Intelligent Autofocus Method for Event Cameras with Time-Frequency Coupled Optimization Metrics

Event cameras, characterized by high dynamic range and microsecond-level temporal resolution, have emerged as effective tools for high-speed industrial monitoring. However, their practical deployment is hindered by challenges in manual focusing, including susceptibility to human perception errors, inefficiency in dynamic environments, and systematic inaccuracies affecting downstream tasks such as vibration analysis or fault diagnosis. To address these limitations, this paper proposes an intelligent autofocus (AF) method based on time-frequency coupled optimization metrics. This method first automatically identifies a coarse focal zone by exploiting the correlation between Event Polarity Ratio (EPR) profiles and defocus degrees, specifically through quantifying the integral area over positive-derivative intervals of the EPR curve. This step reduces the search space by 93% compared to full-scan strategies. Within this zone, precise focal positioning is achieved by coupling two key optimization metrics, i.e., the temporal sparsity to minimize event rate, and the spectral sharpness to maximize high-frequency energy concentration in reconstructed event frames. The proposed time-frequency coupling mechanism synergistically leverages the complementary strengths of time and frequency domain analyses, significantly enhancing focusing accuracy and robustness against motion noise. Experimental validation under laboratory conditions using the radial calibration pattern targets demonstrates that the proposed method can intelligently achieve coarse localization of the focal region followed by precise focusing. Compared to single-metric AF approaches, it exhibits superior robustness and higher reliability, particularly in the presence of noise and dynamic disturbances.