This Homemade 3D Scanner Can See in Total Darkness

A handheld scanner now allows anyone to map underground regions in 3D, even when it is completely dark and no light is coming in. Maker 9nl designed this technique expressly to assist with cave excursions, where GPS is useless and standard surveys are just rough sketches. Commercial versions of this technology can cost upwards of $40,000 and include a restricted software package, but this open source construct throws a kink in the works.

The core element is from a used Velodyne VLP-16 LiDAR sensor, which can be purchased on eBay for roughly $400. Brand new devices will cost a little fortune. This LiDAR emits 300,000 laser pulses per second, measuring the distance to every surface in its range of view, regardless of whether the space is a cave or a dark room. There is no need for a camera or visible light, so don’t go fancy.

A conventional LiDAR setup like this will only scan a level plane because its internal mechanism spins horizontally while shooting a 30-degree vertical beam. However, more motion is required to produce a real 3D image. So 9nl created a special rotating platform to mount the sensor on, which can turn in a second axis. They also added a motor controlled by a Teensy microcontroller to keep the entire lidar setup rotating steadily. The combination provides a true 360-degree scan as the user passes through the corridor.

An inertial measuring unit was employed, the Wheeltech N100, to determine the proper orientation and track position. This little guy measures acceleration and rotation 400 times per second and feeds the data into the software. This in turn calculates how the device is shifting and twisting in real time. If you didn’t have the IMU data, the map would quickly become disjointed, and all of the processing takes place on a Raspberry Pi 5. The Pi uses ROS2, the second iteration of the Robot Operating System. ROS2 manages communication between the LiDAR, IMU, and mapping algorithm, logs raw sensor data, and then applies SLAM (Simultaneous Localization and Mapping) to stitch the points together into a coherent 3D model. The user just walks or crawls inside the cave while holding the battery-powered cage, and the system generates the map in real time.

One major advantage is the overall portability of the equipment, as it is small enough to fit into tight spaces and has enough battery life to endure an adventure. External beacons, anchors, and assistance are also unnecessary. Lasers can cut through the darkness that would obscure any optical approach. So far, the findings from several above-ground tests look promising; the point cloud reveals great, detailed geometry; however, the true problem lies ahead: getting them to operate in a real cave, where there’s muck and uneven footing to cope with. Early runs, however, appear promising, and this could finally make it simpler to abandon those sloppy hand-drawn maps that lack dependable up-and-down information.

9nl intends to release a few more updates soon to break down the various aspects of the project, including LiDAR details, ROS2 setup, motor control with the Simple FOC library, the mechanical design of the gimbal, and how the SLAM integration works.
[Source]