No longer mistaking small dogs for electric vehicles, SingTown Technology introduces an AI visual electric vehicle recognition system.

SingTown Technology基於OpenMV嘅智能相機解決方案，能夠喺複雜嘅真實道路環境中篩選大型車流並準確識別電動車輛。利用人工智能同機器視覺算法，系統可以區分電動車輛同機動車、行人以及其他目標，輸出位置、速度同軌跡資訊，為交通管理、路口執法同安全警示提供可靠嘅視覺感知能力。

In the real-world scenario of urban roads, automatically identifying electric vehicles poses a challenge: electric vehicles are small in size, change speed and direction rapidly, and the difficulties of dense traffic flow, severe occlusions, and complex variations in lighting and weather conditions further increase the difficulty of recognition.

SingTown Technology has proposed an innovative solution: an intelligent camera based on the OpenMV platform, which utilises lightweight AI and computer vision models specifically trained for the contour, handlebar features, body proportions, and motion patterns of electric vehicles to achieve highly robust detection and classification of electric vehicle targets.

The system employs a strategy combining multi-scale detection and temporal tracking: first, suspicious two-wheeled targets are localised in single-frame images through image recognition algorithms, followed by further verification of target identities using inter-frame motion information and trajectory association, thereby effectively reducing the probability of false alarms. To address real-world road conditions, the algorithm incorporates occlusion compensation, backlight compensation, and speed-adaptive thresholds, ensuring stable recognition even in complex environments such as daytime, nighttime, and rain or snow.

OpenMV智能相機可部署於路口、天橋或路邊燈柱。透過多設備協作，實現跨視角目標重識別與軌跡拼接，輸出電動車流量統計、違規逆駛或違例泊車警示等結構化數據。系統支援邊緣部署，識別結果可經邊緣AI本地處理以保障機密與私隱，亦可對接交通管理平台進行信號優化、執法取證或車流分析。

Leveraging SingTown Technology's expertise in machine vision and AI, the OpenMV smart camera provides a lightweight, scalable, and highly practical embedded visual recognition solution for electric vehicles in urban traffic, contributing to enhanced road safety and management efficiency.