Search

In recent years, the surge of artificial intelligence has propelled autonomous driving technology to the forefront, capturing growing interest and enthusiasm. As a sub-module within autonomous driving, research in traffic sign detection and recognition has significantly advanced with a growing focus on utilizing deep learning methods. Nevertheless, in complex real-world road scenarios, traffic signs often exhibit a long-tailed distribution, with the majority of instances concentrated in a few frequent categories and a scarcity in the remaining ones. Considering conventional Traffic Sign Detection and Recognition methods are crafted using manually curated datasets, the class imbalance may detrimentally impact the efficacy of the detection model. In this paper, we first propose a gradient-guided loss reweighting model that dynamically reweights the loss for positive and negative samples based on the cumulative gradients across each category. Additionally, a classification bias-based refinement module is proposed to fine-tune these weights during training, based on the false positive and false negative rate. This serves to suppress a drop in precision for tail categories resulting from the gradient-guided loss reweighting module, thus further balancing the entire training process for improved results. Extensive experiments are performed on the TT100K and GTSDB datasets, yielding significant advancements surpassing state-of-the-art methods.

A timely fault diagnosis is imperative to significantly improve the operational stability and safety of rotating machinery. Until now, several parameterized time-frequency (TF) analysis (TFA) methods that incorporate the proportional relationship of instantaneous frequencies (IFs) have been developed specifically for analyzing vibration signals of varying speed rotating machinery. However, they are sensitive to noise and exhibit limited TF resolution due to the truncation strategy, particularly in the case of narrow frequency intervals between neighboring components. Therefore, this article proposes a low-rank constrained multikernel general parameterized TF transform (MKGPTFT) to generate an ideal TF plane under challenging conditions characterized by substantial noise and closely spaced IFs. First, the low-rank property of proportional IFs is revealed, and this physical prior information serves as the fundament of the proposed method. Second, a multikernel demodulated operator is constructed to replace the single-kernel demodulated operator and the corresponding MKGPTFT is developed to address multicomponent signals. Third, the low-rank constrained MKGPTFT is proposed by incorporating the low-rank constraint into the update process of IFs of MKGPTFT. Thanks to the incorporation of physical knowledge, the proposed low-rank constrained MKGPTFT demonstrates its capability in accurately characterizing multicomponent signals with closely spaced and proportional IFs and generating a high-quality TF plane. The results of simulation and experiment illustrate the proposed method can discern time-varying IFs with greater precision when compared with six typical comparison methods in analyzing strong noise signals and successfully detect the fault of varying speed rotating machine.