Anti-jumping bridge rescue system and emergency escape system

Abstract

The invention relates to the field of municipal safety, and particularly discloses an anti-jumping bridge rescue system and an emergency escape system.The anti-jumping bridge rescue system comprises an image module, the image module comprises an image acquisition module and an image analysis module, and the image acquisition module is installed on the outer side of a bridge deck handrail and used for acquiring image information near the handrail; the image analysis module is used for comparing the image information according to adjacent frames to obtain a static image and a dynamic image, then identifying the handrail in the static image according to the continuity of pixels, analyzing whether the height of the upper edge of the dynamic image exceeding the handrail exceeds a preset value, and if the height exceeds the preset value, determining that the upper edge of the dynamic image exceeds the preset value; if yes, whether the upper edge and the lower edge in the dynamic image are continuous is analyzed, and if yes, a suspected jump bridge signal is generated; and the control module is used for receiving the suspected bridge jumping signal, generating a first control signal and then generating a second control signal after waiting for a preset duration. By adopting the technical scheme of the invention, the recognition accuracy can be ensured, and the falling personnel can be well protected.

Description

Technical Field

[0001] The invention relates to the field of municipal safety, and in particular to an anti-jumping bridge rescue system and an emergency escape system. Background Art

[0002] With the acceleration of urbanization, bridges are becoming more and more important as transportation hubs and urban landscapes. However, it is sad that suicide incidents by jumping off bridges occur from time to time. Of course, it may not be suicide by jumping off bridges, but it may be an accident caused by ignoring dangers during activities on the bridge.

[0003] In existing rescue scenarios, traditional rescue methods face many difficulties. When someone is found about to jump off a bridge or has already jumped off a bridge, rescue forces such as firefighters and police usually use air cushions as a buffer to try to reduce the impact of the jumper landing. However, this method has obvious limitations. On the one hand, the laying of air cushions is greatly restricted by site conditions. For example, in some narrow spaces under bridges and bridges above rivers with turbulent water, it is difficult to quickly and accurately complete the arrangement of air cushions. On the other hand, meteorological factors such as wind direction and wind force will also seriously affect the actual buffering effect of the air cushion. Once the position of the air cushion is offset by the wind, it is very likely that it will not be able to catch the jumper.

[0004] In view of this situation, a document with Chinese patent publication number CN 116999727 A discloses a bridge jumping protection device, which includes a protection net, an opening and closing mechanism for unfolding or closing the protection net, a launching mechanism, a plurality of infrared sensors and a controller; the opening and closing mechanism includes two protection net mounting frames, two hinged seats and two first motors, and the two hinged seats are fixed on the side walls; the protection net mounting frames are rotatably connected to the hinged seats, the protection net is fastened to the protection net mounting frames, and the first motor is installed on the bridge body and is transmission-connected to the protection net mounting frames; the launching mechanism includes two metal rods, two coils and two permanent magnets, a blind hole is opened on the side wall of the bridge body, the metal rods are embedded in the blind holes, and the permanent magnets and the metal rods are arranged correspondingly and fixed on the protection net mounting frames; a plurality of infrared sensors are installed on the guardrails of the bridge body. The invention claims that it can provide timely protection to bridge jumpers and reduce casualties.

[0005] After in-depth research, the applicant found that the above-mentioned bridge jumping protection device uses infrared sensors as triggering conditions, which has a great possibility of false triggering. On the one hand, it is some abnormal human activities (walking across the bridge), such as viewing high-risk actions; on the other hand, it is non-human triggering, such as birds, or other obstructions. In addition, since the protective net is attached to the protective net mounting frame, when a person falls into the protective net, if the impact force is large and there is a lack of appropriate buffer design, it is easy to cause the protective net to detach from the protective net mounting frame, posing a great safety hazard. Therefore, there is an urgent need for a bridge jumping prevention rescue system and emergency escape system that can not only ensure recognition accuracy but also provide better protection for falling personnel. Summary of the invention

[0006] The present invention provides an anti-jumping bridge rescue system and an emergency escape system, which can not only ensure recognition accuracy but also provide better protection for falling personnel.

[0007] In order to solve the above technical problems, this application provides the following technical solutions:

[0008] An anti-jump bridge rescue system, comprising:

[0009] An image module, the image module includes an image acquisition module and an image analysis module. The image acquisition module is installed on the outside of the bridge deck railing and is used to collect image information near the railing; the image analysis module is used to obtain a static image and a dynamic image based on the image information by comparing adjacent frames, and then identify the railing in the static image based on the continuity of pixels, and analyze whether the height of the upper edge of the dynamic image exceeds the railing by more than a preset value. If it exceeds the preset value, analyze whether the upper edge and the lower edge in the dynamic image are continuous. If they are continuous, a suspected bridge jumping signal is generated;

[0010] A control module, configured to receive a suspected bridge jumping signal, generate a first control signal, and then generate a second control signal after waiting for a preset time period;

[0011] At least two actuator motors, both fixedly mounted on the bottom of the bridge deck and used to perform an opening action after receiving a first control signal;

[0012] At least two rotating arms are rotatably connected to the bottom of the bridge deck, and the rotating arms are respectively driven by the execution motor to rotate, and when the execution motor executes the opening action, the rotating arms change from the initial state parallel to the bridge deck to the expanded state;

[0013] A protective net, the protective net is located between the two rotating arms, the inner sides of the two rotating arms are provided with a buckle for fixing the protective net, the buckle includes a base and a hook, the base is rotatably connected to the hook, a plurality of slide grooves are provided on the rotating arm, the base is slidably connected to the slide grooves, and a buffer fixedly connected to the base is provided in the slide groove on the base;

[0014] A sling motor is disposed above the protective net and fixedly connected to the side wall of the bridge deck. A connecting sling is provided between the sling motor and the protective net. The two ends of the connecting sling are respectively fixedly connected to the rotating shaft of the sling motor and the edge of the protective net. The sling motor is also used to rotate after receiving a second control signal so that the connecting sling slowly rises and then slowly descends.

[0015] The basic scheme principle and beneficial effects are as follows: The image acquisition module is precisely installed at a specific position outside the bridge railing, and continuously captures the scene near the railing with an appropriate acquisition frequency and viewing angle range to obtain the original image sequence. The image analysis module performs difference calculation on the pixel points of each frame based on the principle of adjacent frame comparison. Static images and dynamic images are distinguished by setting a static threshold. In static images, the railing outline is identified according to the pixel continuity rule, laying the foundation for the subsequent relative position judgment of dynamic objects and railings. For dynamic images, focus on the relationship between the upper edge of the object and the height of the railing. Once the upper edge exceeds the height of the railing and reaches the preset value, further check the continuity of the upper edge and the lower edge, and comprehensively judge to generate a suspected bridge jumping signal. This process simulates the capture of scene changes and the preliminary judgment logic of the human eye, and uses digital image processing technology to quickly and accurately screen potential bridge jumping risks.

[0016] For example, if the difference in brightness or color value of the same pixel in two adjacent frames is extremely small and within the static threshold range, the pixel is judged to be static, otherwise it is dynamic. By traversing the entire image, the stable railing pixels are classified as the static image part, and the moving human or object pixels form the dynamic image part. When the human head (upper edge) in the dynamic image passes a certain height of the railing and the body (upper and lower edges) presents a coherent shape, the conditions for generating a suspected bridge jumping signal are triggered.

[0017] As the center of the system, the control module receives the suspected bridge jumping signal from the image module in real time. Once received, the first control signal is immediately generated, which is like a starting command and quickly transmitted to at least two execution motors installed at the bottom of the bridge deck. The execution motor starts according to the command and drives the rotating arm connected to it. The rotating arm is designed to rotate and connect to the bottom of the bridge deck, so that it can quickly change from the initial hidden state parallel to the bridge deck to the open state under the power of the motor, providing a support structure for the subsequent deployment of the protective net. The whole process is completed efficiently in a short time to ensure the timeliness of rescue.

[0018] The protective net is placed between the two rotating arms and fixed by a specially designed buckle. The base of the buckle is connected to the hook by rotation, which can not only ensure the stable mounting of the protective net in daily non-emergency conditions, but also flexibly adjust the angle when under stress to prevent stress concentration and falling off. The slide groove and built-in buffer on the rotating arm are even more ingeniously designed. When an object hits the protective net, the buffer absorbs part of the impact force by its own elastic deformation. At the same time, the base can slide in the slide groove to further disperse the force and prevent the protective net from being overstressed at a single point.

[0019] For example, when someone falls onto the protective net, the impact force causes the hook of the buckle to rotate relative to the base, the buffer is compressed, and the base moves in the slide groove, just like the shock absorption system of a car, converting the instantaneous huge impact force into a relatively gentle buffering process, protecting the faller from direct hard collision injuries.

[0020] After receiving the second control signal from the control module, the sling motor starts, and slowly rotates the shaft by connecting the sling and the protective net, so that the connected sling first rises slowly and then slowly falls. On the one hand, this action can further buffer the impact of the faller, simulate the flexible traction during manual rescue, and avoid secondary injuries; on the other hand, by adjusting the height of the protective net, it is easier for subsequent rescuers to approach the faller, thereby improving the rescue efficiency.

[0021] The refined analysis process of the image module, combined with the railing recognition of static images and the multi-condition judgment of dynamic images, effectively distinguishes normal bridge crossing behavior, environmental interference (such as flying birds and debris) and real bridge jumping risks, greatly reducing the false alarm rate. Compared with traditional systems that rely only on a single sensor or simple visual judgment, it can capture potential dangers more accurately, buy precious time for rescue, and avoid ineffective waste of rescue resources.

[0022] The linkage design of the execution motor, rotating arm, protective net and sling motor realizes a quick response from the discovery of suspected bridge jumpers to the deployment of the protective net and buffer adjustment. A series of actions can be completed in a few seconds. Compared with traditional rescue methods such as air cushions, it does not require complicated site preparation and long-term layout, and can provide a protective barrier for bridge jumpers in the first place, greatly improving the success rate of rescue.

[0023] The buckle buffer design of the protective net and the lifting buffer of the cable motor form a multi-buffer system. Whether it is the moment of falling or the subsequent rescue process, it can minimize the impact damage to the body of the bridge jumper, reduce the risk of secondary injuries such as fractures and contusions, and protect the life and health of the bridge jumper.

[0024] In this solution, in terms of recognition accuracy, the image module relies on rigorous algorithms and multi-parameter judgments to accurately screen out the real risk of bridge jumping from massive image information. The static image recognition railing ensures the precise positioning of subsequent dynamic comparison. The multi-dimensional verification of dynamic images, such as height difference, edge continuity and other judgments, effectively avoids interference from environmental factors, and the recognition accuracy rate far exceeds the traditional methods. At the protection level, the protective net is closely coordinated with the rotating arm and the sling motor. The rotating arm quickly opens to provide stable support for the protective net. The buckle buffer structure of the protective net can instantly buffer the impact of falling and disperse the force. The lifting and lowering action of the sling motor can not only buffer but also adjust the height as needed to facilitate rescue by rescue personnel. The purpose of ensuring recognition accuracy and providing better protection for falling personnel is achieved.

[0025] Furthermore, the image acquisition module is used to obtain an original image sequence I(n) containing the railing and surrounding human activities, where n represents the sequence number of the image frame, n=1, 2, 3, ...; the image analysis module is used to obtain a static image I according to the image information by comparing adjacent frames. s (n) and dynamic image I d (n), which is implemented in the following way: suppose two adjacent frames of images I(n) and I(n+1), for each pixel point (x, y) in the image, calculate the absolute value of the pixel value difference ΔI(x, y, n) = |I(x, y, n) - I(x, y, n+1)|, if it is within the preset small area R, it satisfies:

[0026]

[0027] T 1 is the preset static threshold, then the pixels in the area are judged to be static in the current frame. By traversing the entire image, all the pixels judged to be static are combined into a static image I s (n), the remaining pixels form the dynamic image I d (n); where the size of region R is m×m pixels, and m is a positive integer;

[0028] Then in the static image I s In (n), the railing is identified according to the continuity of pixels. Let the pixel point (x i ,y i ) is a suspected railing pixel, where i = 1, 2, ..., N, N is the total number of pixels involved in railing recognition in the static image, and (x i ,y i ) on the horizontal x-axis. If the continuity condition is satisfied: for any adjacent suspected railing pixel point (x i ,y i ) and (x i+1 ,y i+1 ), there is |x i+1 -xi |≤Δx, and on the vertical y-axis, |y i+1 -y i |≤Δy, when the number of pixels that meet the above conditions exceeds the preset proportion P of the number of railing pixels 1 When , it is determined that the railing is identified; wherein Δx is the preset horizontal continuous pixel interval threshold, Δy is the preset vertical continuous pixel interval threshold, P 1 The value range is 0<P 1 <1;

[0029] Analyzing Dynamic Images I d (n) Whether the height of the upper edge of the railing exceeds the preset value, let the coordinate of the upper edge of the railing in the image be The coordinates of the upper edge of the object in the dynamic image are Calculate height difference If Δh>H preset , then analyze whether the upper edge and the lower edge in the dynamic image are continuous. Suppose the coordinates of the lower edge of the object in the dynamic image are Determine the continuity of the upper and lower edges in the vertical direction. If And in the horizontal direction for the upper edge point and the lower edge point have If it is continuous, a suspected bridge trip signal is generated; preset is the preset bridge jumping height threshold, Δh cont is the preset vertical continuity threshold of the upper and lower edges, Δx cont It is the preset upper and lower edge horizontal continuity threshold.

[0030] When distinguishing between static and dynamic images, the image is divided into preset small areas, the absolute value of the pixel value difference is calculated one by one, and compared with the precisely set static threshold. This method is like a fine sieve, filtering out pseudo-dynamic information caused by light and shadow changes, bridge micro-seismicity, etc., accurately analyzing the real static background, and accurately locating key fixed facilities such as railings, laying a solid foundation for subsequent accurate judgment.

[0031] The railing identification link uses the continuity of pixels in the horizontal and vertical directions, combined with an adapted continuous pixel interval threshold and a scientifically set ratio of the number of pixels involved in the judgment, to accurately outline the shape of the railing, eliminating the possibility of mistaking mottled light and shadow or bridge decorations for railings, and providing a precise reference for accurately capturing people's climbing behavior.

[0032] When analyzing dynamic images, a multi-dimensional judgment system is built, from the initial screening of the height of the upper edge of the object relative to the railing to the in-depth verification of the vertical and horizontal continuity of the upper and lower edges. Like a tight filter, it eliminates all interference factors such as flying birds and debris moving in the wind, and directly detects the actions of people who are at risk of jumping off the bridge. The recognition accuracy is far higher than those methods that rely solely on simple visual perception or rough image comparison.

[0033] The absolute value of the pixel difference is calculated and the entire image is traversed accordingly to form static and dynamic images. With the help of the optimized algorithm path, it is completed in a flash, and the time consumption is measured in milliseconds, which greatly reduces the time consumption of the initial image processing. The subsequent railing recognition and dynamic image hazard assessment are closely connected. Once a suspected bridge jumping signal is generated, it is immediately transmitted to the control module, triggering the rescue system to operate at full speed. Compared with traditional models with lengthy image analysis processes and repeated manual identification and verification, the response time of this system is shortened exponentially, which seizes the critical moment for rescue and makes life rescue one step faster.

[0034] Furthermore, the image analysis module is also used to generate a suspected bridge jumping signal for the dynamic image I d The object area in (n) that is judged to be likely to contact the railing and have a risk of jumping off the bridge is subjected to a feature extraction algorithm to extract its shape, texture and color features for verification and judgment. If the verification and judgment fail, the sending of the suspected bridge jumping signal is terminated; if the verification and judgment pass, the suspected bridge jumping signal is sent to the control module.

[0035] Furthermore, the verification judgment also includes: using human body posture estimation to judge the dynamic image sequence I d (n) Identify possible key points of the human body, analyze the relative positions and movement trajectories of these key points, and determine whether there are specific movements such as leaning forward or lifting the legs. Specific movements will cause characteristic changes in the distance and angle between key points. If changes that match the common posture and movement patterns of humans before jumping off a bridge are detected, and the number of continuous frames exceeds the preset frame number threshold F threshold , then the judgment passes the verification judgment.

[0036] Furthermore, the human body posture estimation and judgment further includes: after the human body posture estimation algorithm is used to identify possible human body key points, for each two adjacent frames of image I d (n) and I d (n+1), calculate the distance and angle changes between key points;

[0037] Assume that the set of key points of the human body is K = {k 1 ,k 2 ,…,k M}, where M is the number of key points. For key point k i and k j, where i≠j, and the coordinates in the nth frame image are (x i,n ,y i,n ) and (x j,n ,y j,n ), then the distance D between them ij,n The calculation formula is:

[0038]

[0039] The key point k between two adjacent frames i and k j The distance change ΔD ij，n for:

[0040] ΔD ij，n =|D ij，n+1 -D ij，n |

[0041] At the same time, the calculation consists of three key points k i , k j and k l The angle θ ijl，n , where k l Different from k i and k j Another key point is that according to the angle formula of the vector:

[0042] Let vector

[0043] vector

[0044] but:

[0045]

[0046] θ ijl，n =arccos(cosθ ijl，n )

[0047] The angle change between two adjacent frames Δθ ijl，n for:

[0048] Δθ ijl，n =|θ ijl，n+1 -θ ijl，n |;

[0049] Define the distance and angle change thresholds corresponding to common postures and action patterns before jumping off the bridge, and set the distance change threshold set as {ΔD ij，threshold}, the angle change threshold set is {Δθ ijl，threshold};

[0050] For each key point pair (i, j) and key point triple (i, j, l), if in the continuous N frames, ΔD is satisfiedij，n ≥ΔD ij，threshold And ΔD ij，n ≥ΔD ij，threshold The number of frames reaches N 1 (N 1 ≤N), then the change of the key point pair or triplet is considered to be consistent with the characteristics of the bridge jumping action;

[0051] Define a comprehensive bridge jumping risk score S risk , and its calculation formula is:

[0052]

[0053] Among them, ω ij and ω ijl are the weight coefficients of the key point pair (i, j) and the key point triple (i, j, l), satisfying These weight coefficients are pre-set according to the importance of different key points in the bridge jumping action;

[0054] χ ij and χ ijl is the indicator function. When the change of the key point pair (i, j) or the key point triple (i, j, l) meets the characteristics of the bridge jumping action, χ ij =1 or x ijl =1; otherwise, χ ij =0 or x ijl =0;

[0055] When the comprehensive bridge jumping risk score S risk Exceeding the preset risk score threshold S threshold , and the number of frames that meet the characteristics of the bridge jumping action exceeds the preset frame number threshold F continuously threshold When , the judgment is passed by verification.

[0056] Furthermore, it also includes a pressure detection module, which is fixed on the upper surface of the bridge deck railing. The pressure detection module includes a pressure sensitive unit and a covering member. An elastic member is fixed at the bottom of the covering member to support the covering member and leave an installation gap between the bridge deck railing. The pressure sensitive unit is fixed in the installation gap. The pressure sensitive unit is connected to the control module signal. The pressure sensitive unit is used to detect the pressure information exerted on the covering member and feed it back to the control module. The control module is also used to determine whether the increase in pressure information meets a preset threshold after receiving a suspected bridge jumping signal. If not, the next step of judgment is terminated; if yes, a first control signal is generated, and then a second control signal is generated after waiting for a preset period of time.

[0057] The pressure detection module is installed at a key position on the upper surface of the bridge railing, which can directly sense the changes in pressure applied by objects in contact with the railing. Its unique structural design uses elastic parts to support the cover and form an installation gap to accommodate the pressure sensitive unit, which not only ensures the sensitivity of pressure detection, but also protects the sensitive unit. When the control module receives a suspected bridge jumping signal, the introduction of pressure information becomes the key basis for secondary verification. For example, in windy weather, although the image module may generate a suspected bridge jumping signal due to debris hitting the railing, the pressure detection module can rely on stable and sub-threshold pressure growth data to assist the control module in accurately judging that it is not a real bridge jumping risk, and promptly terminate the subsequent unnecessary rescue action startup process, effectively avoiding the system's false startup due to environmental interference factors, and greatly reducing the ineffective waste of rescue resources.

[0058] Through the real-time signal connection with the control module, the pressure detection module adds a key link to the decision-making chain of the entire rescue system. In daily complex bridge traffic scenarios, such as pedestrians leaning on the railings for a short time and pets moving around the railings, the pressure detection module can continuously monitor the pressure dynamics. Once a real bridge jumping crisis occurs, the jumper will inevitably produce pressure changes that meet the preset threshold characteristics on the cover during the process of climbing and climbing over the railings. At this time, the pressure sensitive unit quickly captures and feedbacks information, prompting the control module to decisively generate the first control signal and start the rapid deployment process of the rescue device. Subsequently, the second control signal is generated after waiting for a preset time, and the actions of various components are coordinated in an orderly manner, so that the rescue response is both accurate and efficient, and the reliability and practicality of the system in real crisis scenarios are comprehensively improved.

[0059] Whether it is a hot summer day, the bridge deck may deform slightly due to thermal expansion and contraction due to high temperature, affecting pressure conduction; or a cold winter day, ice and snow accumulate around the railings and change the pressure distribution; or it is a complex working condition such as heavy rain and passing vehicles causing the bridge deck to vibrate, the accuracy of recognition is guaranteed by complementing the image module.

[0060] Furthermore, the control module is also used to start the pressure signal judgment process when the control module receives the suspected bridge jumping signal: at the initial time t 0 , record the current pressure value as P(t 0 ), and set an initial reference pressure P 0 =P(t 0 ), and at the same time open a time window ΔT to monitor the pressure change;

[0061] Within the time window ΔT, continuously collect the pressure value P(t) and calculate the pressure change rate The value range of t is t 0 <t≤t 0 +ΔT.

[0062] Furthermore, within the time window ΔT, the pressure change rate k satisfies k<K 1 , where K 1 is the preset bird stop pressure change rate threshold, and the pressure value fluctuation amplitude ΔP fluctuate Satisfy ΔP fluctuate <ΔP 1 , ΔP 1 It is the preset bird stay pressure fluctuation threshold. At this time, the control module determines that it is not a dangerous behavior of jumping off the bridge and terminates the subsequent rescue action triggering process.

[0063] Furthermore, within the time window ΔT, the pressure change rate k satisfies k<K 2 , where K 2 is the preset pedestrian stop pressure change rate threshold, and the maximum increment of the pressure value ΔP max Satisfy ΔP max <ΔP 2 , where ΔP 2 For the preset pedestrian stopping pressure increment threshold, the control module also determines it as a non-bridge jumping dangerous behavior and terminates the subsequent rescue action triggering process.

[0064] Furthermore, when the pressure change rate k ≥ K within the time window ΔT 3 , where K 3 is the preset bridge jumping danger pressure change rate threshold, and the maximum pressure value increment ΔP max ≥ΔP 3 When, ΔP 3 When the pressure increment threshold of the bridge jumping danger is preset, the control module determines that the pressure signal characteristic of the bridge jumping danger is met, and then generates a first control signal, and then generates a second control signal after waiting for a preset time.

[0065] After receiving the suspected bridge jumping signal, by setting the initial time t 0 , initial reference pressure P 0 And a specific time window ΔT, to carry out all-round monitoring of pressure changes. Taking the bird stop scene as an example, the system uses the preset bird stop pressure change rate threshold K 1 and pressure fluctuation amplitude threshold ΔP 1 When a bird briefly perches on the railing cover, its light weight will only cause extremely small and short-term pressure changes, and the pressure change rate \(k\) is stably lower than K 1 , pressure fluctuation range ΔP fluctuate It is also much smaller than ΔP 1 The control module quickly determines that it is a non-dangerous behavior and decisively terminates the subsequent rescue process, avoiding frequent triggering of rescue devices due to such common natural phenomena, making the system's misjudgment risk close to zero.

[0066] For daily high-frequency scenes such as pedestrians stopping to appreciate the scenery, the control module uses the pressure change rate threshold K 2 and pedestrian stop pressure increment threshold ΔP 2 Realize intelligent differentiation. When pedestrians lean against the railing and stand for a long time, the pressure changes slowly and the increment is limited. The system monitors the pressure change rate k in the time window ΔT to satisfy k<K 2 , the maximum pressure increment ΔP max Less than ΔP 2 , accurately identify this as normal traffic behavior, terminate the false alarm process in time, and ensure that rescue resources are always on standby to respond to real crises. When facing the real risk of jumping off the bridge, the violent climbing and crossing actions of the bridge jumper will cause the pressure to rise sharply in a short period of time. Once the pressure change rate k≥K 3 And the maximum pressure increment ΔP max ≥ΔP 3 The control module immediately determines that the pressure signal characteristics meet the danger of jumping off the bridge, and generates the first control signal at lightning speed to start the rescue. Subsequently, the second control signal is accurately dispatched according to the preset duration to ensure that the rescue operation is time-sensitive and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a structural schematic diagram of an anti-jump bridge rescue system and an emergency escape system embodiment 1. DETAILED DESCRIPTION

[0068] The following is further described in detail through specific implementation methods:

[0069] The symbols in the drawings of the specification include: rotating arm 1, protective net 2, and sling motor 3.

[0070] Example 1

[0071] An anti-jump bridge rescue system, comprising:

[0072] An image module, the image module includes an image acquisition module and an image analysis module. The image acquisition module is installed on the outside of the bridge deck railing and is used to collect image information near the railing; the image analysis module is used to obtain a static image and a dynamic image based on the image information by comparing adjacent frames, and then identify the railing in the static image based on the continuity of pixels, and analyze whether the height of the upper edge of the dynamic image exceeds the railing by more than a preset value. If it exceeds the preset value, analyze whether the upper edge and the lower edge in the dynamic image are continuous. If they are continuous, a suspected bridge jumping signal is generated;

[0073] A control module, configured to receive a suspected bridge jumping signal, generate a first control signal, and then generate a second control signal after waiting for a preset time period;

[0074] Two actuator motors are fixedly mounted on the bottom of the bridge deck and are used to perform an opening action after receiving a first control signal;

[0075] The two rotating arms 1 are both rotatably connected to the bottom of the bridge deck. The rotating arms 1 are respectively driven by the execution motors to rotate, and when the execution motors execute the opening action, the rotating arms 1 change from the initial state parallel to the bridge deck to the expanded state;

[0076] A protective net 2, the protective net 2 is located between the two rotating arms 1, and a buckle for fixing the protective net 2 is provided on the inner side of the two rotating arms 1, the buckle includes a base and a hook, the base is rotatably connected to the hook, a plurality of slide grooves are provided on the rotating arm 1, the base is slidably connected to the slide grooves, and a buffer fixedly connected to the base is provided in the slide groove on the base;

[0077] A sling motor 3 is provided above the protective net 2 and fixedly connected to the side wall of the bridge deck. A connecting sling is provided between the sling motor 3 and the protective net 2. The two ends of the connecting sling are respectively fixedly connected to the rotating shaft of the sling motor 3 and the edge of the protective net 2. The sling motor 3 is also used to rotate after receiving a second control signal so that the connecting sling slowly rises and then slowly descends.

[0078] The image acquisition module is used to obtain the original image sequence I(n) containing the railing and the surrounding human activities, where n represents the sequence number of the image frame, n=1, 2, 3, ...; the image analysis module is used to compare the image information with adjacent frames to obtain a static image I s (n) and dynamic image I d (n), which is implemented in the following way: suppose two adjacent frames of images I(n) and I(n+1), for each pixel point (x, y) in the image, calculate the absolute value of the pixel value difference ΔI(x, y, n) = |I(x, y, n) - I(x, y, n+1)|, if it is within the preset small area R, it satisfies:

[0079]

[0080] T 1 is the preset static threshold, then the pixels in the area are judged to be static in the current frame. By traversing the entire image, all the pixels judged to be static are combined into a static image I s (n), the remaining pixels form the dynamic image I d (n); where the size of region R is m×m pixels, and m is a positive integer;

[0081] Then in the static image I s In (n), the railing is identified according to the continuity of pixels. Let the pixel point (x i ,y i) is a suspected railing pixel, where i = 1, 2, ..., N, N is the total number of pixels involved in railing recognition in the static image, and (x i ,y i ) on the horizontal x-axis. If the continuity condition is satisfied: for any adjacent suspected railing pixel point (x i ,y i ) and (x i+1 ,y i+1 ), there is |x i+1 -x i |<Δx, and on the vertical y-axis, |y i+1 -y i |≤Δy, when the number of pixels that meet the above conditions exceeds the preset proportion P of the number of railing pixels 1 When , it is determined that the railing is identified; wherein Δx is the preset horizontal continuous pixel interval threshold, Δy is the preset vertical continuous pixel interval threshold, P 1 The value range is 0<P 1 <1;

[0082] Analyzing Dynamic Images I d (n) Whether the height of the upper edge of the railing exceeds the preset value, let the coordinate of the upper edge of the railing in the image be The coordinates of the upper edge of the object in the dynamic image are Calculate height difference If Δh>H preset , then analyze whether the upper edge and the lower edge in the dynamic image are continuous. Suppose the coordinates of the lower edge of the object in the dynamic image are Determine the continuity of the upper and lower edges in the vertical direction. If And in the horizontal direction for the upper edge point and the lower edge point have If it is continuous, a suspected bridge trip signal is generated; preset is the preset bridge jumping height threshold, Δh cont is the preset vertical continuity threshold of the upper and lower edges, Δx cont It is the preset upper and lower edge horizontal continuity threshold.

[0083] Specific use: First, accurately install the image acquisition module on the outside of the bridge railing. Determine its installation location based on factors such as the type and height of the bridge and the flow of people passing by, so as to obtain clear and comprehensive image information.

[0084] The control module uses a high-performance programmable controller and is installed in a control box near the bridge deck to ensure stable communication with the image module, pressure detection module (described later), execution motor and other components. After receiving the suspected bridge jumping signal sent by the image module, the built-in control program is immediately run to generate a first control signal, and a second control signal is generated after waiting for a preset time. The preset time can be adjusted and optimized according to actual factors such as the height of the bridge and the time required for the deployment of the protective net 2 to ensure that the actions of each component are coordinated and orderly.

[0085] Both actuator motors are firmly fixed at the designated positions at the bottom of the bridge deck, ensuring their installation angle and position accuracy so that they can smoothly drive the rotating arm 1. The actuator motors are selected to have high torque and fast response characteristics to meet the needs of quickly executing the opening action in an emergency.

[0086] The rotating arm 1 is made of high-strength, corrosion-resistant metal material, one end of which is rotatably connected to the bottom of the bridge deck through a precision bearing, and the other end is adapted to the buckle structure of the protection net 2. When the execution motor does not receive a signal, the rotating arm 1 is in the initial state of being parallel to the bridge deck, minimizing the occupation and visual impact on the bridge deck passage space; when the first control signal is received, the execution motor rotates rapidly, driving the rotating arm 1 to change to the expanded state in a short time, providing reliable support for the deployment of the protection net 2.

[0087] The protective net 2 is woven from a fiber material that is high-strength, impact-resistant, and has a certain degree of flexibility. The size of the mesh is carefully designed to effectively block falling personnel and reduce wind resistance. The protective net 2 is installed between the two rotating arms 1 and fixed by a specially designed buckle. The base of the buckle is rotatably connected to the hook and is made of wear-resistant, high-strength engineering plastic to ensure that it can rotate flexibly under repeated stress. Several slide grooves are precisely machined on the rotating arm 1, and the base is slidably connected to the slide groove. A buffer fixedly connected to the base is installed in the slide groove. The buffer can be made of high-performance rubber or spring devices. When the protective net 2 is impacted, the buffer can be quickly compressed or stretched to absorb part of the impact force and protect the protective net 2 and the rotating arm 1 structure. At the same time, the base slides in the slide groove to disperse the force and avoid stress concentration.

[0088] The cable motor 3 is a motor with precise speed regulation function, which is installed above the protective net 2 and fixedly connected to the side wall of the bridge deck through a sturdy bracket. A high-strength connecting cable is connected between the cable motor 3 and the protective net 2, and the two ends of the connecting cable are respectively fixedly connected to the rotating shaft of the cable motor 3 and the edge of the protective net 2 through a reliable connection method (such as a metal buckle, etc.). After receiving the second control signal, the cable motor 3 rotates according to the preset speed curve, so that the connecting cable slowly rises and then slowly descends, realizing dynamic adjustment of the height of the protective net 2, further buffering the impact force of the falling person, and facilitating subsequent rescue personnel to approach the faller.

[0089] The image analysis module is also used to generate a suspected bridge jumping signal and analyze the dynamic image I d The object area in (n) that is judged to be likely to contact the railing and have a risk of jumping off the bridge is subjected to a feature extraction algorithm to extract its shape, texture and color features for verification and judgment. If the verification and judgment fail, the sending of the suspected bridge jumping signal is terminated; if the verification and judgment pass, the suspected bridge jumping signal is sent to the control module.

[0090] Specific use: In terms of shape features, calculate the contour perimeter C, area S and aspect ratio AR of the object, where The human body has a relatively specific shape and proportion. Generally, the length-to-width ratio of an adult's body is within a certain range. human-min To AR human-max between.

[0091] In terms of texture features, the gray-level co-occurrence matrix is ​​used to calculate texture parameters such as contrast, correlation, energy, and homogeneity. Human skin and clothing have specific texture features. For example, skin texture is relatively delicate, while clothing texture has its own characteristics due to different materials. These texture parameters will be in a specific value range.

[0092] In terms of color features, the color distribution in the object area is counted and the mean μ of the main colors is calculated. color and standard deviation σ color . Human skin color has a certain color range among different races, and clothing colors are diverse.

[0093] Verification judgment also includes: using human posture estimation judgment to d (n) Identify possible key points of the human body, analyze the relative positions and movement trajectories of these key points, and determine whether there are specific movements such as leaning forward or lifting the legs. Specific movements will cause characteristic changes in the distance and angle between key points. If changes that match the common posture and movement patterns of humans before jumping off a bridge are detected, and the number of continuous frames exceeds the preset frame number threshold F threshold, then the judgment passes the verification judgment.

[0094] The human body posture estimation judgment also includes: after the human body posture estimation algorithm is used to identify possible human body key points, for each two adjacent frames of image I d (n) and I d (n+1), calculate the distance and angle changes between key points;

[0095] Assume that the set of key points of the human body is K = {k 1 , k 2 , …, k M}, where M is the number of key points. For key point k i and k j , where i≠j, and the coordinates in the nth frame image are (x i，n ,y i，n ) and (x j，n ,y j，n ), then the distance D between them ij，n The calculation formula is:

[0096]

[0097] The key point k between two adjacent frames i and k j The distance change ΔD ij，n for:

[0098] ΔD ij，n =|D ij，n+1 -D ij，n |

[0099] At the same time, the calculation consists of three key points k i , k j and k l The angle θ ijl，n , where k l Different from k i and k j Another key point is that according to the angle formula of the vector:

[0100] Let vector

[0101] vector

[0102] but:

[0103]

[0104] θ ijl，n =arccos(cosθ ijl，n )

[0105] The angle change between two adjacent frames Δθ ijl，n for:

[0106] Δθ ijl,n =|θ ijl,n+1 -θ ijl,n |;

[0107] Define the distance and angle change thresholds corresponding to common postures and action patterns before jumping off the bridge, and set the distance change threshold set as {ΔD ij，threshold}, the angle change threshold set is {Δθ ijl,threshold};

[0108] For each key point pair (i, j) and key point triple (i, j, l), if in the continuous N frames, ΔD is satisfied ij，n ≥ΔD ij，threshold And Δθ ijl，n ≥Δθ ijl，threshold The number of frames reaches N 1 (N 1 ≤N), then the change of the key point pair or triplet is considered to be consistent with the characteristics of the bridge jumping action;

[0109] Define a comprehensive bridge jumping risk score S risk , and its calculation formula is:

[0110]

[0111] Among them, ω ij and ω ijl are the weight coefficients of the key point pair (i, j) and the key point triple (i, j, l), satisfying These weight coefficients are pre-set according to the importance of different key points in the bridge jumping action;

[0112] χ ij and χ ijl is the indicator function. When the change of the key point pair (i, j) or the key point triple (i, j, l) meets the characteristics of the bridge jumping action, χ ij =1 or x ijl =1; otherwise, χ ij =0 or x ijl =0;

[0113] When the comprehensive bridge jumping risk score S risk Exceeding the preset risk score threshold S threshold , and the number of frames that meet the characteristics of the bridge jumping action exceeds the preset frame number threshold F continuously threshold When , the judgment is passed by verification.

[0114] Specific usage: In human body images at different angles and in different outfits, the algorithm accurately marks the two-dimensional coordinate position of each key point based on the unique geometric shape and motion continuity characteristics of the human joints, and constructs a preliminary model framework of the human posture.

[0115] Once the key points of the human body are identified, their relative positions are tracked and calculated in real time. In continuous image frames, by comparing the coordinate changes of the same key point between adjacent frames, the displacement vector on the plane is obtained, and then the motion trajectory of each key point is depicted. This dynamic analysis based on time series can capture the subtle posture adjustments of the human body and restore the real-time motion state of the human body.

[0116] Take a person trying to jump off a bridge as an example. From standing close to the railing to preparing to climb over, the vertical position of the shoulder key point relative to the hip key point will gradually increase. At the same time, the movement trajectory of the knee key point will show an obvious displacement toward the railing. These change information is accurately captured and recorded by the system.

[0117] A built-in bridge jumping posture feature library trained with a large number of real bridge jumping cases and simulation experiment data can be built in, which contains the key point distance and angle change patterns corresponding to typical pre-bridge jumping preparation actions such as leaning forward and lifting legs. When analyzing the distance and angle changes of the current human key points and comparing them with the standard patterns in the feature library, if within a certain frame number range, it is detected that these changes are consistent with the common posture and action patterns of humans before bridge jumping, and the continuous frame number exceeds the preset frame number threshold F threshold , it is determined that the current person is in a high-risk state of jumping off the bridge, and the subsequent rescue process is triggered through verification and judgment.

[0118] For example, leaning forward will rapidly reduce the angle between the line connecting the head key point and the shoulder key point and the vertical direction, while the distance between the hip key point and the ankle key point will lengthen. The system can accurately identify the risk of bridge jumping by monitoring the changes in these characteristic parameters in real time.

[0119] Compared with traditional methods that rely only on simple human contour recognition or single action detection, this verification and judgment based on human posture estimation can go deep into the details of human actions. Through multi-key point collaborative analysis, it accurately captures those subtle and iconic preparations for jumping off the bridge, greatly improving the recognition accuracy of the real intention of jumping off the bridge, effectively avoiding misjudging normal stretching, bending and other behaviors as dangerous actions, and reducing the false alarm rate.

[0120] Because it can track the continuous changes in human posture in real time, the system can detect danger signals in advance when a person has just shown a tendency to jump off the bridge, or even before the body has completely crossed the railing. Compared with traditional rescue systems that often start responding only when the person is already in the process of falling, this system gains valuable advance intervention time, allowing rescue operations to be intervened earlier, greatly improving the success rate of rescue.

[0121] Whether it is strong light during the day, weak light at night, or in adverse weather conditions such as wind and rain, as long as the image acquisition module can obtain relatively clear images, the human posture estimation module can accurately identify key points of the human body and analyze posture changes with its powerful algorithm adaptability. It is not affected too much by differences in clothing and body shape, and operates stably in various complex real-life scenarios, ensuring that the rescue system is ready at any time and plays its role accurately.

[0122] Example 2

[0123] Compared with Example 1, the only difference is that it also includes a pressure detection module, which is fixed on the upper surface of the bridge deck railing. The pressure detection module includes a pressure sensitive unit and a covering member. An elastic member is fixed at the bottom of the covering member to support the covering member and the bridge deck railing to leave an installation gap. The pressure sensitive unit is fixed in the installation gap. The pressure sensitive unit is connected to the control module signal. The pressure sensitive unit is used to detect the pressure information exerted on the covering member and feed it back to the control module. The control module is also used to determine whether the increase in pressure information meets a preset threshold after receiving a suspected bridge jumping signal. If not, the next step of judgment is terminated; if yes, a first control signal is generated, and then a second control signal is generated after waiting for a preset period of time.

[0124] The control module is also used to start the pressure signal judgment process when the control module receives a suspected bridge jumping signal: at the initial time t 0 , record the current pressure value as P(t 0 ), and set an initial reference pressure P 0 =P(t 0 ), and at the same time open a time window ΔT to monitor the pressure change;

[0125] Within the time window ΔT, continuously collect the pressure value P(t) and calculate the pressure change rate The value range of t is t 0 <t<t 0 +ΔT.

[0126] In the time window ΔT, the pressure change rate k satisfies k<K 1 , where K 1 is the preset bird stop pressure change rate threshold, and the pressure value fluctuation amplitude ΔP fluctuateSatisfy ΔP fluctuate <ΔP 1 , ΔP 1 It is the preset bird stay pressure fluctuation threshold. At this time, the control module determines that it is not a dangerous behavior of jumping off the bridge and terminates the subsequent rescue action triggering process.

[0127] In the time window ΔT, the pressure change rate k satisfies k<K 2 , where K 2 is the preset pedestrian stop pressure change rate threshold, and the maximum increment of the pressure value ΔP max Satisfy ΔP max <ΔP 2 , where ΔP 2 For the preset pedestrian stopping pressure increment threshold, the control module also determines it as a non-bridge jumping dangerous behavior and terminates the subsequent rescue action triggering process.

[0128] When within the time window ΔT, the pressure change rate k ≥ K 3 , where K 3 is the preset bridge jumping danger pressure change rate threshold, and the maximum pressure value increment ΔP max ≥ΔP 3 When, ΔP 3 When the pressure increment threshold of the bridge jumping danger is preset, the control module determines that the pressure signal characteristic of the bridge jumping danger is met, and then generates a first control signal, and then generates a second control signal after waiting for a preset time.

[0129] When used specifically: the pressure detection module is fixed on the upper surface of the bridge railing. The pressure detection module includes a pressure sensitive unit and a cover. The cover is made of anti-slip and weather-resistant materials. An elastic part is fixed at the bottom to support the cover and the bridge railing to leave an installation gap. The elastic part can be made of silicone or spring, which can ensure that the pressure sensitive unit is in a suitable detection position and can adapt to slight deformations of the bridge deck. The pressure sensitive unit uses a high-precision piezoresistive sensor or a piezoelectric sensor, which is fixed in the installation gap to ensure that the pressure information of the cover can be accurately detected, and is connected to the control module signal through a shielded cable.

[0130] After completing the installation and preliminary debugging of each component, the overall joint debugging of the system is carried out. Various possible scenarios are simulated, including normal pedestrian traffic, birds flying over, debris hitting the railings, and people simulating jumping off the bridge, etc., to comprehensively monitor the response of the system. Through professional monitoring equipment, key performance indicators such as the recognition accuracy and response time of the image module, the signal transmission delay of the control module, the action speed of the execution motor, the deployment effect of the protective net 2, and the judgment accuracy of the pressure detection module are recorded.

[0131] According to the results of the joint debugging, the various parameters of the system are optimized and adjusted. For example, if it is found that the recognition accuracy of the image module decreases under strong light, the threshold parameters for the comparison of adjacent frames can be appropriately adjusted or the pixel value calculation algorithm can be optimized; if the action speed of the execution motor is slow, the motor power supply voltage, the friction of the transmission mechanism and other factors can be checked and corresponding improvements can be made; for the pressure detection module, if there are many misjudgments, the pressure change threshold in different scenarios can be further fine-tuned. Through repeated testing and optimization, it is ensured that the system can operate stably and efficiently in various complex environments, and achieve the design goal of accurately identifying the risk of jumping off the bridge and timely and effective rescue.

[0132] The above are only embodiments of the present invention. The invention is not limited to the fields involved in this implementation case. The common sense such as the known specific structures and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to explain the content of the claims.

Claims

1. A bridge rescue system, characterized in that: include: An image module, the image module includes an image acquisition module and an image analysis module. The image acquisition module is installed on the outside of the bridge deck railing and is used to collect image information near the railing; the image analysis module is used to obtain a static image and a dynamic image based on the image information by comparing adjacent frames, and then identify the railing in the static image based on the continuity of pixels, and analyze whether the height of the upper edge of the dynamic image exceeds the railing by more than a preset value. If it exceeds the preset value, analyze whether the upper edge and the lower edge in the dynamic image are continuous. If they are continuous, a suspected bridge jumping signal is generated; A control module, configured to receive a suspected bridge jumping signal, generate a first control signal, and then generate a second control signal after waiting for a preset time period; At least two actuator motors, both fixedly mounted on the bottom of the bridge deck and used to perform an opening action after receiving a first control signal; At least two rotating arms are rotatably connected to the bottom of the bridge deck, and the rotating arms are respectively driven by the execution motor to rotate, and when the execution motor executes the opening action, the rotating arms change from the initial state parallel to the bridge deck to the expanded state; A protective net, the protective net is located between the two rotating arms, the inner sides of the two rotating arms are provided with a buckle for fixing the protective net, the buckle includes a base and a hook, the base is rotatably connected to the hook, a plurality of slide grooves are provided on the rotating arm, the base is slidably connected to the slide grooves, and a buffer fixedly connected to the base is provided in the slide groove on the base; A sling motor is disposed above the protective net and fixedly connected to the side wall of the bridge deck. A connecting sling is provided between the sling motor and the protective net. The two ends of the connecting sling are respectively fixedly connected to the rotating shaft of the sling motor and the edge of the protective net. The sling motor is also used to rotate after receiving a second control signal so that the connecting sling slowly rises and then slowly descends.

2. The anti-jump bridge rescue system according to claim 1, characterized in that: The image acquisition module is used to obtain an original image sequence I(n) containing the railing and surrounding human activities, where n represents the sequence number of the image frame, n=1, 2, 3, ...; the image analysis module is used to compare the image information with adjacent frames to obtain a static image I s (n) and dynamic image I d (n), which is implemented in the following way: suppose two adjacent frames of images I(n) and I(n+1), for each pixel point (x, y) in the image, calculate the absolute value of the pixel value difference ΔI(x, y, n) = |I(x, y, n) - I(x, y, n+1)|, if it is within the preset small area R, it satisfies: T1 is the preset static threshold, and the pixels in the area are judged to be static in the current frame. By traversing the entire image, all the pixels judged to be static are combined into a static image I s (n), the remaining pixels form the dynamic image I d (n); where the size of region R is m×m pixels, and m is a positive integer; Then in the static image I s In (n), the railing is identified according to the continuity of pixels. Let the pixel point (x i ,y i ) is a suspected railing pixel, where i = 1, 2, ..., N, N is the total number of pixels involved in railing recognition in the static image, and (x i ,y i ) on the horizontal x-axis. If the continuity condition is satisfied: for any adjacent suspected railing pixel point (x i ,y i ) and (x i+1 ,y i+1 ), there is |x i+1 -x i |≤Δx, and on the vertical y-axis, |y i+1 -y i |≤Δy, when the number of pixels satisfying the above conditions exceeds the preset ratio P1 of the number of pixels of the railing, it is determined that the railing is identified; wherein Δx is the preset horizontal continuous pixel interval threshold, Δy is the preset vertical continuous pixel interval threshold, and the value range of P1 is 0<P1<1; Analyzing Dynamic Images I d (n) Whether the height of the upper edge of the railing exceeds the preset value, let the coordinate of the upper edge of the railing in the image be The coordinates of the upper edge of the object in the dynamic image are Calculate height difference If Δh>Hp reset , then analyze whether the upper edge and the lower edge in the dynamic image are continuous. Suppose the coordinates of the lower edge of the object in the dynamic image are Determine the continuity of the upper and lower edges in the vertical direction. If And in the horizontal direction for the upper edge point and the lower edge point have If it is continuous, a suspected bridge trip signal is generated; preset is the preset bridge jumping height threshold, Δh cont is the preset vertical continuity threshold of the upper and lower edges, Δx cont It is the preset upper and lower edge horizontal continuity threshold.

3. The anti-jump bridge rescue system according to claim 2, characterized in that: The image analysis module is also used to generate a suspected bridge jumping signal, for the dynamic image I d The object area in (n) that is judged to be likely to contact the railing and have a risk of jumping off the bridge is subjected to a feature extraction algorithm to extract its shape, texture and color features for verification and judgment. If the verification and judgment fail, the sending of the suspected bridge jumping signal is terminated; if the verification and judgment pass, the suspected bridge jumping signal is sent to the control module.

4. The anti-jump bridge rescue system according to claim 3, characterized in that: The verification judgment also includes: using human body posture estimation to judge the dynamic image sequence I d (n) Identify possible key points of the human body, analyze the relative positions and movement trajectories of these key points, and determine whether there are specific movements such as leaning forward or lifting the legs. Specific movements will cause characteristic changes in the distance and angle between key points. If changes that match the common posture and movement patterns of humans before jumping off a bridge are detected, and the number of continuous frames exceeds the preset frame number threshold F threshold , then the judgment passes the verification judgment.

5. The anti-jump bridge rescue system according to claim 4, characterized in that: The human body posture estimation and judgment further includes: after the human body posture estimation algorithm is used to identify possible human body key points, for each two adjacent frames of image I d (n) and I d (n+1), calculate the distance and angle changes between key points; Assume that the set of key points of the human body is K = {k1, k2, ..., k M }, where M is the number of key points. For key point k i and k j , where i≠j, and the coordinates in the nth frame image are (x i，n ,y i，n ) and (x j，n ,y j，n ), then the distance D between them ij，n The calculation formula is: The key point k between two adjacent frames i and k j The distance change ΔD ij，n for: ΔD ij，n =|D ij，n+1 -D ij，n | At the same time, the calculation consists of three key points k i , k j and k l The angle θ ijl，n , where k l Different from k i and k j Another key point is that according to the angle formula of the vector: Let vector vector but: i ijl，n =arccos(cosθ) ijl，n ) The angle change between two adjacent frames Δθ ijl，n for: Dth ijl，n =|θ ijl，n+1 -θ ijl，n |; Define the distance and angle change thresholds corresponding to common postures and action patterns before jumping off the bridge, and set the distance change threshold set as {ΔD ij，threshold }, the angle change threshold set is {Δθi jl，threshold }; For each key point pair (i, j) and key point triple (i, j, l), if in the continuous N frames, ΔD is satisfied ij，n ≥ΔD ij，threshold And Δθ ijl，n ≥Δθ ijl，threshold The number of frames reaches N1 ( N1≤N), then the change of the key point pair or triplet is considered to be consistent with the characteristics of the bridge jumping action; Define a comprehensive bridge jumping risk score S risk , and its calculation formula is: Among them, ω ij and ω ijl are the weight coefficients of the key point pair (i, j) and the key point triple (i, j, l), satisfying These weight coefficients are pre-set according to the importance of different key points in the bridge jumping action; χ ij and χ ijl is the indicator function. When the change of the key point pair (i, j) or the key point triple (i, j, l) meets the characteristics of the bridge jumping action, χ ij =1 or x ijl =1; otherwise, χ ij =0 or x ijl =0; When the comprehensive bridge jumping risk score S ris k exceeds the preset risk score threshold S threshold , and the number of frames that meet the characteristics of the bridge jumping action exceeds the preset frame number threshold F continuously threshold When , the judgment is passed by verification.

6. The anti-jump bridge rescue system according to claim 5, characterized in that: It also includes a pressure detection module, which is fixed on the upper surface of the bridge deck railing. The pressure detection module includes a pressure sensitive unit and a covering member. An elastic member is fixed at the bottom of the covering member to support the covering member and the bridge deck railing to leave an installation gap. The pressure sensitive unit is fixed in the installation gap. The pressure sensitive unit is connected to the control module signal. The pressure sensitive unit is used to detect the pressure information exerted on the covering member and feed it back to the control module. The control module is also used to determine whether the increase in pressure information meets a preset threshold after receiving a suspected bridge jumping signal. If not, the next step of judgment is terminated; if yes, a first control signal is generated, and then a second control signal is generated after waiting for a preset period of time.

7. The anti-jump bridge rescue system according to claim 6, characterized in that: The control module is also used to start the pressure signal judgment process after receiving the suspected bridge jumping signal: at the initial time t0, record the current pressure value as P(t0), set an initial reference pressure P0=P(t0), and open a time window ΔT to monitor the pressure change; Within the time window ΔT, continuously collect the pressure value P(t) and calculate the pressure change rate The value range of t is t0<t≤t0+ΔT.

8. The anti-jump bridge rescue system according to claim 7, characterized in that: In the time window ΔT, the pressure change rate k satisfies k<K1, where K1 is the preset bird-stay pressure change rate threshold, and the pressure value fluctuation amplitude ΔP fluctuate Satisfy ΔP fluctuate <ΔP1, ΔP1 is the preset bird stay pressure fluctuation threshold. At this time, the control module determines that it is not a dangerous behavior of jumping off the bridge and terminates the subsequent rescue action triggering process.

9. The anti-jump bridge rescue system according to claim 8, characterized in that: In the time window ΔT, the pressure change rate k satisfies k<K2, where K2 is the preset pedestrian stop pressure change rate threshold, and the maximum increment of the pressure value ΔP max Satisfy ΔP max <ΔP2, where ΔP2 is the preset pedestrian stopping pressure increment threshold. The control module also determines it as a non-bridge jumping dangerous behavior and terminates the subsequent rescue action triggering process.

10. The anti-jump bridge rescue system according to claim 9, characterized in that: When within the time window ΔT, the pressure change rate k ≥ K3, where K3 is the preset bridge jumping risk pressure change rate threshold, and the maximum pressure value increment ΔP max ≥ΔP3, where ΔP3 is the preset bridge jumping danger pressure increment threshold, the control module determines that it meets the bridge jumping danger pressure signal characteristics, and then generates a first control signal, and then generates a second control signal after waiting for a preset time.

Images