Thanks for considering Sweep. This sounds like an interesting application. Internally, we have discussed a few potential features related to CAD. It is worth noting that exporting a 2D scan to a specific file format is very different than automating the generation of a model based off scan data. Of course there exists a gradient between these 2 extremes, inside of which we hope to add features.
The first of these features is to simply export our 2D scan (comprised of individual data points) to many popular file formats, including a CAD friendly file format. This might provide users some reference to use in a CAD program, but would not automatically construct a 2D model or drawing.
Ideally your desired format will be supported soon, but if you need it sooner, you can always convert the data. Recording data to a file using the visualizer currently outputs a JSON file. Something we opted for temporarily, as it provides a very simple standard for users to convert from. This file is comprised primarily of individual sensor readings stored in polar coordinates (ie: values for azimuth and distance). The data can be easily converted to cartesian coordinates and stored as a point cloud, assuming you know the target format. If you are unfamiliar with JSON, don’t worry. It is a very simple format. But we should roll out support for basic .csv, PCD and LAS formats very soon, which might make it even easier to convert to your target format. Hopefully that answers your question.
Thinking ahead, we have discussed a feature for creating CAD friendly files in the form of formatted floor plans. In general we will be working on automating feature extraction and ideally floor plan extraction, but will likely add an interactive feature before that. This may take the form of a tool in the visualizer that allows a user to indicate walls overtop the data and then export the user’s outlines as a formatted design file complete with accurate dimensions etc.
Again, it is worth noting that these are all 2D features. The current version of the Sweep sensor is a 2D scanning LiDAR. That means data is only collected in a 2D plane (slice). This kind of data is often used for obstacle avoidance or detection. We have plans to release some DIY info for collecting a 3D scan using our device, but our current software priorities involve the support of features related to 2D scans. Having said that, we do intend to add support and features for 3D LiDAR scans in the future.
I’m not sure to what degree your question was an inquiry about 3D model creation. But we would be interested to hear what kind of tools people would be interested in for the future of Scanse sensor technologies, including the possibility of 3D model generation. There is a large field of work that deals with converting point clouds into useable models. There are many techniques for meshing a point cloud but the end results are usually of a form or quality only useful as a reference in a CAD design project. There do exist software methods for extracting useful CAD designs from high density point clouds (think EdgeWise or InfiPoints) but these have limitations. Usually, such programs are used for commercial surveying where software helps automate the extraction of a known structural primitive (such as a specific type of steel beam, or a pipe). Additionally good results usually come from the automated fitting of primitives based on the manual direction or input of a user. Ie: a user indicates where in the point cloud a certain object should be fit, and a software algorithm performs the fitting. Lastly, these methods generally require 3D point clouds with significantly higher density than what the current Sweep sensor provides. However, these applications are of interest to us and and we are discussing them internally.
Let us know if you have any further questions, or if you run into any specific troubles when converting data.