A Dual-Channel and Multi-Sensor Fusion Framework for Coal Mine Image Dehazing

Area of Science:

• Computer Vision
• Image Processing
• Environmental Sensing

Background:

• Coal mine imagery suffers degradation from dust, haze, and poor lighting.
• Existing dehazing methods are limited by overlooking color, using single algorithms, or requiring extensive data and hardware.
• Deep learning approaches present challenges in efficiency and hardware demands for real-world applications.

Purpose of the Study:

• To propose an efficient and adaptable image dehazing framework for challenging environments like coal mines.
• To enhance image quality by accurately restoring features and details in degraded visuals.
• To develop a method that leverages both visual data and environmental sensor information.

Main Methods:

• Integration of bright and dark channel information to derive contrast features reflecting dust concentration.
• Utilization of dust sensor data to establish adaptive scaling coefficients and dust compensation terms.
• Employing a weighted fusion of channels and a weighted guided filter with dust compensation for refined ambient light and transmission estimation.
• Global color mean used for distinguishing image types to apply appropriate dehazing strategies.

Main Results:

• The proposed method effectively removes haze while preserving image features and details.
• Experimental results show superior performance compared to state-of-the-art methods on coal mine and standard datasets.
• Demonstrated enhanced stability, adaptability, and computational efficiency in dehazing.

Conclusions:

• The framework offers a robust solution for image dehazing in dusty and low-light conditions.
• The integration of sensor data provides state verification and information complementarity for improved perception.
• The method presents a significant advancement in efficient and effective image restoration for industrial environments.