Hello I am using ZED X with Jetson Orin nano to record SVO files and process spatial map. Now, the scan is quite full of holes and untidy. I saw that here Area Memory & VSLAM Mapping | StereoLabs it is possible for the SDK to save .area file to be used in the iterative manner. Is it possible to do the same with spatial mapping? Can we iteratively generate a mesh with multiple SVOs?
Hi Punn,
Short answer: there’s no .area equivalent for spatial mapping. The SDK can’t load a previous mesh and keep integrating into it : enableSpatialMapping() always starts a fresh map, and extractWholeSpatialMap() returns that session’s map only.
What you can reuse is the coordinate frame. Per the docs, mapping data is saved relative to a fixed World Frame, and if Area Memory is enabled with an Area file provided at initialization, the map loads repeatedly across sessions and always appears in the same physical location. So record one .area covering the site, load it on every SVO run, and each mesh comes out in the same frame. Merging then becomes a concatenation rather than an ICP registration. Relocalization needs visual overlap between recordings. Merge offline in CloudCompare or Open3D.
Before going down that road: holes usually mean the single pass is underconstrained. For SVO reprocessing use ZEDfu in OFFLINE mode with NEURAL_PLUS depth, that’s the best-quality path. Also watch getSpatialMappingState() for FPS_TOO_LOW / NOT_ENOUGH_MEMORY, which stop integration silently.
Which SDK version are you on?
Best regards,
Tristan
Stereolabs Support