LiDAR Visualization Module — ROS2 + Qt + VTK

Context: 3D visualization module developed during my collaboration with Logistical (Tokyo, logistical.jp), a company building freight-consolidation software for B2B shippers in Japan — combining smaller shipments into shared truckloads to cut costs and delivery times.

Logistical runs several products (delivery apps, an AI chatbot, and others); I worked specifically on their truck-loading optimization tool, the one project built in C++/ROS2, focused on estimating cargo volume and occupancy from LiDAR data.

Repo: github.com/Logistical-dev/visualpoint — private, as it depends on the company's proprietary ROS2 framework. Feel free to reach out to them directly, or to me for more details. This README documents and showcases the work I did.

---

What it does

The application connects to an existing ROS2 pipeline and renders in real time, in 3D:

• Point clouds — LiDAR point clouds colored by intensity
• Voxels — volumetric occupancy grid
• Object detections — detected object markers

It also displays live metrics (fill ratio, visibility, footprint, total volume, occlusion, max height) computed by the perception pipeline.

---

My role

At first I considered Foxglove for the visualization layer, but the team needed a fully customizable interface (custom metrics panel, filtering controls, specific interactions), so I built it with Qt + VTK instead.

I worked on the C++/Qt/VTK visualization layer, integrating it with an existing ROS2 pipeline (sensors, object detection, voxelization) built by the team. Specifically, I built:

• PointCloudRenderer — converts sensor_msgs::PointCloud2 messages into VTK geometry. Includes a custom lookup table mapping intensity → color, with dynamic color range calculation per frame.
• MarkerRenderer — manages the lifecycle of VTK actors (creation, pose/color updates, deletion) from visualization_msgs::MarkerArray messages, reused for both voxels and object detections.
• The Qt interface itself, built with Qt Designer: a left panel with checkboxes to filter each data type (point cloud, voxels, objects), a clear scene action, and a metrics panel with progress bars and live values driven by the incoming pipeline messages.
• VTK viewport integration: embedded a QVTKOpenGLNativeWidget inside the Qt layout to host the 3D scene, wired it to MainWindow (camera auto-framing based on scene bounds, render triggers on checkbox toggle, double-click-to-fit-view).
• Consumed a thread-safe mailbox system (built by the project's lead dev) to receive ROS2 data on the Qt main thread without blocking rendering.

Tech stack

C++ · Qt (QMainWindow, QVTKOpenGLNativeWidget) · VTK (actors, mappers, lookup tables) · ROS2 (rclcpp, subscriptions, custom msg types) · CMake

---