SkyVision Platform Achieves Precise Detection of Safety Helmet and Reflective Vest Compliance

With the development of the video surveillance field under the new AI Agentic Infra system, more precise behavior recognition has become a key requirement in the emergency security industry. WeLinkirt's SkyVision platform plays an important role in the compliance detection of safety helmets and reflective vests.

User Scenario: A large-scale construction site is a key area for emergency security. With a large number of construction workers and a complex on - site environment, it is necessary to detect the compliance of safety helmets and reflective vests for workers entering the construction area during daily construction processes. The detection objects are safety helmets and reflective vests, aiming to ensure that every worker entering the construction site wears them correctly to safeguard their lives and meet relevant safety regulations.

Pain Points: In the current video surveillance field under the new AI Agentic Infra system, there are higher requirements for precise behavior recognition. However, the existing detection methods at the construction site face many difficulties. The traditional manual inspection method is inefficient and costly in terms of manpower, with a high risk of missed detection. On average, only about 30 people can be inspected per hour, and the missed detection rate is as high as 10%. Some video surveillance systems based on simple algorithms have a high false-alarm rate of 15%. Frequent false alarms not only increase the workload of staff but also affect the timeliness and accuracy of real alarm information. Moreover, as the construction content at the site changes, the detection rules also need to be adjusted accordingly. However, the existing systems have difficulties in model change, taking about 3 hours for each change, which seriously affects the detection efficiency and overall emergency security work.

Technical Principle

The SkyVision zero-code video surveillance AI platform uses advanced deep-learning algorithms combined with the semantic understanding ability of the DaoAI World model. The deep-learning algorithm can automatically learn the feature patterns of safety helmets and reflective vests from a large amount of video image data. Specifically, it uses the Convolutional Neural Network (CNN) to extract features from images and quantitatively analyze features such as the color, shape, and texture of safety helmets and reflective vests in the images.

• Color features: Safety helmets and reflective vests usually have specific colors, such as yellow safety helmets and orange reflective vests. The algorithm can accurately identify these color features and match them with the predefined color ranges.
• Shape features: The shapes of safety helmets and reflective vests have certain regularities. The algorithm can identify the circular contour of the safety helmet and the approximate shape of the reflective vest, and judge the correct wearing status through shape analysis.
• Texture features: The reflective stripes on the surface of the reflective vest have unique textures. The algorithm can confirm the presence and wearing condition of the reflective vest through texture recognition.
• Semantic understanding: The DaoAI World model provides semantic support for the entire detection process, enabling it to understand the meanings of safety helmets and reflective vests in different scenarios, further improving the accuracy and generalization ability of detection.

WeLinkirt Solution and Product Introduction

Centered around the SkyVision zero-code video surveillance AI platform, it can train its own model on - site within hours. On - site staff do not need to have professional programming knowledge. By providing a certain number of positive sample images (1-20 pieces), the platform can complete model training in a short time. In the compliance detection of safety helmets and reflective vests, the platform can quickly learn the features of safety helmets and reflective vests to achieve precise behavior recognition.

The zero-code feature of the SkyVision platform greatly reduces the usage threshold, allowing on - site personnel to quickly get started.

The platform also has a behavior/event recognition function, which can real-time determine whether workers are wearing safety helmets and reflective vests correctly. Once a violation is detected, the edge box will immediately issue a real-time alarm to notify on - site management personnel. At the same time, the platform supports 100% local data storage without leaving the site, ensuring data security and privacy. In addition, the semantic understanding ability of the DaoAI World model enables the platform to maintain a high recognition accuracy under different scenarios and lighting conditions. The semantic false-alarm filtering function in the supporting DaoAI AI AOI software system can further reduce false-alarm situations.

Quantitative Results: After using the SkyVision platform, the detection rate of safety helmets and reflective vests reached 98%, and the missed detection rate was reduced to <2%. The false-alarm rate was reduced by -80%, greatly reducing the ineffective workload of staff. At the same time, the model-changing time was shortened from 3 hours to 5 minutes, greatly improving the detection efficiency and meeting the requirements of rapid response and precise detection for the emergency security work at the construction site.