Live Latency Analysis of IP Cameras in Camera Monitor Systems

Abstract

Camera Monitor Systems are increasingly replacing traditional side and rearview mirrors in the automotive industry. Internet Protocol cameras are well suited for this purpose due to their cost efficiency and configurability. To conserve bandwidth, these cameras transmit compressed video streams that must be decoded by the receiver. This encoding and decoding, along with network transmission and rendering, introduces system latency, i.e., the delay between frame capture and display on the vehicle’s monitor. Strict regulations limit this latency to a maximum of 200 ms. However, existing latency measurement methods require external tools, limiting their use for continuous runtime monitoring. This thesis aims to develop a modular, runtime-capable software tool for latency analysis. It enables the identification of optimal configurations with respect to hardware platform, video codec, network type, and number of simultaneous video streams. Accordingly, the research addresses two key questions: (1) how to measure latency runtime without external tools and (2) which system configuration meets the latency requirements while supporting stable decoding of multiple streams. To this end, a novel latency measurement library is developed using the open-source multimedia framework GStreamer. It measures latency in real time by embedding and extracting timestamps within the video stream. This method supports both inbound latency analysis from the camera side, which requires clock synchronization, and decoding latency assessment. The functionality of the library is validated through stopwatchbased experiments. Subsequently, the tool is used to evaluate various hardware platforms, video codecs (H.264 and MJPEG), network types (Ethernet and Wi-Fi), and number of concurrent streams. The results demonstrate that the proposed software solution reliably measures latency at runtime. While hardware-induced latencies beyond software measurement affect precision, it remains a reliable indicator of latency spikes and elevated latency conditions. The findings highlight the critical role of dedicated hardware decoders in achieving low latency. Decoder performance strongly depends on both the hardware platform and the video codec. Furthermore, a wired network connection between the cameras and the control unit contributes significantly to lower latency. Moreover, latency increases and rendering performance degrades with each additional stream. Nonetheless, with an optimized system setup, decoding of up to four streams simultaneously is possible while still meeting latency requirements. This thesis establishes a practical approach for continuous latency monitoring in Camera Monitor Systems. It contributes to improved system safety through targeted configuration and timely detection of latency issues.

Date of Award	2025
Original language	English
Supervisor	Franz Leopold Wiesinger (Supervisor) & Stefan Larndorfer (Supervisor)