Overloaded trucks pose a severe threat to highway traffic, increasing the rate and severity of accidents, while damaging road infrastructure. Enabling an efficient and low-cost overload detection method at existing weighing stations would facilitate regulatory compliance. This paper presents a real-time, accurate truck overload detection system that leverages existing surveillance cameras installed at weighing stations. To achieve this goal, we applied computer vision algorithms on video from an indoor camera monitoring a digital display and on another outdoor camera monitoring the station’s weighing bridge. The truck’s actual weight is obtained by reading the numeric digit display on images from the indoor camera, while the truck’s maximum load capacity is estimated by recognizing the its wheel layout using the video from the outdoor camera. Through evaluation using video data from a weighing station, the optical digit recognition algorithm achieves an accuracy of 99.0%, while the load capacity estimation algorithm achieves an accuracy of 93.18%. Furthermore, our system can achieve an accuracy of over 90% among 21 overload events. Finally, we conclude with a discussion of potential optimizations and other future work.

numeric digit recognition;smart weighing station;vehicle classification;overload detection