Show only the CCTV cameras in view, thinned to one per ground cell when zoomed out

The CCTV layer kept every loaded camera live: each view change touched every
icon, every notification rebuilt every camera's state, and the picker listed
the whole catalog. With the default 4,000 cameras, a continental drag ran at
152-179 fps with 18-24 frames over 30 ms (worst 112 ms). It now runs at
235 fps with none. The layer now:

- shows icons only on or near the screen, and when zoomed out keeps one camera
  per world-anchored cell (healthy feeds first), so panning keeps the same
  cameras;
- grounds and draws coverage only for the cameras in view, the active camera
  and its coverage neighbours, retrying a camera at most twice per view regime;
- reruns the view pass while the camera moves, paced by its own cost, and
  projects with one matrix per pass instead of a Cesium call per camera;
- keeps the billboard buffer-usage hint fixed, so showing or hiding a few
  icons no longer rewrites the vertex data of every billboard;
- lists the cameras in view (active camera first) in the picker, closing with
  the count of cameras outside the view, and builds the full camera list in
  the UI state only when it is read.
This commit is contained in:
Sameh Khamis committed 2026-10-29 21:31:59 -07:00
1 parent f13c60c477
commit 93942d0dbe
32 files changed
+1301 -71

No files matched your search

+39
View File
@@ -60,6 +60,45 @@ export const GEO_PROGRESS_NOTIFY_INTERVAL_MS = 300;
export const GEO_PROGRESS_NOTIFY_BATCH_LIMIT = 10;
// View thinning: camera icons show only on or near the screen, and each
// camera competes for a world-anchored grid cell sized to about VIEW_CELL_PX
// at its distance from the viewer (in power-of-two metre steps, so a pan
// keeps the same winners until a camera's distance crosses a step). Cells
// smaller than VIEW_MIN_CELL_M do not thin:
// close up every camera shows, co-located ones included.
export const VIEW_CELL_PX = 40;
export const VIEW_MIN_CELL_M = 60;
// Icons this far past the screen edge (fraction of its size) still show, so
// a pan does not reveal empty ground before the next view pass.
export const VIEW_ICON_MARGIN = 0.5;
// While the camera moves, the view pass reruns at most this often.
export const VIEW_MOTION_REFRESH_MS = 250;
// A motion tick is skipped unless the camera moved this fraction of its
// ellipsoid height, or turned this many degrees, since the last pass.
export const VIEW_MOTION_MIN_MOVE_FRACTION = 0.01;
export const VIEW_MOTION_MIN_TURN_DEG = 0.5;
// Motion passes are spaced at least this many times the last pass's cost, so
// a large catalog spends at most about a tenth of a long move on them.
export const VIEW_MOTION_COST_FACTOR = 10;
// A view camera's geometry is attempted at most this many times per surface
// regime (a cold elevation cell can leave the first attempt unresolved); the
// retry runs this long after the drain that left it unresolved.
export const VIEW_GEOMETRY_ATTEMPTS = 2;
export const VIEW_GEOMETRY_RETRY_MS = 1500;
// Throttle for placeholder repaints — the projection RAF loop must not
// re-fill a 1080p canvas on every frame while a feed image is still loading.