Event camera calibration plays an irreplaceable role in event camera-based visual applications. Event cameras report per-pixel brightness changes as a stream of asynchronous events, making them insensitive to the spatial distribution of feature points. Consequently, traditional methods such as checkerboard and circle calibration algorithms, which rely on spatial intensity distribution for feature point extraction, are ill-suited for event camera calibration. To address this issue, a flexible and accurate event camera calibration method is presented in this work. The method analyzes the frequency characteristics of event data streams, introducing the groundbreaking use of Fourier transform in event camera calibration. Instead of relying on intensity, we extract feature points from the robust phase domain, which mitigates the impact of event camera noise. Event frames are derived from the event data streams, and we recover two crossed phase maps using Fourier transform. Feature points are then detected using the law of phase distribution and refined with sub-pixel accuracy. Subsequently, camera parameters are estimated. The performance of our method was rigorously validated from different perspectives, and the results show that the accuracy of proposed method is substantially improved compared to the conventional methods. The root mean square errors (RMSE) of the proposed method are consistently below 0.06 pixels, with some cases achieving an impressive 0.03 pixels. The accuracy of binocular reconstruction reaches 0.394 mm.

Spiking Neural Networks (SNNs) achieve an impressive performance due to low power consumption and quick inference on neuromorphic hardware. Among various SNNs training methods, the ANN-SNN conversion approach can achieve performance levels comparable to those of Artificial Neural Networks (ANNs). However, the spike firing rate of SNNs needs to be aligned with the activation of ANNs over longer time steps, which leads to a performance decline of SNNs at short time steps, limiting their applicability. In response to the challenge, this research proposes an innovative framework for bearing fault diagnosis. The framework first trains an ANNs model with a multi-scale convolutional attention mechanism (MCNN-AM) that has the capability of feature extraction and noise resistance. Subsequently, the ANN is trained and converted to an SNN using the quantization clip-floor-shift (QCFS) activation function. During the inference phase, an optimization strategy based on residual membrane potential (SRP) is introduced to effectively reduce the SNNs response latency while maintaining diagnostic accuracy. The proposed framework enhances the capability of SNNs for diagnosing faults in bearings and enables the deployment of SNNs on compact and mobile platforms.

Memristors can be used to simulate the magnetic induction effect of action potentials in biological neurons. The diversity and complexity of firing patterns generated by existing memristive neuron models remain limited, which hinders their ability to fully characterize the rich electrical activities observed in biological nervous systems. To address this issue, this article introduces a novel memristive multiring Rulkov neuron model (MMRNM) by incorporating a universal discrete memristor into the classical Rulkov neuron model, enabling the investigation of more complex neuronal firing patterns and firing multistability. Dynamical analysis indicates that the MMRNM exhibits abundant firing behaviors, such as periodic firing, chaotic firing, hyperchaotic firing, and firing pattern transitions. Meanwhile, the MMRNM can generate multiring firing patterns with an arbitrary number of rings. Through initial offset boosting control, two types of firing multistability, namely homogeneous and heterogeneous firing multistability, are observed. To verify the performance, an STM32-based hardware platform is designed to capture the firing behaviors experimentally. Finally, a novel encryption scheme is designed by leveraging the chaotic and hyperchaotic sequences of the MMRNM, and performance evaluations demonstrate that the scheme possesses robust security.

Event cameras report per-pixel brightness changes as a stream of asynchronous events, making them insensitive to the spatial distribution of feature points. Thus, traditional calibration methods, such as checkerboard and circle patterns, are inadequate for precise event camera calibration. Here, we propose an accurate calibration method based on blinking binary stripes to address this limitation. Two binary stripes, which are orthogonal, are displayed on a screen and blink at a specific frequency. The triggered data streams are converted into frames, which are analyzed using the Fourier technique. Thus, the feature points are located from the phase domain, which differs from conventional methods. The superior performance of the proposed method is verified in calibration experiments.