Improved YOLOv5-Based Electric Bicycle Detection Algorithm in Elevators Using State Space Models.

Saved in:

Bibliographic Details
Title:	Improved YOLOv5-Based Electric Bicycle Detection Algorithm in Elevators Using State Space Models.
Authors:	Lingzhi Wang, Yingfan Wu
Source:	Engineering Letters. Nov2025, Vol. 33 Issue 11, p4584-4592. 9p.
Subjects:	Electric bicycles, Elevators, Real-time computing, Object recognition (Computer vision), Feature extraction, State-space methods, Edge computing, Fire risk assessment
Abstract:	As electric bicycle use grows, indoor chargingrelated fires have surged, posing serious risks to life and property. Therefore, a real-time and accurate method for detecting electric bicycles in elevators is of great research significance. Addressing complex equipment issues and elevated false positive rates in existing detection methods, this paper proposes an electric bicycle detection algorithm for elevators based on an improved YOLOv5 using a State Space Model (SSM). First, lightweight GhostConv replaces the standard convolution to reduce the number of parameters. Second, the SSM-Conv module is introduced to replace the C3 structure, leveraging the context modeling of the Mamba block to enhance feature extraction capabilities. Finally, a 2D selective scanbased SS2D-Fusion module is designed for feature fusion and integrated into the Neck part, aiming at improving feature fusion capabilities through cross-layer information interaction and enhanced context modeling. Additionally, we incorporate a sample-difficulty-aware SlideLoss into the training framework to dynamically balance the loss contributions between hard and regular samples. Experimental results demonstrate that the improved YOLOv5n model achieves a 1.6% increase in mAP0.5 and a 6.9% increase in mAP0.5:0.95 compared to the original model, while reducing the number of parameters by 14.7%. Consequently, the proposed model enhances detection accuracy and reduces the parameter count, making it well-suited for edge computing devices. It effectively improves both the accuracy and real-time performance of electric bicycle detection in elevators. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	As electric bicycle use grows, indoor chargingrelated fires have surged, posing serious risks to life and property. Therefore, a real-time and accurate method for detecting electric bicycles in elevators is of great research significance. Addressing complex equipment issues and elevated false positive rates in existing detection methods, this paper proposes an electric bicycle detection algorithm for elevators based on an improved YOLOv5 using a State Space Model (SSM). First, lightweight GhostConv replaces the standard convolution to reduce the number of parameters. Second, the SSM-Conv module is introduced to replace the C3 structure, leveraging the context modeling of the Mamba block to enhance feature extraction capabilities. Finally, a 2D selective scanbased SS2D-Fusion module is designed for feature fusion and integrated into the Neck part, aiming at improving feature fusion capabilities through cross-layer information interaction and enhanced context modeling. Additionally, we incorporate a sample-difficulty-aware SlideLoss into the training framework to dynamically balance the loss contributions between hard and regular samples. Experimental results demonstrate that the improved YOLOv5n model achieves a 1.6% increase in mAP0.5 and a 6.9% increase in mAP0.5:0.95 compared to the original model, while reducing the number of parameters by 14.7%. Consequently, the proposed model enhances detection accuracy and reduces the parameter count, making it well-suited for edge computing devices. It effectively improves both the accuracy and real-time performance of electric bicycle detection in elevators. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X