Vehicle-Type Recognition Method for Images Based on Improved Faster R-CNN Model.

Area of Science:

• Computer Vision and Image Processing
• Artificial Intelligence in Transportation

Background:

• Increasing vehicle numbers cause traffic congestion, accidents, and crime, complicating parking management.
• Vehicle-type recognition technology offers a solution to reduce human workload in vehicle management.
• Image technology is crucial for integrated traffic management systems.

Purpose of the Study:

• To propose an improved Faster R-CNN model for accurate vehicle-type recognition.
• To enhance recognition accuracy by combining features from different convolution layers.
• To optimize the model using contextual features and bounding box strategies for better performance.

Main Methods:

• An improved Faster R-CNN model was developed for vehicle-type recognition.
• Combined output features from different convolution layers to boost recognition accuracy.
• Integrated contextual features and an object bounding box optimization strategy.

Main Results:

• The improved model effectively identifies vehicle types, including cars, SUVs, and vans.
• Achieved average precision (AP) of 83.2% for cars, 79.2% for SUVs, and 78.4% for vans.
• The mean average precision (mAP) showed a 1.7% improvement over the traditional Faster R-CNN model.

Conclusions:

• The proposed enhanced Faster R-CNN model demonstrates superior performance in vehicle-type recognition.
• This technology is significant for improving integrated traffic management and reducing operational burdens.
• The method provides a robust solution for automated vehicle classification in real-world scenarios.