Method for realizing hoisting risk detection by using monocular vision

Abstract

The invention relates to the technical field of computer vision, in particular to a method for achieving hoisting risk detection through monocular vision. The method comprises the following steps: determining a target detection area, automatically collecting video data of a hoisting operation site through a camera, obtaining a target image through frame extraction processing, performing recognition processing on the target image, obtaining an external rectangular frame, taking the target image under an ideal distance as a reference, calculating the diameter of a hoisting arm in the target image, and calculating the diameter of the hoisting arm according to the external rectangular frame. Determining a rotation center point, obtaining a first offset point and a second offset point, establishing a first to-be-detected area in cooperation with the left lower vertex of the external rectangular frame, extracting a second to-be-detected area by using a computer vision technology and a deep learning model, and judging whether all personnel obtaining the target image fall in the second to-be-detected area; and if the hoisting risk is detected, early warning is given out immediately. The method provided by the invention is more accurate and reliable in detection result, higher in real-time performance and lower in cost.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision, and in particular to a method for detecting lifting risks using monocular vision. Background Art

[0002] A crane, commonly known as a crane, is a machine that performs cyclic and intermittent motion. A working cycle of a crane includes vertically picking up the load, moving the load, and unloading the load. Each working cycle uses the crane's lifting mechanism, which includes wire ropes, pulleys, hooks, and rings. Lifting is the process of installing and positioning the load by the crane. Lifting operations can complete tasks such as equipment movement, emergency rescue, lifting, mechanical installation, and rescue, and play an important role in construction, oil and gas field operations, and other sites. The application of lifting is very extensive, and the probability of lifting safety risks has also increased significantly. Lifting safety risks are mostly caused by various illegal operations, and one of the hazard factors is standing under the boom. In the "Ten No-Standing" rules for safe production issued by relevant industries, "It is strictly forbidden to stand under the running load or under the operating radius of the crane boom" is the first item.

[0003] In fact, in order to avoid the safety risks of hoisting, relevant industries have gradually formed corresponding risk monitoring mechanisms, but they are limited to arranging relevant supervisors near the hoisting site, or using traditional video surveillance to monitor the risks of on-site hoisting. In actual production, since the crane's boom will rotate according to specific operation requirements, supervisors and video surveillance will have blind spots, and cannot accurately judge the specific situation under the boom in the blind spot, resulting in inadequate supervision. In addition, the current supervision method has poor real-time performance and cannot prevent problems before they occur.

[0004] Computer vision technology is a hot emerging field in China. With the development of computer vision technology, more and more researchers are trying to apply this technology to the detection of personnel under the crane arm. However, binocular vision is generally used. Although it has good accuracy, the hoisting environment is mostly a construction site or an oil and gas field operation site, which has the characteristics of strong light. Binocular vision is sensitive to light and has poor detection effect under strong light, making it unsuitable for use in the above environment. At the same time, the amount of calculation required for binocular vision is high, and the real-time performance does not meet the actual needs. In addition, many hoisting construction sites have large operating areas and complex operating scenes, and hundreds of cameras need to be installed. Compared with a single camera, the application cost of binocular vision is higher. Therefore, in view of the above shortcomings, a method for hoisting risk detection using monocular vision is proposed. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] The present invention provides a method for detecting hoisting risks using monocular vision, so as to overcome the problems of poor safety monitoring and detection effect, poor real-time performance and high application cost in the prior art of hoisting site.

[0007] (II) Technical solution

[0008] In order to solve the above problems, the present invention provides a method for detecting lifting risks using monocular vision, comprising:

[0009] Step S1: determining a target detection area, and installing at least one camera in the target detection area, automatically collecting video data of the hoisting operation site through the camera, performing frame extraction processing on the collected video data, and obtaining a target image;

[0010] Step S2: performing recognition processing on the target image obtained in step S1, identifying the crane boom, the crane operating room, the suspended objects and personnel on the ground, and obtaining their respective circumscribed rectangular frames;

[0011] Step S3: using the target image at the ideal distance as a reference, calculating the diameter of the boom in the target image, and determining the rotation center point of the operating room or the boom according to the circumscribed rectangular frame obtained in step S2;

[0012] Step S4: acquiring a first offset point and a second offset point according to the rotation center point acquired in step S3 and the diameter of the boom in the target image;

[0013] Step S5: creating a first area to be detected according to the rotation center point and the lower left vertex of the circumscribed rectangular frame obtained in step S3, and the first offset point and the second offset point obtained in step S4;

[0014] Step S6: using computer vision technology and deep learning models, extracting the area formed by the ground of the first area to be detected obtained in step S5 as the second area to be detected;

[0015] Step S7: Based on the second area to be detected obtained in step S6, determine whether all the persons in the target image obtained in step S1 fall within the second area to be detected. If they do, a lifting risk is detected and an early warning is issued immediately.

[0016] Preferably, in step S1, the target detection area is a lifting operation site, and the target images all perform operations from "key information identification to risk detection and early warning".

[0017] Preferably, in step S3, the target image at the ideal distance at least satisfies the conditions that the image contains all objects that need to be detected and identified, and the size of the crane occupies 1 / n of the screen.

[0018] Preferably, the diameter of the boom in the target image is calculated based on an empirical value of the target image scale at an ideal distance, and the calculation formula is:

[0019] d g =e×r g ;

[0020]

[0021] Where, d g The diameter of the boom in the target image is calculated based on the empirical value of the target image scale at the ideal distance; e is the empirical value of the target image scale at the ideal distance; r g is the reference size in the target image; m is the number of times the camera focus is adjusted when the ideal distance is reached; s i is the size of the crane in the image taken after the camera focus is adjusted for the i-th time; r i It is the reference size of the crane in the image taken after the camera focus is adjusted for the i-th time.

[0022] Preferably, in step S3, the rotation center point is the intersection of the extended line of the left side of the circumscribed rectangular frame of the boom and the extended line of the lower side of the circumscribed rectangular frame of the operating room, and the intersection is used as the origin to form a two-dimensional rectangular coordinate system.

[0023] Preferably, in step S4, the first offset point and the second offset point are both obtained by offsetting the relative positions of the circumscribed rectangular frame of the hanging object and the horizontal axis and the vertical axis of the coordinate system.

[0024] Preferably, when the suspended object is above the horizontal axis, the first offset point is obtained by offsetting the rotation center point in the positive direction along the longitudinal axis by the diameter of the suspension arm on the target image; when the suspended object is below the horizontal axis, the first offset point is obtained by offsetting the rotation center point in the negative direction along the longitudinal axis by the diameter of the suspension arm on the target image.

[0025] Preferably, when the hanging object is to the right of the longitudinal axis, the second offset point is obtained by offsetting the lower left vertex of the circumscribed rectangular frame of the hanging object upward along the edge where it is located by the distance of the diameter of the hanging arm on the target image; when the hanging object is to the left of the longitudinal axis, the second offset point is obtained by offsetting the lower right vertex of the circumscribed rectangular frame of the hanging object upward along the edge where it is located by the distance of the diameter of the hanging arm on the target image.

[0026] (III) Beneficial effects

[0027] Compared with the existing binocular vision detection method, the method for detecting lifting risks using monocular vision provided by the present invention adopts a single camera for shooting and detection. On the premise that the accuracy meets the requirements, it is insensitive to light and is suitable for working environments with strong light. It has smaller calculation amount, higher real-time performance and lower application cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a method for detecting lifting risks using monocular vision according to an embodiment of the present invention;

[0029] Figure 2 This is an example of a simplified diagram of a target image for implementing a method for detecting lifting risks using monocular vision in an embodiment of the present invention.

[0030] Legend: External rectangular frame ACDB-crane boom; external rectangular frame NOPQ-crane operating room; external rectangular frame RUTS-suspended objects on the ground; external rectangular frame EKLF-personnel; external rectangular frame GXYZ-personnel; V-first offset point; W-second offset point; external rectangular frame MVUW-first area to be detected; external rectangular frame HIUW-second area to be detected. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Figure 1 FIG. 1 is a flow chart of a method for detecting lifting risks using monocular vision according to an embodiment of the present invention. Figure 1 As shown, the present invention provides a method for detecting lifting risks using monocular vision, comprising:

[0033] Step S1: determine a target detection area, and install at least one camera in the target detection area, automatically collect video data of the lifting operation site through the camera, perform frame extraction processing on the collected video data, and obtain a target image.

[0034] In step S1, the target detection area is a hoisting operation site, and the target images all perform operations from "key information identification to risk detection and early warning".

[0035] Step S2: The target image acquired in step S1 is processed for recognition, the crane boom, the crane operating room, the suspended objects and personnel on the ground are recognized, and their respective outer rectangular frames are acquired.

[0036] Step S3: Using the target image at the ideal distance as a reference, calculate the diameter of the boom in the target image, and determine the rotation center point of the operating room or the boom according to the circumscribed rectangular frame obtained in step S2.

[0037] In step S3, the target image at the ideal distance at least satisfies the condition that the image contains all objects that need to be detected and identified, and the size of the crane occupies 1 / n of the screen, wherein the diameter of the boom in the target image is calculated according to the empirical value of the scale of the target image at the ideal distance, and the calculation formula is:

[0038] d g =e×r g ;

[0039]

[0040] Where, d g The diameter of the boom in the target image is calculated based on the empirical value of the target image scale at the ideal distance; e is the empirical value of the target image scale at the ideal distance; r g is the reference size in the target image; m is the number of times the camera focus is adjusted when the ideal distance is reached; s i is the size of the crane in the image taken after the camera focus is adjusted for the i-th time; r i It is the reference size of the crane in the image taken after the camera focus is adjusted for the i-th time.

[0041] It should be noted that the position of the camera at the work site is often fixed, but the crane is mobile. If the camera is too far away from the crane, the shooting will be unclear, and if the camera is too close to the crane, the shooting will be incomplete. Therefore, there is an ideal distance between the camera and the crane. Only at the ideal distance can the captured image be accurately detected and analyzed. The target image captured at the ideal distance at least meets the following conditions: the image contains all objects that need to be detected and identified, and the size of the crane occupies 1 / n of the screen, where n is determined according to actual usage requirements.

[0042] In addition, in practical applications, the rotation center point is the intersection of the extended line of the left side of the circumscribed rectangular frame of the boom and the extended line of the lower side of the circumscribed rectangular frame of the operating room, and the intersection is used as the origin to form a two-dimensional rectangular coordinate system.

[0043] Step S4: According to the rotation center point obtained in step S3 and the diameter of the boom in the target image, a first offset point and a second offset point are obtained.

[0044] In step S4, the first offset point and the second offset point are both obtained by offsetting the relative positions of the circumscribed rectangular frame of the hanging object and the horizontal axis and the vertical axis of the coordinate system.

[0045] Therefore, when the suspended object is above the horizontal axis, the first offset point is obtained by offsetting the rotation center point in the positive direction along the longitudinal axis by the diameter of the suspension arm on the target image; when the suspended object is below the horizontal axis, the first offset point is obtained by offsetting the rotation center point in the negative direction along the longitudinal axis by the diameter of the suspension arm on the target image.

[0046] Similarly, when the suspended object is to the right of the longitudinal axis, the second offset point is obtained by offsetting the lower left vertex of the circumscribed rectangular box of the suspended object upward along the edge where it is located by the distance of the diameter of the boom on the target image; when the suspended object is to the left of the longitudinal axis, the second offset point is obtained by offsetting the lower right vertex of the circumscribed rectangular box of the suspended object upward along the edge where it is located by the distance of the diameter of the boom on the target image.

[0047] Step S5: Create a first area to be detected based on the rotation center point and the lower left vertex of the circumscribed rectangular frame obtained in step S3, and the first offset point and the second offset point obtained in step S4.

[0048] Step S6: Using computer vision technology and deep learning models, extract the area formed by the ground of the first area to be detected obtained in step S5 as the second area to be detected.

[0049] Step S7: Based on the second area to be detected obtained in step S6, determine whether all the persons in the target image obtained in step S1 fall within the second area to be detected. If they do, a lifting risk is detected and an early warning is issued immediately.

[0050] In practical applications, the rotation center of the crane operating room or the boom is first found based on the circumscribed rectangular frame of the crane operating room and the boom. Then, based on the rotation center, the load, and the diameter of the boom on the target image, the projection of the boom on the ground when the boom picks up the load is accurately found. In addition, when calculating the diameter of the boom on the target image, the empirical value of the target image scale at the ideal distance and the reference size in the target image are introduced to further make the overall detection result more accurate and reliable.

[0051] It should be noted that in this method, the lifting risk is detected by using the bottom coordinates of the personnel's external rectangular box rather than whether the top coordinates fall in the dangerous area. The detection result is more accurate and effectively avoids the problem of wasting time and effort due to a large number of false alarms, which seriously affects normal lifting.

[0052] In order to further understand the workflow of the method for implementing the risk detection of hoisting using monocular vision, in this embodiment, the following steps are taken when detecting the risk of hoisting operation:

[0053] Step 1: Determine the target detection area and install at least one camera in the target detection area. Automatically collect video data of the lifting operation site through the camera, perform frame extraction processing on the collected video data, and obtain the target image.

[0054] In this embodiment, the target detection area is the lifting operation site. At least one camera is set at the lifting operation site to automatically collect video data of the lifting operation site. The collected video data is subjected to frame extraction processing to obtain the target image. Steps 2 to 7 are performed for each frame of the target image. The frame extraction interval is set according to actual business needs, such as extracting 5 frames of target images every second.

[0055] Step 2: The target image obtained in step 1 is processed for recognition, the crane boom, the crane operating room, the suspended objects and personnel on the ground are identified, and their respective circumscribed rectangular frames are obtained.

[0056] In this embodiment, Figure 2 This is an example of a target image diagram of a method for detecting a risk of lifting using monocular vision according to an embodiment of the present invention. Figure 2 As shown, the crane's boom, the crane's operating room, the suspended object on the ground, and the personnel are identified in the target image, and their circumscribed rectangular frames are the crane's boom ACDB, the crane's operating room NOPQ, the suspended object RUTS on the ground, the personnel EKLF, and the personnel GXYZ.

[0057] Step 3: Using the target image at the ideal distance as a reference, calculate the diameter of the boom in the target image, and determine the rotation center point of the operating room or the boom according to the circumscribed rectangular frame obtained in step S2.

[0058] The diameter of the boom in the target image is calculated based on the empirical value of the target image scale at the ideal distance, and the calculation formula is:

[0059] d g =e×r g

[0060] Where, d g The diameter of the boom in the image is calculated based on the empirical value of the target image scale at the ideal distance; e is the empirical value of the target image scale at the ideal distance; r g is the reference size in the target image.

[0061] In practical applications, the ideal distance is usually unknown and needs to be determined by continuously adjusting the focal length of the camera at the work site. Calculate the empirical value e of the target image scale at the ideal distance:

[0062]

[0063] Where m is the number of times the camera focus is adjusted to achieve the ideal distance, and s is i is the size of the crane in the image taken after adjusting the camera focus for the i-th time, r i It is the reference size of the crane in the image taken after the camera focus is adjusted for the i-th time.

[0064] It should be noted that the diameter of the boom in the target image is not easy to measure accurately, so an easily measurable dimension of an object in the image is selected as a reference dimension, which is obtained through manual measurement or other measurement methods.

[0065] In this embodiment, n=3, that is, the image captured at the ideal distance at least meets the following conditions: the image includes the crane boom, the crane operating room, the suspended objects on the ground, and the personnel, and the size of the crane occupies 1 / 3 of the screen. Assume that when the ideal distance is reached, the camera focus is adjusted three times, that is, m=3.

[0066] The first corresponding

[0067] The second corresponding

[0068] The third corresponding

[0069] but

[0070] In this embodiment, the height of the crane operating room in the target image at the selected ideal distance is used as the reference size r g =1.29cm, then the diameter of the boom in the target image is d g =1 / 3×1.29=0.43cm.

[0071] In this embodiment, the rotation center point M of the operating room or the boom is determined: the intersection point M of the extension line of the left length AC of the boom circumscribed rectangular frame ACDB and the extension line of the lower length PQ of the operating room outer rectangular frame NOPQ is used as the origin to form a two-dimensional rectangular coordinate system, and point M is equivalent to the rotation center of the operating room or the boom.

[0072] Step 4: According to the rotation center point obtained in step 3 and the diameter of the boom in the target image, obtain the first offset point and the second offset point.

[0073] In this embodiment, the relative positions of the external rectangular frame RUTS of the hanging object and the horizontal axis and the vertical axis of the coordinate system are determined, and an offset operation is performed to obtain a first offset point V and a second offset point W, specifically:

[0074] When the hanging object is above the horizontal axis, the first offset point V is obtained by offsetting the diameter of the hanging arm in the target image along the longitudinal axis from the rotation center point M in the positive direction; when the hanging object is below the horizontal axis, the first offset point V is obtained by offsetting the diameter of the hanging arm in the target image along the longitudinal axis from the rotation center point M in the negative direction;

[0075] At the same time, when the suspended object is to the right of the longitudinal axis, the lower left vertex U of the circumscribed rectangular frame RUTS of the suspended object is pointed upward along the side UR to the direction of R, and the second offset point W is obtained by offsetting the diameter of the boom in the target image; when the suspended object is to the left of the longitudinal axis, the lower right vertex T of the circumscribed rectangular frame RUTS of the suspended object is pointed upward along the side TS to the direction of S, and the second offset point W is obtained by offsetting the diameter of the boom in the target image.

[0076] In this embodiment, the hanging object is in the fourth quadrant of the coordinate system, so the first offset point V is obtained by offsetting the rotation center point M along the vertical axis in the negative direction by 0.43 cm of the diameter of the hanging arm in the target image, and the second offset point W is obtained by offsetting the lower left vertex U of the circumscribed rectangular frame RUTS of the hanging object along its edge UR upward, pointing in the direction of R, by 0.43 cm of the diameter of the hanging arm in the target image.

[0077] Step 5: Create a first area to be detected based on the rotation center point and the lower left vertex of the circumscribed rectangular frame obtained in step 3, and the first offset point and the second offset point obtained in step 4.

[0078] In this embodiment, the rotation center point M, the lower left vertex U of the circumscribed rectangular frame RUTS of the suspended object, the first offset point V and the second offset point W are connected to obtain the first area to be detected MVUW. It should be noted that, in this embodiment, MW is equivalent to the projection of the boom on the ground, and the first area to be detected MVUW formed by expanding the projection with the diameter of the boom on the figure is equivalent to the projection area of ​​the boom on the ground.

[0079] In practical applications, the rotation center of the crane operating room or the boom is first found based on the circumscribed rectangular frame of the crane operating room and the boom. Then, based on the diameter of the rotation center, the load and the boom in the image, the projection of the boom on the ground when the boom picks up the load is accurately found, making the overall detection result more accurate and reliable.

[0080] Step 6: Using computer vision technology and deep learning models, extract the area formed by the ground of the first area to be detected obtained in step 5 as the second area to be detected.

[0081] In this embodiment, computer vision technology and deep learning models are used to extract the area formed by the ground in the first area to be detected MVUW as the second area to be detected HIUW. The second area to be detected HIUW is the dangerous area under the boom. In actual lifting operations, the dangerous area often does not need to be accurately located directly below the boom, but should be determined as an area below the boom, taking into account the impact of the hoisted objects during the lifting process, as in the area HIUW in the embodiment.

[0082] Step 7: Based on the second area to be detected obtained in step 6, determine whether all the personnel in the target image obtained in step 1 fall within the second area to be detected. If they do, a lifting risk is detected and an early warning is issued immediately.

[0083] In this embodiment, it is determined whether the bottom coordinates of the external rectangular frame EKLF, GXYZ of all personnel fall within the second area to be detected HIUW. If so, it is determined that the personnel corresponding to the external rectangular frame are in the dangerous area under the boom, that is, the hoisting risk is detected, and an early warning message can be immediately sent to the monitoring center and / or the hoisting site. In the embodiment of the present invention, only the bottom coordinates of the external rectangular frame EKLF of the personnel fall within the second area to be detected HIUW, that is, the hoisting risk is detected. At this time, the hoisting risk early warning message can be immediately sent to the monitoring center. At the same time, an audio and / or optical alarm signal can be immediately issued at the hoisting construction site to promptly drive the personnel corresponding to the external rectangular frame EKLF out of the dangerous area.

[0084] In actual applications, the bottom coordinates of the person's circumscribed rectangular box are used instead of the top coordinates to determine whether there is a person under the boom during the boom picking up the suspended object. If the bottom coordinates of the person's circumscribed rectangular box fall in the dangerous area, it must mean that the person corresponding to the circumscribed rectangular box is in the dangerous area under the boom, that is, the lifting risk is detected. If the top coordinates of the person's circumscribed rectangular box are used for judgment, the detection result is not accurate, because in the two-dimensional image, the person's head often enters the dangerous area, and then the lifting risk warning information is immediately issued, but in fact the person has not walked to the dangerous area, that is, in fact the person has not entered the dangerous area under the boom, and there is no lifting risk. Such inaccurate detection results will cause a large number of false alarms, which will waste time and effort in vain, and seriously affect the normal lifting operation.

[0085] The embodiment of the present invention only exemplarily takes personnel as the object of lifting risk detection. In fact, all physical elements in the dangerous area under the boom that pose a safety risk to the lifting operation can be used as the object of lifting risk detection. Using monocular vision to detect whether these objects cause lifting risks due to being in the dangerous area under the boom falls within the scope of the present invention.

[0086] Compared with the existing binocular vision detection method, the method for detecting lifting risks using monocular vision provided by the present invention adopts a single camera for shooting and detection. On the premise that the accuracy meets the requirements, it is insensitive to light and is suitable for working environments with strong light. It has smaller calculation amount, higher real-time performance and lower application cost.

[0087] The above implementation modes are only used to illustrate the present invention, but not to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the present invention. The patent protection scope of the present invention should be defined by the claims.

Claims

1. A method for detecting lifting risks using monocular vision, characterized in that: include: Step S1: determining a target detection area, and installing at least one camera in the target detection area, automatically collecting video data of the hoisting operation site through the camera, performing frame extraction processing on the collected video data, and obtaining a target image; Step S2: performing recognition processing on the target image obtained in step S1, identifying the crane boom, the crane operating room, the suspended objects and personnel on the ground, and obtaining their respective circumscribed rectangular frames; Step S3: using the target image at the ideal distance as a reference, calculating the diameter of the boom in the target image, and determining the rotation center point of the operating room or the boom according to the circumscribed rectangular frame obtained in step S2; Step S4: acquiring a first offset point and a second offset point according to the rotation center point acquired in step S3 and the diameter of the boom in the target image; Step S5: creating a first area to be detected according to the rotation center point and the lower left vertex of the outer rectangular frame obtained in step S3, and the first offset point and the second offset point obtained in step S4; Step S6: using computer vision technology and deep learning models, extracting the area formed by the ground of the first area to be detected obtained in step S5 as the second area to be detected; Step S7: Based on the second area to be detected obtained in step S6, determine whether all the persons in the target image obtained in step S1 fall within the second area to be detected. If they do, a lifting risk is detected and an early warning is issued immediately.

2. The method for detecting lifting risks using monocular vision according to claim 1 is characterized in that: In step S1, the target detection area is the lifting operation site, and the target images all perform the operations of "key information identification to risk detection and early warning".

3. The method for detecting lifting risks using monocular vision according to claim 1, characterized in that: In step S3, the target image at the ideal distance at least satisfies the conditions that the image contains all objects that need to be detected and identified, and the size of the crane occupies 1 / n of the screen.

4. The method for detecting lifting risks using monocular vision according to claim 3 is characterized in that: The diameter of the boom in the target image is calculated based on the empirical value of the target image scale at the ideal distance, and the calculation formula is: d g =e×r g ; Where, d g The diameter of the boom in the target image is calculated based on the empirical value of the target image scale at the ideal distance; e is the empirical value of the target image scale at the ideal distance; r g is the reference size in the target image; m is the number of times the camera focus is adjusted when the ideal distance is reached; s i is the size of the crane in the image taken after the camera focus is adjusted for the i-th time; r i It is the reference size of the crane in the image taken after the camera focus is adjusted for the i-th time.

5. The method for detecting lifting risks using monocular vision according to claim 1, characterized in that: In step S3, the rotation center point is the intersection of the extended line of the left side of the circumscribed rectangular frame of the boom and the extended line of the lower side of the circumscribed rectangular frame of the operating room, and the intersection is used as the origin to form a two-dimensional rectangular coordinate system.

6. The method for detecting lifting risks using monocular vision according to claim 1, characterized in that: In the step S4, the first offset point and the second offset point are both obtained by offsetting the relative positions of the circumscribed rectangular frame of the hanging object and the horizontal axis and the vertical axis of the coordinate system.

7. The method for detecting risks of hoisting using monocular vision according to claim 6, characterized in that: When the suspended object is above the horizontal axis, the first offset point is obtained by offsetting the rotation center point in the positive direction along the longitudinal axis by the diameter of the suspension arm on the target image; when the suspended object is below the horizontal axis, the first offset point is obtained by offsetting the rotation center point in the negative direction along the longitudinal axis by the diameter of the suspension arm on the target image.

8. The method for detecting risks of hoisting using monocular vision according to claim 6, characterized in that: When the hanging object is to the right of the longitudinal axis, the second offset point is obtained by offsetting the lower left vertex of the circumscribed rectangular frame of the hanging object upward along the edge where it is located by the distance of the diameter of the hanging arm on the target image; when the hanging object is to the left of the longitudinal axis, the second offset point is obtained by offsetting the lower right vertex of the circumscribed rectangular frame of the hanging object upward along the edge where it is located by the distance of the diameter of the hanging arm on the target image.

Images