I Made a Point Cloud Using Tesla Cameras

🚀 Try SurveyXR for free: 👉 https://surveyxr.com 🚀 Join The Survey School: Want to elevate your survey knowledge and stay ahead of the latest workflows in GNSS, RTK, LiDAR, photogrammetry, mobile mapping, and reality capture? 👉 https://thesurveyschool.com 🧪 Submit your research abstract for The Survey School Research Symposium: 👉 https://thesurveyschool.com/symposium 🛰 Check out Emlid products here: 👉 https://store.emlid.com/rami In this video, I show how I used Tesla Model Y dashcam footage to create a survey mapping workflow and generate a point cloud from a consumer vehicle. The idea started with a simple question: can the dashcam footage from a Tesla be synchronized with GNSS data, calibrated with computer vision techniques, and processed inside photogrammetry software to create geospatial data? This video walks through the full research workflow I developed over the last year, including camera calibration, camera localization, lever arm offsets, GNSS data collection, frame extraction, time synchronization, and photogrammetry processing. We start by calibrating the Tesla cameras using a checkerboard target to solve for the intrinsic camera parameters, including focal length, principal point, and distortion. Then we localize the camera orientations using surveyed control points and calculate the exterior parameters for each camera on the vehicle. After that, I measure the lever arm offsets between the Tesla cameras and the GNSS receiver mounted on the roof. This allows the software to correctly position each extracted frame based on the camera location rather than only the GNSS antenna position. Once everything is calibrated, I collect dashcam video while driving around a library entrance and circular island. The GNSS receiver logs position data, the Tesla records video from its cameras, and then SurveyXR synchronizes the GNSS trajectory with the video timestamps to extract geotagged frames. Finally, I bring those extracted frames, camera parameters, and geolocation data into Pix4Dmatic to process the imagery and generate a point cloud from Tesla dashcam footage. This is not presented as a replacement for traditional survey methods yet. It is a research project and proof-of-concept showing how consumer vehicle cameras, GNSS, and photogrammetry can be combined into a mobile mapping workflow. Future videos will focus on improving the results, adding more sensors, and validating the accuracy against traditional survey tools and terrestrial laser scanning. If you are interested in surveying, geospatial technology, GNSS, LiDAR, photogrammetry, mobile mapping, or reality capture, make sure to subscribe. I’ll continue documenting this research as the system improves. If you are working on geospatial research of your own, submit your abstract for The Survey School Research Symposium. We want to showcase new ideas, workflows, and experiments that can move the surveying and mapping industry forward. CHAPTERS 0:00 Turning Tesla dashcam footage into a survey point cloud 1:39 The idea: Tesla cameras + GNSS + calibration 4:14 Camera calibration with the checkerboard 8:08 Survey control and camera localization 13:48 Measuring lever arm offsets 16:53 Collecting mapping data with the Tesla 19:22 Processing everything in Survey XR 27:35 Building the point cloud in Pix4Dmatic 31:05 Reviewing the final results #Tesla #MobileMapping #Surveying #Geomatics #GNSS #Photogrammetry #PointCloud #Emlid #SurveyXR #Pix4D #RealityCapture #LandSurveying #Geospatial #ComputerVision #Dashcam