Rock blasting evaluation - image recognition method based on deep learning

The image analysis of rock blasting fragmentation and half-hole rates serves as a pivotal basis for evaluating the quality of geotechnical blasting and pre-splitting blasting. Using image recognition technology to conduct image analysis of rock blasting fragmentation and half-hole rate can direct the control of rock blasting parameters, which is of great significance for improving rock breaking quality and shovel loading efficiency. In order to efficiently evaluate the quality of rock blasting in mines, this paper developed a blasting effect image analysis and calculation model and recognition algorithm based on the established machine learning database, and carried out recognition and analysis work on the half-hole rate and rock blasting fragmentation of pre-splitting blasting. Research has shown that the half-hole rate derived from manual statistics and image recognition is 68.16% and 67.15%, respectively, with an average error of 1.49%. This indicates that the image recognition method has good reliability in evaluating the quality of pre-splitting blasting, and can provide feedback on the half-hole rate obtained through image recognition to guide the adjustment of blasting parameters and improve the quality of on-site blasting. Based on the recognition outcomes of mineral rock block size images, the "S" shaped cumulative distribution pattern of block size was discovered, and the Pearson curve function was fitted. And a segmented function R-R block size distribution correction model was proposed, which can better characterize the characteristics of blasting block size distribution and serves as a guide for accurate prediction of block size distribution. Overall, image recognition technology can significantly improve the efficiency and accuracy of blasting evaluation, solve many drawbacks of traditional manual analysis methods, and holds great application prospects in blasting quality evaluation.