Back to Blog

Based on RK3588/Sophon BM1684 AI Box: Integrated Intelligent Video AI Analysis System Deployment Plan (Part 3) Hard Hat, Drowsiness/Absence from Post Detection, Electric Vehicles, Smoking Scenarios

#人工智能#音视频#计算机视觉
      1. Hard Hat and Reflective Vest Detection Algorithm
        1. Function Description

Hard hat and reflective vest detection involves pre-setting a monitoring area in a surveillance scene. If personnel within this area are not wearing a hard hat or reflective vest, an alarm is triggered promptly. The detected target's resolution in a 1080p image should be no less than 30*30 pixels.

        1. Recommended Scenarios

Scenario Requirements: Suitable for indoor and outdoor use, avoiding direct sunlight or backlight conditions; at night, only full-color images are supported. If the camera does not support full-color mode, please use white light for illumination.

Target Distance Requirements: Events beyond 3 meters can be detected from a standard 3-meter pole height.

Target Pixel Requirements: The detected target's resolution in a 1080P image should be greater than or equal to 30*30 pixels.

Recall Rate: 95%

      1. Electric Vehicle Detection Algorithm
        1. Function Description

The electric vehicle detection algorithm, based on AI visual analysis technology, automatically identifies whether an electric vehicle enters the video image. If an electric vehicle is detected, an alarm is triggered on screen.

        1. Recommended Scenarios

Scenario Requirements: Suitable for indoor and outdoor use, avoiding direct sunlight or backlight conditions; at night, only full-color images are supported. If the camera does not support full-color mode, please use white light for illumination.

Target Distance Requirements: Events beyond 3 meters can be detected from a standard 3-meter pole height.

Target Pixel Requirements: The detected target's resolution in a 1080P image should be greater than or equal to 30*30 pixels, with electric vehicle occlusion not exceeding 30%.

Supports detection of two-wheeled, three-wheeled, four-wheeled, and other electric vehicles.

Recall Rate: 90%

      1. Person Fall Detection Algorithm
        1. Function Description

Person fall detection involves pre-setting a monitoring area in a surveillance scene. When a person falling is detected within this area, an alarm is triggered. The detected target's resolution in a 1080p image should be no less than 30*30 pixels.

        1. Recommended Scenarios

Scenario Requirements: Suitable for indoor and outdoor use, avoiding direct sunlight or backlight conditions; at night, only full-color images are supported. If the camera does not support full-color mode, please use white light for illumination.

Target Distance Requirements: Events beyond 3 meters can be detected from a standard 3-meter pole height.

Target Pixel Requirements: The detected target's resolution in a 1080P image should be greater than or equal to 30*30 pixels.

Recall Rate: 90%

      1. Person Smoking Detection Algorithm
        1. Function Description

Person smoking detection involves pre-setting a monitoring area in a surveillance scene. When a person smoking is detected within this area, an alarm is triggered. The detected target's resolution in a 1080p image should be no less than 30*30 pixels.

        1. Recommended Scenarios

Scenario Requirements: Suitable for indoor and outdoor use, avoiding direct sunlight or backlight conditions; at night, only full-color images are supported. If the camera does not support full-color mode, please use white light for illumination.

Target Distance Requirements: Events beyond 3 meters can be detected from a standard 3-meter pole height.

Target Pixel Requirements: The detected target's resolution in a 1080P image should be greater than or equal to 30*30 pixels.

Recall Rate: 90%

      1. Pedestrian Structured Detection (Face Recognition) Algorithm
        1. Function Description

Pedestrian structured detection (face recognition) algorithm involves pre-setting a monitoring area in a surveillance scene. When personnel not in the whitelist database are detected within this area, an alarm is triggered. The detected target's resolution in a 1080p image should be no less than 30*30 pixels.

        1. Recommended Scenarios

Scenario Requirements: Suitable for indoor and outdoor use, avoiding direct sunlight or backlight conditions; at night, only full-color images are supported. If the camera does not support full-color mode, please use white light for illumination.

Target Distance Requirements: Events beyond 3 meters can be detected from a standard 3-meter pole height.

Target Pixel Requirements: The detected target's resolution in a 1080P image should be greater than or equal to 30*30 pixels.

      1. Personnel Absence from Post/Drowsiness Detection Algorithm
        1. Function Description

Personnel absence from post/drowsiness detection involves pre-setting a monitoring area in a surveillance scene. When personnel are detected absent from their post or drowsy within this area, an alarm is triggered. The detected target's resolution in a 1080p image should be no less than 30*30 pixels. The time threshold for absence from post/drowsiness can be customized.

        1. Recommended Scenarios

Scenario Requirements: Suitable for indoor and outdoor use, avoiding direct sunlight or backlight conditions; at night, only full-color images are supported. If the camera does not support full-color mode, please use white light for illumination.

Target Distance Requirements: Events beyond 3 meters can be detected from a standard 3-meter pole height.

Target Pixel Requirements: The detected target's resolution in a 1080P image should be greater than or equal to 30*30 pixels.

Recall Rate: 90%