Jump label

Service navigation

Main navigation

You are here:

Main content

Publication on „A Low Cost Modular Radio Tomography System for Bicycle and Vehicle Detection and Classification“ to be presented on IEEE SYSCON 2021 conference

Another paper regarding CNI's research activities on a low-cost and modular radio-based detection and classification system for road users has been accepted for IEEE International Systems Conference (SYSCON) 2021. Compared to existing solutions, the proposed approach meets multiple crucial requirements, including cost-efficiency, robustness, accuracy, and privacy-preservation, while classifying heterogeneous road users (e.g., cyclists, passenger cars) accurately.

Vehicle_Scenario_Small

The proposed system exploits radio tomography in terms of attenuation patterns and high-resolution channel information provided by WLAN Channel State Information (CSI) and Ultra-Wideband (UWB) Channel Impulse Response (CIR) data. Therefore, we have implemented a four-step data processing pipeline, including data acquisition of real-world traces (fingerprints), data preprocessing (i.e., smoothing and normalization), state-of-the-art Machine learning-based detection and classification, and data exploitation. Specifically, we have gathered different road users' fingerprints on a cyclist path and a busy one-lane road.

We assess the suitability of WLAN CSI and UWB CIR concerning reliable detection and classification tasks of heterogeneous road users using the ML models Artificial Neural Network (ANN), Random Forest (RF), and Support Vector Machine (SVM). Fig. 1 illustrates the five most relevant channel parameters for detecting bicycles extracted from WLAN CSI and UWB CIR using RF and 10-fold cross-validation, respectively. While the coarse-grained RSSI parameter suits best for WLAN, the quotient of the estimated First Path Power (FPP) and the Channel Impulse Response (CIR) power is most important when using UWB. Using the latter FPP/CIR and ANN, we achieve a bicycle detection (binary classification) accuracy of 100%. Further analysis results are provided in our paper.

Bicycle_Detection_CNI_Website

Figure 1: Bicycle detection: Five most relevant channel parameters for WLAN CSI and UWB using Random Forest and 10-fold cross-validation, respectively.

The paper containing the results is accepted for presentation in the course of the upcoming 15th Annual IEEE International Systems Conference (SYSCON) in April 15 - May 15.

Acknowledgment:

This work has been supported by the PuLS project (03EMF0203B) funded by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) and the German Research Foundation (DFG) within the Collaborative Research Center SFB 876 “Providing Information by Resource-Constrained Analysis”, projects A4 and B4.

Reference:

  • M. Haferkamp, B. Sliwa, C. Wietfeld, "A Low Cost Modular Radio Tomography System for Bicycle and Vehicle Detection and Classification", In 2021 Annual IEEE International Systems Conference (SysCon), Vancouver, Canada, April 2021. (accepted for presentation).  [bibtex] [arxiv] [pdf]