SHARE3DCAM Traffic Accident Scene Recording Technical Solution

I. Background

1. Traditional Methods

In the past, due to technological limitations, 3D LiDAR scanning equipment was not widely used for recording traffic accident scenes, which were primarily documented using traditional manual methods.

Traditional recording methods involve sketching the scene by hand and taking on-site photographs. Once back at the office, the draughtsman uses the sketches and photographs to reconstruct the accident scene and prepare site plans in drafting software.

This approach relies heavily on the experience of both the traffic police officers documenting the scene and the draughtsmen responsible for the final drawings. It is relatively inaccurate and does not support free-view scene review, which can make accident liability determination more difficult in some cases.

2. Handheld SLAM Solution

Given the practical requirements of traffic accident scene recording, the equipment used must be highly accurate, portable and easy to operate.

To address these needs, this solution uses the handheld SLAM laser 3D scanner SHARE SLAM S20 together with the supporting point cloud processing software SHARE PointClouds Studio. The processed data can be imported into AutoCAD to produce site plans with spatial positioning information, thereby improving drawing accuracy.

Where more realistic visual effects are required, the data can also be used for 3D Gaussian modelling for subsequent viewing and measurement. This is an optional enhancement and is not a mandatory requirement under the standard.

II. Solution Overview

1. Laser 3D Scanner and Supporting Software

Data acquisition hardware: SHARE SLAM S20, used to quickly capture raw point cloud data.

Engineering control app: SHARE Capture, used to start and stop projects, configure scan parameters and view real-time data.

Data pre-processing and 3DGS modelling: SHARE PointClouds Studio, used to reconstruct point clouds and output modelling data. With built-in 3DGS functionality, high-quality 3DGS models can be generated directly.

2. Drawing Software

File conversion software: Autodesk ReCap, used to convert point cloud data into a format that can be recognized by drafting software.

Drawing software: AutoCAD, used to draft directly on the converted data.

3. Cloud Storage Software

File management software: Google Drive, which can be used on mobile phones, computers and other devices to back up files to the cloud with one click, and to view, download and share them at any time.

4. Technical Workflow Overview

Site survey: Before commencing work, visit the site to familiarize yourself with the surroundings, plan the scanning route and organize a technical briefing for the team.

Raw data collection: Use the SHARE SLAM S20 to capture raw data. For areas of interest such as number plates, pause for 1 to 2 seconds to allow the camera to take photographs.

Data pre-processing: Use SHARE PointClouds Studio for pre-processing to generate point cloud data and distortion-corrected imagery.

Drawing preparation: Convert the point cloud data into the required format and import it into AutoCAD. Then prepare the site plan in accordance with the relevant specifications.

Modelling and measurement (optional): Perform 3D Gaussian modelling and measurement in SHARE PointClouds Studio.

III. Workflow Steps

1. Field Data Collection and Pre-processing

Take the SHARE SLAM S20 to the accident scene, switch on the device and connect it to your mobile phone. Use SHARE Capture to control the operation, scan the accident vehicle and its surroundings, and carry out a detailed scan of the vehicle itself.

Once the scan has been completed and saved, display the real-time point cloud view in the app to both parties involved in the incident for confirmation. After both parties confirm the result, the field collection work is complete.

Then remove the SD card from the SHARE SLAM S20, insert it into a card reader, connect the reader to your phone via the Type-C port, copy the scanned data to the phone, compress the files and upload them to cloud storage. This allows back-office staff to process the data more quickly and reduces the overall turnaround time.

2. Back-office Processing

When downloading raw data from cloud storage, you can enable real-time transfer. After the upload from the mobile device is complete, the data can be downloaded automatically to the computer without additional manual intervention.

Process the data in SHARE PointClouds Studio and enable Output Distortion-Corrected Photographs in the Mapping Parameters tab to obtain the reconstructed on-site point cloud and distortion-corrected photographs.

3. ReCap Format Conversion

The point cloud data generated during processing cannot be imported directly into CAD software for editing and drafting, so it must be converted using third-party software.

In ReCap, select New Project, then choose Import Point Cloud to import the reconstructed LAS point cloud data. Next, click Import File, Index Scan and Start Project in sequence. After the import is complete, select Save As in the upper-left corner and save the data in the RCP format recognized by AutoCAD to complete the conversion.

4. Floor Plan Drafting

Create a new blank drawing in AutoCAD, open the Insert tab, select the Attach tool and import the converted RCP data.

After the import is complete, trace along the areas that need to be drawn.

Using the point cloud data as a reference, road edges, lane markings and vehicle outlines can be drawn, while key information such as vehicle positions, points of impact and skid marks can be annotated. The final drawing can then be produced in accordance with the relevant standards for traffic accident scene diagrams.

The final output can be used for accident determination, insurance assessment and forensic evidence collection.

5. 3D Gaussian Modelling and Measurement (Optional)

In SHARE PointClouds Studio, the 3DGS feature requires login access. Before scanning, the photo interval must be set to 0.5 seconds or less to capture sufficient features. During processing, both Point Cloud Colouring and Distortion-Corrected Photo Output must be enabled.

After the point cloud has been processed, click 3D Gauss, import the output folder generated in the previous step, and configure the scene type, processing preferences and number of iterations before starting the modelling process.

The generated 3D Gaussian model is automatically loaded into the Point Cloud Manager interface. Select the Measure tool from the top toolbar, then choose Two-Point Measurement to perform measurements.

6. Estimated Time Required

Note: Using a simulated accident scenario involving a collision between a bicycle and a car as an example, the scanned area is approximately 10 m². Assuming the raw data size is about 1 GB per minute of scanning, a 3-minute scan produces around 3 GB of data, or approximately 3,000 MB. The cloud upload speed is calculated at 5 MB/s, and the download speed is calculated at 15 MB/s.