A comprehensive swarming intelligent method for optimizing deep learning-based object detection by unmanned ground vehicles

Area of Science:

• Artificial Intelligence
• Computer Vision
• Robotics

Background:

• Unmanned ground vehicles (UGVs) rely on artificial intelligence for environmental perception.
• Deep learning object detection, particularly Faster R-CNN, is crucial for UGV perception.
• Existing region proposal networks (RPNs) have limitations in exploration space.

Purpose of the Study:

• To develop an improved region proposal network (RPN) for enhanced object detection in UGVs.
• To address the limitations of existing RPNs in adapting to diverse object shapes and learning from negative samples.
• To boost the overall performance of deep learning-based object detection for UGV applications.

Main Methods:

• Introduced a boosted region proposal network (BRPN) with three key enhancements.
• Designed a novel enhanced pooling network for adaptability to various object shapes.
• Improved the BRPN loss function for better learning of negative samples, optimized using the Grey Wolf Optimizer (GWO).
• Integrated a novel Genetic Algorithm-Support Vector Machine (GA-SVM) classifier to enhance classification capabilities.

Main Results:

• The BRPN demonstrated adaptability to objects of different shapes.
• Optimization of the BRPN loss function using GWO significantly improved its performance.
• The GA-SVM classifier enhanced the overall classification accuracy.
• The proposed deep learning object detection method achieved excellent experimental results on PASCAL VOC 2007, VOC 2012, and KITTI datasets.

Conclusions:

• The boosted region proposal network (BRPN) effectively enhances object detection for unmanned ground vehicles.
• The proposed method shows significant improvements in adapting to object variations and classification accuracy.
• This research contributes to advancing environmental perception capabilities in autonomous systems.