Improved PCNN Rock Muck Image Segmentation Based on Zebra Optimization Algorithm and Morphology

In order to solve the problem of uneven segmentation caused by fixed threshold, a blast pile image segmentation model based on an improved pulse coupled neural network is proposed, with distortion and adhesion issues taken into consideration. This method involves setting up an image acquisition device on site based on the principle of distortion and selecting a sphere as a reference, while simultaneously segmenting rock chips as foreground pixels. The segmentation process consists of two stages. Fisrt,the zebra optimization algorithm is employed in combination with Kapur entropy as the fitness function to optimize the structural parameters of the pulse coupled neural network. Secondly, morphological operations are introduced to enhance segmentation performance and improve the precision of the resulting rock chip morphology. The research results demonstrate that the proposed model further improves segmentation precision compared to traditional models, and the model is feasible. For complex shape scenes, the proposed model still has advantages in accuracy, although its segmentation results tend to retain more reliable areas.