Parametric Information Extraction and Data Cleaning Method in Construction Design Drawing

The construction design drawing contains a lot of effective parametric information of ground features. To improve the utilization rate of parametric information in construction design drawings, a method of parametric information extraction and data cleaning in construction design drawings is proposed. By using Contourlet transformation, the multi-angle enhancement of spatial domain and transformation domain of construction design drawings is realized, and the overall clarity of construction design drawings is optimized. The construction design drawing is mapped to CIE LAB color space by SLIC superpixel segmentation method. The characteristics of the drawing are fused to obtain the probability feature map, construct the relative dominance matrix of the target, share the weights in the convolution process, and update the discriminant parameters of machine learning in real time. The model is trained to balance the machine learning process and the discrimination process, and the convolution obtained feature maps are connected as the discrimination basis to judge the authenticity of the drawings, and thus the parametric information of the drawings is extracted. Based on the density clustering algorithm, the anomaly of parametric information extraction is identified, the abnormal part of parametric information extraction is solved and eliminated, and the abnormal cleaning of parametric information extraction is completed. The experimental results show that: The proposed method can effectively classify the subject of parametric information in drawings, significantly reduce the misrepresentation of parametric information extraction, and improve the readability of drawings.