CN-117475343-B - Intelligent monitoring method for cross operation of refining device area

CN117475343BCN 117475343 BCN117475343 BCN 117475343BCN-117475343-B

Abstract

The invention provides an intelligent monitoring method for cross operation of a refining device region, which comprises the following steps of fixedly erecting binocular cameras in the horizontal direction and the vertical direction of a cross operation region respectively, shooting pictures of the cross operation region, acquiring RTSP video streams of the two binocular cameras in real time for intelligent analysis, respectively carrying out camera calibration, binocular correction and stereo matching on the pictures shot by the two binocular cameras, then carrying out intelligent analysis and judgment on the pictures shot by the two binocular cameras, and carrying out personnel event alarm on the personnel station of the cross operation region of the lower layer if the pictures shot by the two binocular cameras are judged to be personnel stations of the cross operation region of the suspected lower layer, and respectively outputting two shooting images by the two binocular cameras to control alarm equipment for acousto-optic and voice reminding. The invention alarms the abnormal situation, informs the manager and the operator to timely treat the illegal action, and prevents casualties such as object striking, mechanical injury and the like in the process of cross operation.

Inventors

LI QIANDENG
LI JIANXIANG
SUN DEQING
CHENG SIJIA
LIU TING
ZHANG XIAOHUA
WU RUIQING
LIU YANG

Assignees

中国石油化工股份有限公司
中石化安全工程研究院有限公司
中石化管理体系认证(青岛)有限公司

Dates

Publication Date: 20260508
Application Date: 20220719

Claims (9)

1. The intelligent monitoring method for the cross operation of the refining device area is characterized by comprising the following steps of: S1, fixedly erecting binocular cameras in the horizontal direction and the vertical direction of a cross operation area respectively, shooting pictures of the cross operation area, wherein the cross operation area comprises an upper layer of construction platform and a lower layer of construction platform, the erection height of the binocular cameras is the height of the lower layer of construction platform of the cross operation, and the erection distance is based on the condition that the cross operation environment can be covered, and real-time acquisition of RTSP video streams of the binocular cameras is performed for intelligent analysis; S2, respectively carrying out camera calibration, binocular correction and stereo matching on pictures shot by the two binocular cameras, and then judging the pictures shot by the two binocular cameras; S3, if the two binocular cameras in two directions are judged to be suspicious of people standing in the lower-layer cross operation area, alarming the personnel standing in the lower-layer cross operation area, respectively outputting two shooting images by the two binocular cameras, and controlling alarm equipment to carry out acousto-optic and voice reminding; in the step S2, the pictures shot by the two binocular cameras are respectively and simultaneously determined, and the method includes the following sub-steps: s21, carrying out personnel target detection on the shot picture, and calculating the coordinates of the central point below each rectangular frame according to the rectangular coordinates of the personnel, namely the coordinates corresponding to the feet of the personnel, wherein the coordinates represent the station position of each personnel; s22, recording the image height as H, comparing the ordinate value of each personnel position with H/2, recording as an upper-layer operator if the ordinate value is greater than H/2, and recording as a lower-layer operator if the ordinate value is less than H/2; s23, calculating the horizontal coordinate difference value of each upper-layer operator and each lower-layer operator in sequence, judging whether the horizontal coordinate difference value is smaller than a threshold value d1, and if the horizontal coordinate difference value is smaller than the threshold value, further judging the depth information distance relation; s24, according to the parallax map, gray values corresponding to station coordinates of an upper-layer operator and a lower-layer operator are obtained, a gray difference value is calculated, whether the gray difference value is smaller than a threshold value d2 is judged, and if the gray difference value is smaller than the threshold value d2, the binocular camera in the direction judges that a person standing in a suspected lower-layer cross operation area; And S25, if the binocular cameras in the two directions are judged to be suspected of standing persons in the lower-layer cross operation area, judging to be a standing person event in the lower-layer cross operation area.
2. The intelligent monitoring method for the cross operation of the refining device area according to claim 1, wherein in the step S2, two binocular cameras are respectively calibrated by using OpenCV, and the specific implementation steps are as follows: a. the binocular camera is horizontally placed and fixed on the same reference plane; b. a checkerboard calibration plate is placed in front of the binocular camera, so that the calibration plate is completely displayed in a shot picture, and the calibration plate is kept flat in the calibration process; c. under the condition that the brightness of the calibration plates is enough and uniform, a plurality of calibration plates are placed in the actual working distance range at equal intervals, and 20-30 Zhang Biaoding images are acquired; d. And (3) automatically importing a calibration image by using Matlab to perform automatic calibration, reading a calibration result into an OpenCV, and performing subsequent image calibration and matching.
3. The intelligent monitoring method for the cross operation of the refining device area according to claim 2, wherein in the step b, the calibration plate occupies 1/4 to 1/2 of the whole shooting picture.
4. The intelligent monitoring method for the cross operation of the refining device area according to claim 2, wherein the internal parameters of the two binocular cameras are obtained after the calibration is completed, and the internal parameters comprise fx, fy, cx, cy, distortion coefficients [ k1, k2, p1, p2, k3] and the relative positions between the left camera and the right camera.
5. The intelligent monitoring method for the cross operation of the refining device area according to claim 4, wherein in the step S2, the specific implementation process of binocular correction and stereo matching is that after each parameter of a camera is obtained through double-target determination, a correction rotation matrix R, a projection matrix P and a re-projection matrix Q are obtained through stereoRectify in OpenCV, calibration mapping parameters are obtained through initUndistortRectifyMap functions, then remap is used for correcting input left and right images, the corrected left and right image matching points are on the same line, and a parallax image is obtained after parallax image binocular matching is calculated through SGBM algorithm in OpenCV.
6. The intelligent monitoring method for the cross-country operations of the refining device according to claim 1, wherein in the step S21, a YOLOv personnel target detection frame is adopted for personnel detection of the shot picture, the characteristic information is firstly extracted through the backbone network Darknet-53, and then classification and positioning are performed through the detection network.
7. The intelligent monitoring method for the cross operation of the refining device area according to claim 6, wherein a Darknet-53 network structure adds residual modules in the network, and each residual module is composed of two convolution layers and a shortcut link.
8. The intelligent monitoring method for the cross operation of the refining device area according to claim 6, wherein the YOLOv adopts an up-sampling and fusion method, prediction target frames are respectively carried out on three layers of characteristic diagrams of 13, 26 and 52, and 3 prediction branches adopt a full convolution structure.
9. The intelligent monitoring method for the cross operation of the refining device area according to claim 6, wherein aiming at YOLOv detection frames, based on original classification results and positioning results, output results of target key point thermodynamic diagrams are added, so that tasks of target detection and human body key point detection are integrated, obtained human body key point position information comprises a head part, a chest part, a waist part and a foot part, human body height information is further obtained according to the key point position information, and setting of self-adaptive correction discrimination thresholds d, d1 and d2 is performed.

Description

Intelligent monitoring method for cross operation of refining device area Technical Field The invention relates to the technical field of cross operation monitoring, in particular to an intelligent monitoring method for cross operation of a refining device area. Background The cross operation is to perform different operations on the same working surface or to perform different or the same operations on different working surfaces in the same three-dimensional space. The construction site is frequently provided with the operation of vertical three-dimensional intersection and the operation of high places which are simultaneously carried out in a space through state, which belong to the category of the cross operation, and safety accidents such as falling injury, high place falling, mechanical striking, chemical burning, fire disaster and the like are very easy to occur, so that the safety hazards are more, and the management difficulty is high. The existing safety protection measures for the cross operation are mainly realized by means of establishing a contact communication mechanism to reasonably adjust operation activities and avoiding cross as much as possible, strengthening personnel training, informing risks, compiling emergency plans and the like, and the technical means are lacking. CN107247456a discloses a safety anti-collision system for multi-equipment cross operation area, which adopts various sensors installed on different equipment to continuously read data, can realize distance judgment among various equipment through the distance relation among computing equipment and alarm through an audible and visual alarm to remind an operator to carefully operate, thus improving detection efficiency and safety, CN101570304 discloses a display device for cross operation of a tower crane, which is characterized in that an alarm bell and a camera are installed on a crane hook roadster of the tower crane, when the crane works, the electric bell and the camera can move along with the roadster, and an image and bell are utilized to avoid and remind constructors of avoiding damage of suspended objects, and CN113095525A provides an intelligent control cross operation system for power grid maintenance, which comprises a data storage unit, a detection unit and a high-level application unit, and helps the maintainers to timely and accurately know maintenance operation information, prevent maintenance operation and high-voltage test cross operation, and effectively reduce operation risks. The existing cross operation safety management and control technology mainly focuses on adopting auxiliary sensor detection equipment or prejudging tracks to perform early warning in advance, is high in hardware cost and complex in field erection construction, and lacks applicability in tank areas, device areas and the like in petrochemical industry. Therefore, it is necessary to provide a method for identifying and detecting the stereo cross operation, which is used for reminding a manager to strengthen early warning and monitoring and avoiding personal injury in the operation process. Disclosure of Invention Aiming at the problems in the prior art, the invention provides an intelligent monitoring method for the cross operation of the refining device area, which has reasonable design, overcomes the defects in the prior art and has good effect. In order to achieve the above purpose, the present invention adopts the following technical scheme: an intelligent monitoring method for the cross operation of a refining device area comprises the following steps: S1, fixedly erecting binocular cameras in the horizontal direction and the vertical direction of a cross operation area respectively, shooting pictures of the cross operation area, wherein the cross operation area comprises an upper layer of construction platform and a lower layer of construction platform, the erection height of the binocular cameras is the height of the lower layer of construction platform of the cross operation, and the erection distance is based on the condition that the cross operation environment can be covered, and real-time acquisition of RTSP video streams of the binocular cameras is performed for intelligent analysis; S2, respectively carrying out camera calibration, binocular correction and stereo matching on pictures shot by the two binocular cameras, and then judging the pictures shot by the two binocular cameras; S3, if the binocular cameras in two directions shoot pictures and judge that people are suspected to be standing in the lower-layer cross operation area, alarming is carried out on the people standing in the lower-layer cross operation area, the two binocular cameras respectively output two shooting images, and the alarm equipment is controlled to carry out acousto-optic and voice reminding. Further, in step S2, two binocular cameras are calibrated by using OpenCV, which specifically includes the following steps: a. the binocular camera is hor