基于对数变换和多尺度特征融合的螺栓盒处图像增强

Image Enhancement of Bolt Box Areas Based on Logarithmic Transformation and Multi-scale Feature Fusion

摘要:

针对建筑机器人视觉采集装配式建筑单元的螺栓连接盒图像时存在的整体昏暗、螺栓杆处明暗不均和纹理不清晰等问题,提出了一种基于对数变换和多尺度特征融合的图像增强算法。针对螺栓连接盒整体昏暗环境,采用对数变换方法,结合Gamma校正对图像动态扩展优化。通过动态调整参数,提升图像在不同光照条件下的表现,恢复潜在的细节信息。引入多尺度特征融合方法,将图像在不同尺度下的信息通过动态权重分配进行融合处理,从而显著提升图像的纹理清晰度。针对螺栓杆区域,采用对比度受限自适应直方图均衡化(CLAHE)进行强化处理。对图像进行边缘锐化和噪声抑制,确保增强的图像平滑且细节清晰。与当前主流图像增强算法进行对比,本文算法在亮度、清晰度和细节纹理上表现更优,结构相似性、峰值信噪比平均值分别提高了5.89%和4.73%,均方误差降低了12.55%。

Abstract:

To address the issues encountered in robotic visual acquisition of bolt connection box images in prefabricated building units (such as overall low brightness, uneven illumination on bolt rods, and unclear texture details), this paper proposes an image enhancement algorithm based on logarithmic transformation and multi-scale feature fusion. For the problem of overall dim lighting in bolt connection box images, a logarithmic transformation is adopted and combined with Gamma correction to optimize dynamic range expansion. By adaptively adjusting parameters, the proposed method improves image performance under varying illumination conditions and effectively recovers latent detail information. Furthermore, a multi-scale feature fusion strategy is introduced, in which image information at different scales is integrated through dynamic weight allocation, thereby significantly enhancing texture clarity. Targeted enhancement is applied to the bolt rod region using the Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve local contrast. In addition, edge sharpening and noise suppression techniques are employed to ensure that the enhanced images remain smooth while preserving fine details. Comparative experiments with mainstream image enhancement algorithms demonstrate that the proposed method achieves superior performance in terms of brightness, clarity, and texture detail. Specifically, the average Structural Similarity Index Measure (SSIM) and Peak Signal-to-Noise Ratio (PSNR) are improved by 5.89% and 4.73%, respectively, while the Mean Squared Error (MSE) is reduced by 12.55%.