Back to Blog

AI Intelligent Box Empowers Sinosteel Tianyuan Equipment Factory to Upgrade Safety Protection

#人工智能#安全#大数据

Sinosteel Anhui Tianyuan Technology Co., Ltd., established on March 27, 2002, is a listed company controlled by China Sinosteel Corporation, a central state-owned enterprise. Its main products include permanent magnet ferrite devices, NdFeB devices, trimanganese tetraoxide, strontium ferrite pre-sintered materials, and various magnetic separators.

During Sinosteel Tianyuan's intelligent upgrade process, a series of algorithms for factory safety management and control were adopted and custom-developed.

图片

△Sinosteel Tianyuan AI Algorithm Platform

Efficient Custom Development of New Algorithms

Steel equipment factories have vast areas, numerous heavy machinery, and complex environments. Operating personnel can easily be obscured, creating blind spots for safety monitoring. If only fixed cameras are used, they can suffer from perspective distortion, leading to misjudgments when equipment is in motion. Therefore, multi-point monitoring perspectives are needed to supplement coverage. A typical scenario is: when overhead cranes are in operation, there is a risk of falling objects causing casualties. At this time, personnel are strictly prohibited from being within a designated range near the overhead crane.

The professional team, through business logic analysis and algorithm development evaluation, ultimately determined a solution: custom-developing an overhead crane hook recognition algorithm. By installing cameras on both sides of the overhead crane, multi-point monitoring checks whether personnel enter or exit the area below when the crane is moving. If the overhead crane (and thus the entire camera view) is moving and personnel appear within the ROI (Region of Interest), an alert is triggered in the backend. If the overhead crane (and the camera) is stationary and personnel appear within the ROI, no alert is triggered. This algorithm design, utilizing a moving overhead perspective, can effectively identify whether the scene (overhead crane) is in motion, reduce interference from obstructions and perspective angles, and ensure personnel are clearly identified.

图片

图片

△Overhead Crane Hook Recognition Algorithm

In addition, it also provides various algorithms such as open flame and smoke detection, smoking detection, fire lane obstruction recognition, and hard hat detection. These promptly identify safety risks, powerfully ensuring safe production on site.

图片

△Open Flame and Smoke Detection

图片

△Smoking Detection

图片

△Fire Lane Obstruction Recognition

图片

△Hard Hat Detection

Overcoming Challenges, High Customer Recognition

During the implementation of this project, due to the high mounting height of existing cameras in the aisle environment and the presence of glare, the images were not clear enough, affecting recognition accuracy. The team was stationed on-site with the client, collecting multiple rounds of field data for algorithm optimization, which significantly improved the accuracy of the on-site algorithms. Both the high-quality algorithm delivery and professional project implementation services received high praise from the client. We look forward to collaborating with Sinosteel Group to empower various regional sites nationwide and contribute to intelligent and safe modern steel manufacturing.