Abstract: The primary challenge encountered by unmanned technology during the unloading phase in open-pit mines is safety hazards, particularly concerning the stability and normative detection of engineering structures at the edges of unloading area. To tackle this issue, a point cloud model analysis algorithm, driven by parallel perception theory and named AC-VIT, is proposed for the real-time and stable detection of the stability and normativity of engineering structures at the edges of open-pit coal mine unloading areas. Initially, three-dimensional point cloud data are captured using unmanned dump trucks equipped with rearward LiDAR scanning. These data are then processed through grid averaging methods, statistical filtering, and mapping to discrete grid models. Preliminary terrain marking is conducted via height field gradient feature extraction, in conjunction with the improved AC-VIT neural network for normative recognition and classification. The AC-VIT model, leveraging parallel computation solely based on a self-attention mechanism and multi-level attention mechanisms, effectively captures long-distance dependencies. Furthermore, a parallel simulation environment for the unloading area is established based on the actual production environment of the Haerwusu open-pit coal mine in Inner Mongolia, within a simulated artificial scene environment, to gather a vast array of diverse artificial scene data. Utilizing this data, in conjunction with actual scene data, the algorithm undergoes a parallel execution to design and perform parallel perception computing experiments, facilitating the effective training of the detection algorithm and scientific evaluation. Experimental outcomes demonstrate that the AC-VIT algorithm, underpinned by parallel perception theory, attains an accuracy rate of 98%, surpassing the accuracy and efficiency of traditional neural network models. The successful deployment of the AC-VIT algorithm not only elevates the intelligence level in open-pit mine unloading operations, but also furnishes robust technical support for the safety detection of other analogous engineering structures. The algorithm introduced herein presents a more efficient, safe, and intelligent approach for the detection of engineering structures at unloading area edges, bearing significant relevance for achieving high-performance, high-reliability, and high-automation in open-pit mine operations.