基于特征分布的建筑墙体裂缝图像识别方法

针对当前以人工智能为基础的墙体裂缝识别主要以图像识别为主,容易受到裂痕特征分布不均匀的影响,识别精度不高的问题,提出基于特征分布和高斯混合模型的建筑墙体裂缝图像识别方法.采用Harris角点检测算法对墙体图像进行角点求解处理,对建筑墙体图像进行预处理;通过选择掩模平滑法对墙体图像进行增强处理,将特征分布和高斯混合模型相结合,实现对建筑墙体裂缝图像的高精度识别.结果表明,该方法识别精度较高且识别时间短,预处理效果明显增强.

Image recognition method for building wall cracks based on feature distribution

Aiming at the problem that the current wall crack identification based on artificial intelligence is mainly composed of image recognition, which is easily affected by the uneven distribution of crack feature, and the identification accuracy is not high, a building wall crack image recognition method based on both the feature distribution and Gaussian mixture model was proposed. The corner solution processing for the wall image was carried out with the Harris corner detection algorithm, and the building wall image was preprocessed. Through selecting the mask smoothing method, the enhancement processing for the wall body image was performed. Through combining the feature distribution and Gaussian mixture model, the high-precision recognition of building wall crack image was realized. The results show that the proposed method has higher recognition precision and shorter identification period, and the pretreatment effect is obviously enhanced.