Invasive plant species pose severe ecological and economic threats to agricultural productivity, biodiversity, and ecosystem stability worldwide, necessitating rapid and accurate detection methodologies for timely intervention. Traditional monitoring approaches, including manual surveys and laboratory-based identification, are labor-intensive, time-consuming, and impractical for large-scale or remote ecosystems. Edge artificial intelligence (Edge-AI) represents a transformative paradigm that enables real-time, on-device machine learning inference directly at the point of data acquisition, eliminating cloud dependency and enabling immediate decision-making in field conditions. This review examines the integration of edge computing platforms, embedded machine learning models, and multi-sensor imaging technologies for autonomous identification of invasive plant species in diverse ecological and agricultural settings. We discuss critical aspects of on-device model deployment, including lightweight convolutional neural networks, model quantization, feature extraction optimization, and inference acceleration techniques tailored for resource-constrained hardware. Applications spanning precision agriculture, ecological restoration, protected area management, and rapid response systems are evaluated alongside performance metrics, validation strategies, and real-world deployment case studies. The review concludes by addressing fundamental challenges including hardware limitations, environmental variability, model generalization, energy constraints, and scalability barriers, while highlighting future opportunities for integrating Edge-AI systems with unmanned aerial vehicles, Internet of Things ecosystems, and automated management platforms to achieve comprehensive, cost-effective invasive species surveillance and control.