A safe distance warning method

Abstract

The application provides a safety distance early warning method, relates to the technical field of image early warning, and comprises the following steps: acquiring images of a first target object and a second target object of a to-be-measured distance, marking a first measuring point and a second measuring point on the first target object and the second target object of the to-be-measured distance in the images, setting a reference line in the images, formulating calibration parameters of actual distances and corresponding pixel distances in the images according to actual shape sizes of the first target object and the second target object and the corresponding pixel distances, taking an included angle between a distance measuring line formed by the first measuring point and the second measuring point and the reference line as a selection basis of the calibration parameters, taking a product of a pixel distance of the distance measuring line and the selected calibration parameters as a measured distance of the first target object and the second target object, and comparing the measured distance with a preset safety threshold to make a warning judgment. The early warning method can monitor actual distances between industrial objects in different working scenes through monocular video and make safety early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image early warning, in particular to a safety distance early warning method. BACKGROUND

[0002] In production, it is often necessary to set a safety distance for industrial objects that may produce harmful reactions in phase mixing or phase contact, for example, oxygen and acetylene are generally used for metal cutting operations, and the oxygen-acetylene flame formed in the combustion process has high temperature and can cut common metals. In the intelligent mine maintenance area, oxygen and acetylene are used to ignite the oxygen-acetylene flame, and the high temperature of the combustion can be used for metal cutting, welding, and steel plate shaping and correction, and the use is very extensive. However, in the industrial scene, acetylene is a flammable material, and oxygen is a combustible material. If acetylene leaks, acetylene mixes with air, and meets a spark or open flame, it will explode violently. The explosion will cause the oxygen cylinder to break and leak oxygen, and the oxygen will make the explosion more violent and uncontrollable, causing serious accidents and causing significant loss of life and property. At present, the safety distance early warning of oxygen cylinders and acetylene cylinders relies on manual inspection and judgment. Manual detection is time-consuming and labor-intensive, and the human eye cannot continuously and stably complete these highly repetitive tasks, which can easily cause missed detection and has a low degree of intelligence.

[0003] In order to solve the problem of automatic detection and early warning of the safety distance of oxygen cylinders and acetylene cylinders, some intelligent detection techniques and devices have appeared. For example, Chinese patent application No. CN202123202272.4 discloses a safety distance detection device for oxygen cylinders and acetylene cylinders during construction, which includes a transmitting end and a receiving end. The transmitting end is arranged on one of the oxygen cylinder and the acetylene cylinder, and the receiving end is arranged on the other. Distance sensors and temperature sensors are arranged on the oxygen cylinder and the acetylene cylinder, which can detect the distance between the two in real time during use. Once the dangerous distance is placed, an alarm is sent and the electric valve is closed, thereby achieving the purpose of improving the safety of use. In addition, Chinese patent application No. CN201922227101.3 also discloses an oxygen cylinder and acetylene cylinder safety distance automatic induction valve device, which includes a device main body integrated with an electric control valve and a power module. The electric control valve is electrically connected with the power module. The electric control valve is provided with a distance sensing unit and an alarm unit. The distance sensing unit is electrically connected with the electric control valve and the alarm unit respectively. Two device main bodies are respectively installed in front of the oxygen cylinder valve and the acetylene cylinder valve. The distance between the oxygen cylinder and the acetylene cylinder is automatically sensed and monitored by the distance sensing unit, and the opening / closing of the electric control valve and the alarm action of the alarm unit are automatically controlled according to the monitoring result. It can be seen that the above two schemes are to install detection devices on industrial objects that may produce harmful reactions, i.e. oxygen cylinders and acetylene cylinders, to determine the distance between the oxygen cylinder and the acetylene cylinder. This method not only requires the installation of new accessories on industrial objects with hazards, but also brings inconvenience in use. The added accessories also reduce the reliability in harsh use environment, so there are defects in this scheme.

[0004] With the development of image intelligent recognition technology, using machine vision to measure the required distance can not only further improve the intelligent level of monitoring, but also can ensure the accuracy of safety distance monitoring and recognition, and at the same time, it can also ensure the accuracy of safety distance monitoring and recognition. SUMMARY

[0005] In order to solve the problem of monitoring and early warning of the safety distance between industrial objects that may produce harmful reactions by contacting or mixing in production, the present application provides a safety distance early warning method, which can monitor the actual distance between industrial objects in different working scenes by monocular video and perform safety early warning.

[0006] The technical scheme adopted by the embodiment of the present application to solve the technical problem is: a safety distance early warning method, characterized by: acquiring images of a first target object and a second target object of a distance to be measured, marking a first measurement point and a second measurement point on the first target object and the second target object of the distance to be measured in the images, setting a reference line in the images, formulating a calibration parameter of an actual distance and a corresponding pixel distance in the images according to actual shape sizes of the first target object and the second target object and the corresponding pixel distances, taking an included angle between a distance measuring line formed by the first measurement point and the second measurement point and the reference line as a selection basis of the calibration parameter, and at this time, a product of a pixel distance of the distance measuring line and the selected calibration parameter is a measurement distance of the first target object and the second target object, and a warning judgment is made by comparing the measurement distance with a preset safety threshold.

[0007] In a specific embodiment, image data sets of the first target object and the second target object in different working scenarios are acquired, the image data sets are trained through a semantic segmentation model, and the trained model is called to identify the first target object and the second target object in an image that needs to be measured.

[0008] In a specific embodiment, the reference line set in the images of the first target object and the second target object of the distance to be measured is parallel to a horizontal axis of a pixel coordinate system in the images.

[0009] In a specific embodiment, the first measurement point and the second measurement point on the first target object and the second target object are marked through the following steps:

[0010] S1, the first target object and the second target object in the image are masked, the masked area is separated and recombined with the background to obtain a binary image;

[0011] S2, noise of the binary image of the first target object and the second target object is eliminated;

[0012] S3, a non-continuous region in a region where the first target object and the second target object are located in the binary image is filled by inversion;

[0013] S4, a minimum bounding box of the first target object and the second target object in the binary image is acquired, a bottom edge midpoint of the minimum bounding box of the first target object is marked as the first measurement point, and a bottom edge midpoint of the minimum bounding box of the second target object is marked as the second measurement point.

[0014] In a specific embodiment, the selection basis of the calibration parameter is:

[0015] When the included angle between the distance measuring line and the reference line is 0°≤ the included angle ≤ 30°, a calibration parameter with width as a reference is selected, which is formulated according to actual width sizes of the first target object and the second target object and the corresponding pixel distances;

[0016] 60°≤ the angle between the ranging line and the reference line ≤ 90°, the calibration parameter with height as reference is selected by the actual height size of the first target object and the second target object and the corresponding pixel distance thereof;

[0017] 30°< the angle between the ranging line and the reference line < 60°, the calibration parameter with width and height as comprehensive reference is selected by the actual width and height size of the first target object and the second target object and the corresponding pixel distance thereof.

[0018] In a specific embodiment, the calibration parameter with width as reference is established by the actual width size w1, w2 of the first target object and the second target object and the corresponding pixel distance W1, W2 of the width of the minimum enclosing frame thereof in the image. 宽 :

[0019]

[0020] In a specific embodiment, the calibration parameter with height as reference is established by the actual height size h1, h2 of the first target object and the second target object and the corresponding pixel distance H1, H2 of the height of the minimum enclosing frame thereof in the image. 高 :

[0021]

[0022] In a specific embodiment, the calibration parameter with width and height as comprehensive reference is established by the actual width size w1, w2 of the first target object and the second target object and the corresponding pixel distance W1, W2 of the width of the minimum enclosing frame thereof in the image, actual height size h1, h2 and the corresponding pixel distance H1, H2 of the height of the minimum enclosing frame thereof in the image. 均 :

[0023] wherein

[0024] In a specific embodiment, a correction coefficient is set according to the angle between the ranging line and the reference line, and the measured distance of the first target object and the second target object is processed by the corresponding correction coefficient to reduce the error.

[0025] In a specific embodiment, multiple measured distances of the first target object and the second target object are calculated respectively by multiple images with different shooting angles, compared with the safety threshold, and the frequency of the measured distances exceeding the safety threshold is compared with the preset frequency threshold to make a pre-warning judgment.

[0026] In a specific embodiment, a plurality of images of different shooting angles are acquired, and the midpoint of the distance measuring line pixel distance between the first target object and the second target object is located in the center region of the image, the center region is the region surrounded by the three equal lines of the image pixel width and the three equal lines of the height, and the average value of the plurality of measured distances of the first target object and the second target object is compared with the preset safety threshold to make a pre-warning judgment.

[0027] The advantages of the embodiment of the application are:

[0028] 1. The method analyzes the real-time collected target object image to determine whether the actual distance between the target objects is less than the safety threshold, which endangers production safety, and timely safety warning is performed, labor is saved, detection efficiency and intelligent level are improved, and the possibility of accident occurrence is reduced.

[0029] 2. The method calculates and analyzes the actual distance by the included angle of the target object connecting line and the set reference line, simplifies the mathematical calculation model, establishes the corresponding relationship between the image pixels and the actual distance, and improves the analysis and calculation speed; and the accuracy of the monitoring data is improved by combining the error correction method.

[0030] 3. The method sets multiple shooting preset positions for the monitored target objects, and then multiple verifications are performed on the distances between the target objects in multiple angle images, thereby improving the reliability and safety of the monitoring results.

[0031] 4. The method trains the collected picture data set of each target object in different working scenes by using a semantic segmentation model, then detects the image by calling the trained model, identifies the target object to be measured, so that the monitoring mode is more intelligent, and the recognition rate of the target in the image is also improved. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The working flowchart of the safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder is a safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder.

[0033] Figure 2 The platform building schematic diagram of the safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder is a safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder.

[0034] Figure 3 The minimum rectangular bounding box and the included angle between the distance measuring line and the reference line of the safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder are a safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder.

[0035] Figure 4 The logic flowchart of the safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder is a safety distance pre-warning method for detecting the safety distance between the oxygen cylinder and the acetylene cylinder.

[0036] Figure 5 A central area division method for a safety distance early warning method. DETAILED DESCRIPTION

[0037] The embodiment of the present application provides a safety distance early warning method, solves the problem of safety distance monitoring and early warning between industrial objects which may produce harmful reactions by contacting or mixing in production, and the general idea is as follows:

[0038] Embodiment 1

[0039] The present application provides a safety distance early warning method, by acquiring the images of a first target object and a second target object of a distance to be measured, marking a first measuring point and a second measuring point on the first target object and the second target object of the distance to be measured in the images, setting a reference line in the images, formulating calibration parameters of actual distances and corresponding pixel distances in the images according to actual shape sizes of the first target object and the second target object and the corresponding pixel distances, taking an included angle between a distance measuring line formed by the first measuring point and the second measuring point and the reference line as a selection basis of the calibration parameters, and taking a product of a pixel distance of the distance measuring line and the selected calibration parameters as a measuring distance of the first target object and the second target object, the measuring distance is compared with a preset safety threshold to make a pre-warning judgment. By analyzing the target object images collected in real time, the shape sizes of the target objects and the corresponding pixel distances in the images are utilized, and appropriate calibration parameters are selected in combination with relative geometric position relationships between the target objects and the image reference line, so that whether the actual distance between the target objects is less than the safety threshold is calculated and determined, production safety is harmed, safety pre-warning is made in time, labor is saved, detection efficiency and intelligent level are improved, and the possibility of accident occurrence is reduced.

[0040] In order to better illustrate the application method of the embodiment, the working scene of the oxygen cylinder and the acetylene cylinder in the maintenance area of the smart mine in actual application is taken as an example to illustrate that the first target object is the oxygen cylinder and the second target object is the acetylene cylinder.

[0041] Please refer to Figure 1, first, the video data of the oxygen cylinder and acetylene cylinder in the maintenance area of the smart mine in different working scenes can be collected, the image can be labeled by means of the image labeling tool Labelme, the DeepLabv3+ is used to process each frame image in the video data, and the target contour is obtained, specifically including: collecting clear and unobstructed video frame images of the oxygen cylinder and acetylene cylinder under different working scenes, using the semantic segmentation model DeepLabv3+ to train the collected oxygen cylinder and acetylene cylinder picture data set. By adjusting the PTZ parameters of the industrial field camera, the clear and unobstructed oxygen cylinder and acetylene cylinder video data are recognized, and the trained model is called to detect the frame image; the oxygen cylinder and acetylene cylinder in the picture are detected, the recognized oxygen cylinder and acetylene cylinder are marked with different color masks, the oxygen cylinder and acetylene cylinder target contour is obtained, the mask area and the background of the recognized oxygen cylinder and acetylene cylinder are separated and reorganized based on the HSV color space segmentation method, the mask area containing the oxygen cylinder and acetylene cylinder in the image is set to 1, and the remaining background area is set to 0, and the binary image containing only the oxygen cylinder and acetylene cylinder is reorganized; the image noise is removed by using the Gaussian filtering method, and the FloodFill flooding filling algorithm is used to fill the discontinuous size hole area in the white area of the filtered binary image until all the pixel points in the closed area are completely filled, and then the oxygen cylinder and acetylene cylinder contour image is obtained.

[0042] The reference line is parallel to the horizontal axis of the pixel coordinate system in the image, then the minAreaRect function in OpenCV is used to form a minimum rectangular enclosing frame for the target image after semantic segmentation, the center point coordinates, width and height pixel values and rotation angle relative to the reference line of the oxygen cylinder and acetylene cylinder are obtained, and the pixel distance D pixel between the two bottom points in different working scenes is calculated according to the minimum enclosing frame of the oxygen cylinder and acetylene cylinder.

[0043] Please refer to Figure 2 , the calibration parameters P in this scene are formulated by analyzing and operating the actual different positions of the oxygen cylinder and acetylene cylinder.

[0044] W1 represents the pixel value of the width of the minimum rectangular enclosing frame of the oxygen cylinder, w2 represents the actual size width of the acetylene cylinder, W2 represents the pixel value of the width of the minimum rectangular enclosing frame of the acetylene cylinder, and the calibration parameter P 宽 with width as the reference is obtained:

[0045]

[0046] The actual size height of the oxygen cylinder is represented by h1, the pixel value of the minimum rectangular surrounding frame height of the oxygen cylinder is represented by H1, the actual size height of the acetylene cylinder is represented by h2, and the pixel value of the minimum rectangular surrounding frame height of the acetylene cylinder is represented by H2. The calibration parameter P is obtained by taking height as a reference 高 :

[0047]

[0048] The actual width size w1 and w2 of the oxygen cylinder and the acetylene cylinder and the corresponding pixel distance W1 and W2 of the minimum surrounding frame thereof in the image, and the actual height size h1 and h2 and the corresponding pixel distance H1 and H2 of the minimum surrounding frame thereof in the image, are taken as comprehensive references for the width and height. The calibration parameter P is obtained 均 :

[0049] wherein

[0050] Please refer to Figure 3 In general, when the included angle between the distance measuring line and the reference line is 0°≤ included angle ≤ 30°, the calibration parameter P taking the width as a reference can be selected 宽 ; when the included angle between the distance measuring line and the reference line is 60°≤ included angle ≤ 90°, the calibration parameter P taking the height as a reference can be selected 高 ; and when the included angle between the distance measuring line and the reference line is 30°< included angle < 60°, the calibration parameter P taking the width and height as comprehensive references can be selected 均 .

[0051] After the calibration parameter is selected according to the included angle between the distance measuring line formed by the midpoints of the bottom edges of the minimum surrounding frames of the oxygen cylinder and the acetylene cylinder and the reference line, the measured distance Distance of the oxygen cylinder and the acetylene cylinder can be calculated by the product of the pixel distance D of the distance measuring line formed by the midpoints of the bottom edges of the minimum surrounding frames thereof and the corresponding calibration parameter pixel :

[0052] Distance = D pixel · P; wherein P is selected from P 宽 , P 高 , and P 均 according to the included angle between the distance measuring line and the reference line.

[0053] In actual use, due to the existence of system error, the average value of the measured data deviates from its true value under the same conditions, so it is necessary to correct the calculated measurement distance. For constant system error, correction measures can be taken, and a correction coefficient is set according to the included angle between the ranging line and the reference line. The measurement distance of the oxygen cylinder and the acetylene cylinder is processed through the corresponding correction coefficient to reduce the error. For example, in actual production, when 0°≤the included angle between the ranging line and the reference line≤30°, the measurement distance of the oxygen cylinder and the acetylene cylinder is smaller than the true value, and the correction value of 20% is used to eliminate the system error in this scenario; when 60°≤the included angle between the ranging line and the reference line≤90°, the measurement distance of the oxygen cylinder and the acetylene cylinder is smaller than the true value, and the correction value of 14% is used to eliminate the system error in this scenario; when 30°<the included angle between the ranging line and the reference line<60°, the measurement data of the oxygen cylinder and the acetylene cylinder is larger than the true value, and the correction value of 11% is used to eliminate the system error in this scenario, so as to effectively avoid the occurrence of safety accidents and the generation of missed detection. Therefore, please refer to Figure 4 , in order to realize the judgment calculation of the program in the computer, the angle threshold of the included angle between the ranging line and the reference line can be set, the relationship between the included angle and the angle threshold is judged, for example, the angle threshold m1 is 30°, and the angle threshold n1 is 60°, so as to select the corresponding calibration parameter and correction coefficient to calculate and predict the actual distance. According to the calculated measurement distance and the safety threshold, when it is less than the safety threshold, the related data of the oxygen cylinder and the acetylene cylinder is saved, and when it exceeds the safety threshold, safety warning is carried out.

[0054] Embodiment 2:

[0055] The difference from embodiment 1 is that in this embodiment, multiple measurement distances of the first target object oxygen cylinder and the second target object acetylene cylinder are calculated respectively from multiple images with different shooting angles, and compared with the safety threshold, and the frequency of exceeding the safety threshold is compared with the preset frequency threshold to make a warning judgment, so as to further improve the reliability of the warning.

[0056] Specifically, please continue to refer to Figure 1When the measured distance calculated by analyzing and calculating the oxygen cylinder and acetylene cylinder in the photographed image exceeds the safety threshold, in order to improve the measurement accuracy and reduce the false alarm rate, multiple images of different angles can be obtained by rotating the camera to obtain video frame images of the oxygen cylinder and acetylene cylinder in the working scene at different shooting angle positions, and multiple verification is performed on the oxygen cylinder and acetylene cylinder frame images at multiple angles. If the multiple images of different angles photographed by the camera during rotation have some images whose calculated and predicted measured distance is safe compared with the safety threshold, and some images whose calculated and predicted measured distance is unsafe compared with the safety threshold, the frequency of exceeding the safety threshold is compared with the preset frequency threshold, and a warning is made if the frequency of exceeding the safety threshold is greater than the frequency threshold. As an optional solution, multiple images of different angles can also be obtained by rotating the camera to determine whether the midpoint of the pixel distance of the distance line between the first target object and the second target object is within the center region of the two-dimensional image. The center region is divided as follows: Figure 5 In order to realize smaller error of the gas cylinder in the region, the center region is divided as follows: the 1 / 3 and 2 / 3 positions of the two-dimensional image are determined from left to right and from top to bottom, for example, L1 and L2 are the three equal division lines of the 1 / 3 and 2 / 3 positions of the two-dimensional image from left to right, and L3 and L4 are the three equal division lines of the 1 / 3 and 2 / 3 positions of the two-dimensional image from top to bottom. The region surrounded by L1, L2, L3 and L4 is set as the center region, that is, the region formed by connecting the intersection points a, b, d and c of the three equal division lines as shown in Figure 5 The midpoint of the pixel distance of the line connecting the bottom edges of the minimum enclosing frame of the oxygen cylinder and the acetylene cylinder is located in the image center region as much as possible. The images of different shooting angles with the midpoint of the pixel distance of the distance line located in the center region are selected, and the average value of the multiple measured distances of the oxygen cylinder and the acetylene cylinder calculated from each image is compared with the preset safety threshold to make a warning judgment. After the warning is made, the kafka software can be used for message pushing and consumption, the video recording instruction is sent to the Deepstream architecture, the alarm video is intelligently recorded, the alarm information is pushed to the server and stored, and finally transmitted to the on-site engineer through the display terminal, sent to the service end for manual inspection and confirmation, to avoid safety accidents.

[0057] In summary, the safety distance warning method provided by the application can monitor the actual distance between industrial objects in different working scenes through monocular video and make safety warning, which can effectively solve the problem of safety distance monitoring and warning between industrial objects that may produce harmful reactions by contacting or mixing in production.

[0058] It should be noted that the above-mentioned embodiments are merely used to clearly illustrate the technical solutions of the present application, and should not be construed as limitations to the present application. Based on the above-mentioned embodiments, those skilled in the art can make other variations or modifications without departing from the spirit of the present application. The present application is not required to enumerate all of the embodiments, and the variations or modifications made without departing from the spirit of the present application should fall within the scope of the present application.

Claims

1. A method for early warning of safe distance, characterized in that: Acquire images of the first and second target objects whose distances are to be measured; Mark the first measurement point and the second measurement point for the first target object and the second target object whose distance is to be measured in the image, respectively, and set reference lines in the image, wherein the reference lines set in the image of the first target object and the second target object whose distance is to be measured are parallel to the horizontal axis of the pixel coordinate system in the image; Based on the actual shape and size of the first target object and the second target object and their corresponding pixel distance, calibration parameters are established for the actual distance in the image and their corresponding pixel distance. The angle between the ranging line formed by the line connecting the first and second measurement points and the reference line is used as the basis for selecting calibration parameters. At this time, the product of the pixel distance of the ranging line and the selected calibration parameter is the measurement distance between the first target and the second target. The first and second measurement points on the first and second target are marked by the following steps: S1. Mask the first and second target objects in the image, separate and reconstruct the masked areas from the background to obtain a binary image; S2. Remove noise from the binary images of the first and second target objects; S3. Invert and fill the discontinuous regions in the areas where the first and second target objects are located in the binary image; S4. Obtain the minimum bounding box of the first target object and the second target object in the binary image. Mark the midpoint of the bottom edge of the minimum bounding box of the first target object as the first measurement point and the midpoint of the bottom edge of the minimum bounding box of the second target object as the second measurement point. The selection criteria for the calibration parameters are as follows: When the angle between the ranging line and the reference line is ≤30°, the calibration parameters with width as the reference are selected based on the actual width dimensions of the first target object and the second target object and their corresponding pixel distance. When the angle between the ranging line and the reference line is 60°≤ and 90°≤, the calibration parameters with height as the reference are selected based on the actual height dimensions of the first target object and the second target object and their corresponding pixel distance. When the angle between the ranging line and the reference line is less than 60° and the angle between the ranging line and the reference line is less than 30°, the calibration parameters are determined by the actual width and height of the first target object and the second target object and their corresponding pixel distance, with width and height as comprehensive references. The method for determining the calibration parameter with width as a reference is as follows: based on the actual width dimensions w1 and w2 of the first and second target objects and the width of their minimum bounding box in the image as the corresponding pixel distances W1 and W2, the calibration parameter P_width with width as a reference is obtained. ； The method for determining the calibration parameter with height as a reference is as follows: based on the actual height dimensions h1 and h2 of the first and second target objects and the corresponding pixel distances H1 and H2 of their minimum bounding box heights in the image, the calibration parameter P_height with height as a reference is obtained. ； The method for determining calibration parameters using width and height as a comprehensive reference is as follows: Based on the actual width dimensions w1 and w2 of the first and second target objects and the width of their minimum bounding box in the image as the corresponding pixel distances W1 and W2, and the actual height dimensions h1 and h2 of the second target objects and the height of their minimum bounding box in the image as the corresponding pixel distances H1 and H2, the calibration parameter P, using width and height as a comprehensive reference, is obtained. ； Among them, a correction coefficient is set according to the angle between the ranging line and the reference line, and the measured distance between the first target and the second target is processed by the corresponding correction coefficient to reduce the error; The measured distance is compared with a preset safety threshold to make a warning judgment; wherein, image datasets of the first target object and the second target object under different working scenarios are acquired, the image datasets are trained by a semantic segmentation model, and the trained model is called to identify the first target object and the second target object in the image where the distance needs to be measured.

2. The safe distance early warning method as described in claim 1, characterized in that, By calculating multiple measured distances of the first and second targets from multiple images taken from different angles and comparing them with a safety threshold, a warning judgment is made by comparing the frequency of exceeding the safety threshold with a preset frequency threshold.

3. The safe distance early warning method as described in claim 2, characterized in that, Multiple images are acquired from different shooting angles, with the midpoint of the pixel distance between the first target and the second target located in the imaging center region. The center region is the area enclosed by the trisection of the image pixel width and the trisection of the image height. The average value of multiple measured distances between the first target and the second target is calculated from each image and compared with a preset safety threshold to make a warning judgment.

Images