Table 1 Comparison of existing methods for EV charging Port detection and localization.

Zhang et al.¹³

HSI color model for segmentation;

Median filtering for noise reduction;

Morphological operations and Canny edge detection

Sensitive to extreme lighting;

Limited robustness at long distances

Accuracy: 100% (under controlled conditions)

Tadic⁸

Depth sensor for 3D scene reconstruction;

Morphological operations for object extraction;

Simple image processing pipeline

Requires controlled illumination;

Limited to CCS2 sockets

Detection rate: 94%

Zhao et al.¹¹

Modified YOLOv4 for recognition;

Meanshift clustering for noise removal;

Affine correction for coordinate refinement

Requires large labeled dataset;

Limited to indoor environment

Detection success rate: 100%

Avg. processing time: 27ms

Mahhadevan

et al.¹²

SWIN-Transformer for global context;

SimAM attention mechanism

High computational complexity;

Requires powerful hardware

mAP0.5: 81.4

FasterEVPoints

Integrating PConv, SE, and a modified PAN with FasterNet to balance speed and accuracy;

PnP algorithm and BA optimization algorithm for Pose estimation;

Modified NMS for post-processing

Potential inaccuracies for different charging ports;

Limited adaptability under extreme conditions

Accuracy rate: 95%;

mAP0.5:0.95: 33.3