CN-122024151-A - Method, device, medium and equipment for detecting radius of rotation of personnel entering machine

CN122024151ACN 122024151 ACN122024151 ACN 122024151ACN-122024151-A

Abstract

The invention discloses a method, a device, a medium and equipment for detecting the radius of a mechanical rotation of a person entering. The method comprises the steps of obtaining a monitoring video of mechanical operation, screening out images containing mechanical operation scenes from the monitoring video, manually marking the images containing the mechanical operation scenes, obtaining marking data, dividing the marking data, constructing a target detection model, obtaining a trained target detection model, obtaining a picture to be detected, inputting the picture to be detected into the trained target detection model, inputting the picture to be detected into a target detection network after image decoding and image preprocessing operation, obtaining coordinates, types and confidence of a person and a rotating machine based on detection results, calculating the distance between the person and the center point of the rotating machine and the length of a mechanical arm of the rotating machine based on detection results of a single image, and judging whether the person is in the rotating radius of the rotating machine. The method can judge whether the personnel are in the rotating radius of the rotating machine in real time, and improves the quality of the image.

Inventors

ZHANG HUI
ZHAO JINLING
ZHANG YAO
YANG KAIZAN
WANG JUN
HAO HUAJIE
PU HONGBIN
WANG TING
DONG WENTAO

Assignees

中国石油天然气股份有限公司

Dates

Publication Date: 20260512
Application Date: 20241111

Claims (10)

1. A method for detecting the radius of a person entering a machine, the method comprising the steps of: S10, acquiring a monitoring video V i of mechanical operation based on a monitoring camera connected to an operation site; S20, screening out an image I i containing a mechanical operation scene from the monitoring video V i ; S30, manually labeling the image I i containing the mechanical operation scene, obtaining labeling data I ann , and dividing the labeling data I ann ; S40, constructing a target detection model, and training the labeling data I ann to obtain a trained target detection model; s50, obtaining a picture I test to be detected; S60, inputting the picture I test to be detected into a trained target detection model, inputting the picture I test to be detected into a target detection network after image decoding and image preprocessing operations, and acquiring the coordinates, the category and the confidence of personnel and rotary machinery based on a detection result; And S70, calculating the distance D cent between a person and the center point of the rotary machine and the length W of the mechanical arm of the rotary machine based on the detection result of the single image, and judging whether the person is in the rotation radius of the rotary machine, wherein if D cent < = 1.5 x W, the person is in the rotation radius of the rotary machine.
2. The method for detecting the radius of rotation of a person entering a machine according to claim 1, wherein step S30 includes: s310, manually labeling the image I i containing the mechanical operation scene to obtain labeling data I ann ; S320, dividing the labeling data I ann into a training sample I train and a verification sample I valid .
3. The method for detecting the radius of rotation of a person entering a machine according to claim 2, wherein the step S40 includes S410 of constructing a target detection model based on deep learning; S420, performing a Mosaic data enhancement training on the training sample I train and the verification sample I valid ; And S430, training the training sample I train and the verification sample I valid after the Mosaic data is enhanced, and obtaining a trained target detection model.
4. The method for detecting the radius of rotation of a person entering a machine according to claim 1, wherein step S50 includes: S510, extracting frames of the video stream of the monitoring camera connected to the operation site based on FFmpeg, and extracting 1 frame as a detection picture I test every N frames.
5. The method for detecting the radius of rotation of a machine by a person according to claim 1, further comprising, prior to step S60: s520, inputting the images to be detected belonging to the same group to be detected into a trained target detection model, and extracting coordinates of the target.
6. The method for detecting the radius of rotation of a person entering a machine according to claim 1, wherein step S60 comprises: s610, inputting the picture I test to be detected into a trained target detection model, and inputting the picture I test to be detected into a target detection network after image decoding and image preprocessing operations; S620, filtering redundant repeated frames by suppressing the detection result through NMS non-maximum value; s630, acquiring the coordinates of the person and the rotary machine based on the detection result Category(s) Confidence level Where n is the number of people or rotating machinery detected in the image.
7. The method for detecting the radius of rotation of a person entering a machine according to claim 1, wherein step S70 comprises: s710, judging whether personnel and rotating machinery exist in the detection result of the single image; S720, if the area ratio of the personnel to the rotary machine is calculated; S730, if the area of the person is 10% -30% of the area of the rotary machine, determining that the person and the rotary machine are around the rotary machine; S740, calculating the distance D cent between the center point of the person and the rotating machine and the length W of the mechanical arm of the rotating machine; S750, if D cent < = 1.5 x w, the person is within the radius of rotation of the rotating machine.
8. The detection device of personnel entering mechanical radius of rotation, characterized by comprising: The video acquisition module is used for acquiring a monitoring video V i of mechanical operation based on a monitoring camera connected to an operation site; the image screening module is used for screening an image I i containing a mechanical operation scene from the monitoring video V i ; The manual labeling module is used for manually labeling the image I i containing the mechanical operation scene, obtaining labeling data I ann and dividing the labeling data I ann ; The model construction and training module is used for constructing a target detection model, training the labeling data I ann and acquiring a trained target detection model; The picture acquisition module is used for acquiring a picture I test to be detected; The detection module is used for inputting the picture I test to be detected into a trained target detection model, inputting the picture I test to be detected into a target detection network after image decoding and image preprocessing operations, and acquiring the coordinates, the category and the confidence of personnel and rotary machinery based on detection results; And the calculation module is used for calculating the distance D cent between the center point of the person and the rotating machine and the length W of the mechanical arm of the rotating machine based on the detection result of the single image, judging whether the person is in the rotating radius of the rotating machine, and if D cent < = 1.5 x W, the person is in the rotating radius of the rotating machine.
9. A computer readable storage medium comprising instructions which, when run on a computer, cause the computer to perform the method of detecting the radius of a person entering a machine as claimed in any one of claims 1 to 7.
10. A computing device, the computing device comprising: at least one processor, memory, and input output unit; wherein the memory is used for storing a computer program, and the processor is used for calling the computer program stored in the memory to execute the method for detecting the radius of the mechanical rotation entered by the person according to any one of claims 1-7.

Description

Method, device, medium and equipment for detecting radius of rotation of personnel entering machine Technical Field The present invention relates to the field of mechanical detection technologies, and in particular, to a method, an apparatus, a medium, and a device for detecting a radius of rotation of a person entering a machine. Background The equipment is usually moved soil in the oilfield site, hoisting operation, mechanical rotating equipment such as an excavator, a suspension arm and the like, personnel intrude into the rotating radius of the equipment during operation, so that the risk of mechanical injury is caused, and a key safety measure is to ensure that the personnel can timely detect the equipment when the personnel is in the rotating radius of the machinery, and take necessary safety measures to prevent accidents. Conventional approaches typically rely on hardware devices such as mechanical sensors, photosensors, or lidars, but they may be limited by environmental noise, equipment failure, or expensive maintenance costs. Effective identification is needed by a video monitoring means to avoid similar risks. With the rapid development of video detection technology, intelligent monitoring methods based on image and video data are becoming more attractive. By using deep learning models such as a deep Convolutional Neural Network (CNN), whether a worker enters a mechanical rotation radius can be detected in a real-time video stream, so that the safety of a workplace is improved. This approach not only reduces reliance on additional sensors, but also provides greater accuracy and real-time. However, introducing video detection into the field of mechanical radius of rotation detection faces challenges such as poor image quality, problems with detection errors caused by illumination variations, target size and pose variations. Disclosure of Invention The invention mainly aims to provide a method, a device and equipment for detecting the radius of rotation of a person entering a machine, and aims to solve the technical problems that in the prior art, the image quality is poor, and detection errors are caused by illumination changes, target size and posture changes. The detection method for the mechanical turning radius of the personnel comprises the steps of S10, obtaining a monitoring video V i of mechanical operation based on a monitoring camera connected to an operation site, S20, screening an image I i containing a mechanical operation scene from the monitoring video V i, S30, manually marking the image I i containing the mechanical operation scene, obtaining marking data I ann, dividing the marking data I ann, S40, constructing a target detection model, training the marking data I ann, obtaining a trained target detection model, S50, obtaining a picture I test to be detected, S60, inputting the picture I test to be detected into a target detection network after image decoding and image preprocessing operation, obtaining the coordinates, types and confidence of the personnel and the turning machine based on the detection result of the single image, S70, calculating the center point distance D cent between the personnel and the turning machine and the length W of a mechanical arm of the turning machine, and judging whether the personnel is within the turning radius of the machine or not, and if the personnel is within the turning radius of the machine is within cent =1.W=5. In some embodiments, the step S30 includes S310 of manually labeling the image I i containing the mechanical operation scene to obtain labeling data I ann, and S320 of dividing the labeling data I ann into a training sample I train and a verification sample I valid. In some embodiments, step S40 includes constructing a target detection model based on deep learning, step S420, performing a Mosaic data enhancement training on the training sample I train and the verification sample I valid, and step S430, performing a training on the training sample I train and the verification sample I valid after the Mosaic data enhancement, and obtaining a trained target detection model. In some embodiments, the step S50 includes S510, extracting frames of the video stream of the monitoring camera of the access job site based on FFmpeg, and extracting 1 frame as a detection picture I test every N frames. In some embodiments, before step S60, the method further comprises the step of S520, inputting the images to be tested belonging to the same group to be tested into a trained target detection model, and extracting coordinates of the target. In some embodiments, the step S60 includes inputting the picture I test to be detected into a trained target detection model, inputting the picture I test to a target detection network after image decoding and image preprocessing operations, filtering the detection result through NMS non-maximum suppression to redundant repeated frames, S620, and acquiring the coordinates of personnel and rotating machinery b