Metal object detection for electric vehicle inductive power transfer systems based on hyperspectral imaging

With the rapid development of wireless charging technology for electric vehicles (EVs), metal object detection (MOD) in charging devices has been widely considered for the operational safety of the system. In this paper, a robust and time-saving MOD method based on the hyperspectral imaging technique and support vector machine is proposed. Since hyperspectral characteristics of different objects highly depend on their materials regardless of sizes and shapes, the proposed method can achieve a good generalization ability after training with very few datasets. In particular, the proposed method can detect a very small-sized metal object regardless of the operation status of the charging system, which is a considerable challenge for conventional methods. Experimental results verify the effectiveness and reliability of the proposed method. The pixel-based detection accuracies for ferromagnetic metal and nonferromagnetic metal objects are 93.4% and 94.2%, respectively, and the object-based detection accuracy for metal objects reaches 100%.

Automated summary

A robust and time-saving metal object detection method based on hyperspectral imaging and support vector machine is proposed in this paper, achieving good generalization ability with very few training datasets. The method can detect very small-sized metal objects with a 100% object-based detection accuracy.