior, and any attempt to pan or tilt the model seems to tilt the camera around a point far from the 3d tileset, making camera control very awkward.
I’ve tried two approaches to fix this: reimplementing the picking behavior at a low level, and using the viewedEntity tracking behavior. Since I’d still like users to be able to pan across the point cloud, the tracking behavior doesn’t really work for me, and all my attempts at picking from a point cloud have been unsuccessful.
Based on these attempts, I have two questions:
Is there a way to retain standard camera behavior while removing the globe?
If not, is there a way to disable globe picking while rendering the globe invisible, or something to that effect? Ideally, if the user clicks a point on the globe (unknowingly, since it is not visible), the camera will not respond by panning/tilting around that distant point.
I am aware this is a fairly unusual use case, and cesium may not be the best tool for the job, but the ability to render point clouds in the browser with varying LOD would be very useful in my application.