Rebar Binding Crosspoints Detection Method Based on Deep Learning and Feature Projection Curve

Focused on the problems of cumbersome operation, low efficiency, and high cost in the traditional manual rebar binding process, a rebar binding crosspoints detection method based on deep learning and feature projection curve is proposed. Firstly, a dataset of 3 300 rebar binding crosspoints is built by original data collection and data enhancement, and a MobileNetV3-SSD rebar binding crosspoints recognition model is built based on TensorFlow deep learning framework to achieve automatic recognition and classification of rebar bound or unbound crosspoints, which is used to extract the prediction box of unbound crosspoints. Secondly, an image post-processing procedure is developed for the extracted prediction box region of unbound crosspoints, and a crosspoints localization method combining control factor α and feature projection curve is introduced to achieve the localization of unbound crosspoints. Finally, the algorithm is evaluated by performance test experiments to determine the optimal threshold of the recognition model and localization method. Experimental results show that the macro precision and macro recall of the proposed model are 95.40% and 96.70%, respectively, and the localization accuracy of the proposed method is 90.45% with an average relative error of 6.15%. The proposed method can achieve fast, non-contact and intelligent detection of rebar binding crosspoints, which has good robustness and application value.