Automated Object Measurement with a Total Station Using Instance Segmentation

(Geodata Technology M.Eng. Thesis Project — THWS, 2025)

---

Overview

This repository contains the public version of the software prototype developed during my master's thesis in Geodata Technology (M.Eng.) at Technische Hochschule Würzburg-Schweinfurt (THWS), 2025.

The project evaluates whether a Computer Vision (CV) pipeline can support automated object measurement with a total station. It combines:

• YOLOv8 (Large) instance segmentation
• Canny edge detection
• Progressive probabilistic Hough line Transform
• Harris Corner Detection
• DBSCAN clustering
• GeoCOM-based device control of a Leica MS50 (all GeoCOM commands removed in this public version)

⚠️ Important: The original measurement hardware, the trained YOLO model, and the physical object used for testing are unique and not included. Images from other scenes cannot be segmented without retraining a model.

---

Workflow Overview

The pipeline consists of six main CV steps (a–f) and an optional measurement step:

(a) Example image after acquisition (e.g., from the total station). (b) Preprocessed image patch after segmentation, cropping, and masking. (c) Result of the Canny edge detection showing fragmented edge traces. (d) Detected line structures obtained using the Hough Transform. (e) Identified object corner points using Harris Corner Detection, grouped via DBSCAN clustering. (f) Placeholder icon representing the TXT file in which the corner-point pixel coordinates are stored. (Optional) Total station control with the pipeline result via GeoCOM (GeoCOM commands not included in this repository, only schematic flowchart)

---

Scripts and Their Role in the Workflow

1. Object_Segmentation_and_Canny_Edge_Detection_App.py (Steps a–c)

• Runs YOLOv8 instance segmentation
• Crops the detected object (in this workflow the used steel beam)
• Opens an interactive Canny GUI with manual + automatic methods
• Saves cropped object and edges

2. Hough_Line_Transform_App.py (Step d)

• Interactive GUI for the Progressive probabilistic Hough Transform
• Tunable parameters + live preview of detected lines
• Saves a line-only image on black background

3. Harris_Corner_detection_and_DBSCAN_Clustering_App.py (Step e)

• Harris Corner Detection + DBSCAN clustering
• Tunable parameters + live preview of detected corner clusters
• Saves: visual output and list of final cluster centers (clustered object corners)

4. main_total_station_control.py

• Central script intended to control a total station during the automated measurement workflow
• Sequentially triggers all three CV modules (segmentation → Hough → Harris) and integrates their outputs into the measurement loop
• All GeoCOM commands removed; placeholders indicate where users must insert their own device-specific total-station control
• Preserves the original architecture so the full pipeline can be rebuilt on Leica instruments or adapted to other manufacturers

---

Repository Structure

project/
│
├── README.md
├── requirements.txt
│
├── assets/
│   ├── computer_vision_pipeline.jpg
│   └── total_station_workflow_.jpg
│
├── segmentation_and_canny/
│   └── Object_Segmentation_and_Canny_Edge_Detection_App.py        # (a)-(c) YOLOv8 (Large) segmentation + Canny
│
├── hough_transform/
│   └── Hough_Line_Transform_App.py                                # (d) Progressive probabilistic Hough Transform
│
├── harris_corner_detection/
│   └── Harris_Corner_detection_and_DBSCAN_Clustering_App.py       # (e) Harris Corner Detection + DBSCAN corner clustering
│
└── total_station_control/
    └── main_total_station_control.py                              # (d) Structure for total-station workflow (GeoCOM removed)

---

Requirements

Install dependencies:

pip install -r requirements.txt

---

License

This project is licensed under the MIT License – see the LICENSE file for details.

---