J. Eur. Opt. Society-Rapid Publ. 21, 31( 2025) 17
Fig
. 18. Detection results on the Cement Street dataset.
Fig. 19. Detection results on the Holly dataset.
higher false alarms in others. Supervised detection methods generally provided results with higher false alarms, while unsupervised detection methods tended to yield lower detection rates. Table 4 presents the objective evaluation metrics, specifically the AUC values of detection results for different dimensionality reduction methods under various detection methods. The proposed method achieved the best performance in four out of the six detection methods and demonstrated the highest average detection performance, proving that the proposed dimensionality reduction method effectively supports downstream tasks.
3.5 The impact on time consumption
This section will examine the impact of dimensionality reduction on the running time of detection methods. Figure 21 shows the variation of running time under different detection methods, with the y-axis representing the