Platform area intrusion intelligent detection method based on pressure sensing and vision fusion

Abstract

The invention discloses a platform area intrusion intelligent detection method based on pressure sensing and vision fusion. The method comprises the following steps: synchronously acquiring video data of a platform monitoring area and track pressure sensor data; preprocessing the video data and identifying a potential intrusion area and an intrusion state thereof by using a target detection algorithm; meanwhile, the rail pressure value is filtered, and the train state is judged according to a preset pressure threshold value; a decision module is used for comprehensive judgment; only when the train is in a moving dangerous state and an invasion state is detected by a video, dangerous invasion is judged, and sound-light alarm and remote alarm are triggered; and a safety signal is output under other conditions. According to the invention, by fusing the vision and pressure sensing information, the false alarm rate of single sensor detection is effectively reduced, and the accuracy and reliability of judging the dangerous intrusion event of the platform are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing and public safety, and in particular to an intelligent intrusion detection method for platform areas based on pressure sensing and visual fusion. Background Technology

[0002] Platform area intrusion detection is a core component of public transportation safety management. Its core challenge lies in accurately distinguishing between "genuine dangerous intrusions" and "normal operational behaviors" (such as passengers boarding and alighting when the train stops). Existing technical solutions all have significant shortcomings in addressing this critical issue.

[0003] Existing technologies mainly use traditional motion detection algorithms such as background subtraction, optical flow, and frame difference, as well as edge detection operators like Canny. First, existing visual algorithms can only "see" objects moving or crossing boundaries in the image, but they cannot "understand" the operational context of the platform at that time, i.e., they cannot determine whether the train is stopped or moving. They may misinterpret normal passenger boarding and alighting behavior when the train is stopped as dangerous intrusion, thus generating a large number of troublesome false alarms and causing the "boy who cried wolf" effect in the system. Second, complex changes in lighting, weather conditions (rain and snow), and large-scale background movement and shadows brought about by trains entering and leaving the station can seriously interfere with the stability of visual algorithms, further increasing the false alarm and false negative rates. In addition, although other sensors can reliably determine whether the train is stopped, they cannot detect whether there are people intruding into dangerous areas. Even if video surveillance and vehicle sensing systems are deployed simultaneously, they are often independent of each other, and the data is not deeply integrated and used for intelligent decision-making. Summary of the Invention

[0004] This invention provides an intelligent intrusion detection method for platform areas based on the fusion of pressure sensing and vision, in order to overcome the problems of false alarms caused by the erroneous judgment of pure vision algorithms and the independent monitoring systems.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: A smart intrusion detection method for platform areas based on pressure sensing and vision fusion includes: S1. Collect video data of the platform monitoring area through an image sensor, and simultaneously collect track pressure values ​​through a track pressure sensor; S2. Preprocess the video data of the platform monitoring area and obtain the visual detection confidence level through the target detection algorithm; obtain the intrusion status of the potential intrusion area based on the visual detection confidence level and the preset minimum alarm threshold; the intrusion status of the potential intrusion area includes intrusion status and non-intrusion status. S3. Filter and denoise the track pressure value to obtain the filtered pressure value; determine the train status based on the filtered pressure value and the preset pressure threshold; the train status includes a safe stopping status and a dangerous moving status. S4. Based on the train status and the intrusion status of the potential intrusion area, obtain the alarm decision signal output by the decision module: If the train is in a safe stopping state, then the intrusion status of the potential intrusion area is ignored, and a safety signal is output. If the train is in a moving danger state and the potential intrusion area is in an intrusion non-intrusion state, then a safety signal is output. If the train is in a moving danger state and the potential intrusion area is in an intrusion state, then a danger intrusion signal is output. S5. When the alarm decision signal is a dangerous intrusion signal, the alarm intensity is obtained; based on the alarm intensity, the audible and visual alarm is triggered to output the corresponding alarm signal and send the alarm signal to the monitoring center and the terminal of the guard personnel.

[0006] Furthermore, the expression used for filtering and denoising the track pressure values ​​is as follows:

[0007] In the formula, This is the filtered track pressure value at the current moment; This represents the original orbital pressure value at the current moment. This is the filtered track pressure value from the previous moment; These are the filter coefficients; Sampling time.

[0008] Furthermore, based on the filtered pressure value and the preset pressure threshold, the formula used to determine the train status is as follows:

[0009] In the formula, Decision parameters representing the current train status; This is the preset pressure threshold.

[0010] Furthermore, the decision-making rules of the decision-making module can be expressed by the following formula:

[0011] In the formula, For alarm decision signals; Train status; For visual detection confidence; This is the minimum alarm threshold.

[0012] Furthermore, based on the decision-making module's judgment rules, an alarm probability model is constructed using a conditional probability model; the probability of an alarm occurring during an intrusion is as follows:

[0013] In the formula, This represents the probability that the system will correctly trigger an alarm in the event of a real intrusion. This represents the probability that the pressure sensor correctly identifies the train as being in a moving state when the train is actually in a dangerous moving state. This is the probability of correctly detecting an intrusion based on video data from the platform monitoring area when an intrusion actually occurs. Among them, based on the pressure sensor data obtained from laboratory tests and field tests, the following was obtained: , expressed as: ; Based on the station intrusion data obtained from the simulated intrusion test experiment, we obtain The expression is: ; The probability of false alarms occurring during non-intrusion events is as follows:

[0014] In the formula, Indicates the probability of a system error alarm; This indicates the probability that the pressure sensor incorrectly identifies the train as moving; This indicates the probability of erroneously detecting an intrusion based on video data from the platform monitoring area. Among them, based on the false alarm data of the pressure sensor obtained when the train is stopped, the following was obtained: The expression is: ; Based on the error detection data of the platform monitoring area video data under non-intrusion conditions, the following results were obtained: The expression is: .

[0015] Furthermore, when the alarm decision signal is a dangerous intrusion signal, the alarm strength is obtained; the expression for the alarm strength is:

[0016] In the formula, The overall alarm intensity for that day; This is the alarm response coefficient; This serves as the alarm decision signal for the current moment; This represents the overall alarm intensity at the previous moment.

[0017] Beneficial Effects: This invention presents an intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion. By introducing a decision-making method combining pressure sensors and a vision system, this invention uses the direct and reliable "train status" signal provided by the pressure sensor as the main decision-making thread, constructing a "state-driven" alarm logic. This logic explicitly stipulates that when the pressure sensor indicates a "safe stopping state," the system unconditionally suppresses the alarm. This mechanism, in principle, completely eliminates false alarms caused by normal boarding and alighting, theoretically reducing such false alarms to near zero. Furthermore, the visual detection result can only trigger an alarm if the train is in a dangerous moving state. This architecture ensures that the system's final decision is based on the most reliable physical signals, greatly improving the overall robustness of the system. Pressure sensing and vision technology complement each other, with pressure sensors providing stable state perception unaffected by light or weather, while the vision system provides accurate spatial intrusion detection. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the intelligent intrusion detection method for platform areas according to the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] This embodiment provides an intelligent intrusion detection method for platform areas based on the fusion of pressure sensing and vision, such as... Figure 1 As shown, it includes: S1. Collect video data of the platform monitoring area through an image sensor, and simultaneously collect track pressure values ​​through a track pressure sensor; S2. Preprocess the video data of the platform monitoring area and obtain the visual detection confidence level through the target detection algorithm; obtain the intrusion status of the potential intrusion area based on the visual detection confidence level and the preset minimum alarm threshold; the intrusion status of the potential intrusion area includes intrusion status and non-intrusion status. S3. Filter and denoise the track pressure value to obtain the filtered pressure value; determine the train status based on the filtered pressure value and the preset pressure threshold; the train status includes a safe stopping status and a dangerous moving status. S4. Based on the train status and the intrusion status of the potential intrusion area, obtain the alarm decision signal output by the decision module: If the train is in a safe stopping state, then the intrusion status of the potential intrusion area is ignored, and a safety signal is output. If the train is in a moving danger state and the potential intrusion area is in an intrusion non-intrusion state, then a safety signal is output. If the train is in a moving danger state and the potential intrusion area is in an intrusion state, then a danger intrusion signal is output. S5. When the alarm decision signal is a dangerous intrusion signal, the alarm intensity is obtained; based on the alarm intensity, the audible and visual alarm is triggered to output the corresponding alarm signal and send the alarm signal to the monitoring center and the terminal of the guard; otherwise, monitoring continues.

[0022] Specifically, the track pressure sensor, based on the piezoresistive effect principle, yields the following formula for resistance versus stress:

[0023] In the formula, This represents the change in relative resistance. Sensitivity; In response to the situation; The above formula describes the process from strain to resistance change; where strain is the external pressure. The pressure applied to a special material (such as semiconductor silicon) inside the sensor causes mechanical deformation of that material; within the material's elastic range, the pressure... With strain It is directly proportional; the resistance of a sensitive material changes with its own strain. It is the relative change in resistance. This formula realizes the fundamental transformation from a mechanical quantity (deformation) to an electrical quantity (change in resistance); Tiny resistance changes It is difficult to measure and process directly; therefore, a signal conditioning circuit, typically a Wheatstone bridge, is usually connected inside or outside the sensor. This Wheatstone bridge can process minute resistance changes. Convert and amplify into an output voltage Ideally, after calibration, the final output voltage of the entire sensor will be... With applied pressure The relationship is linear; The following formula describes the effect of external pressure To the sensor's final output voltage The complete process:

[0024] In the formula, This is the output voltage of the track pressure sensor; Sensitivity; External pressure; This is the zero-point offset voltage, i.e., the output voltage when the pressure is zero; pressure Cause strain ,strain The piezoresistive effect causes a change in resistance. Resistance change The voltage is ultimately converted into an output voltage via a measuring circuit (Wheatstone bridge). ;Voltage The original pressure sample value was obtained through sampling and quantization using an analog-to-digital converter (ADC). .

[0025] Specifically, the preprocessing of the platform video data includes grayscale denoising. In this embodiment, the target detection algorithm uses the Canny edge detection algorithm to extract the static structure of the platform, and at the same time uses the neighbor frame difference algorithm to analyze the differences between consecutive frames and dynamically capture the contours of moving targets.

[0026] Preferably, the expression used for filtering and denoising the track pressure value is:

[0027] In the formula, This is the filtered track pressure value at the current moment; This represents the original orbital pressure value at the current moment. This is the filtered track pressure value from the previous moment; These are the filter coefficients; Sampling time.

[0028] Preferably, the formula used to determine the train status based on the filtered pressure value and the preset pressure threshold is:

[0029] In the formula, The current train status; The preset pressure threshold; in, To output a safe stopping status when pressure generated by a train stopping at the platform is detected; When no pressure is detected (no train in the platform area) or when moving vibration is detected (train is running in the platform area), a movement hazard status is output.

[0030] In this embodiment, even if there is no train in the platform area, a train may still be approaching at high speed and will enter the platform area in the next second; this situation is a dangerous situation, so the only safe window is for the train to stop at the platform.

[0031] Preferably, the decision-making rule of the decision module can be expressed by the following formula:

[0032] In the formula, For alarm decision-making; Train status; Visual detection confidence is a measure of the certainty of a computer vision algorithm in determining whether a real intrusion target has been detected; it is a value between 0 and 1. This is the minimum alarm zone.

[0033] Specifically, the decision rules include unconditional suppression rules for safe states and collaborative confirmation rules for dangerous states; The unconditional suppression rule for the safe state is used when the train status indication is a safe stopping state (current train status). When this occurs, regardless of whether the potential intrusion area is intrusion (visual detection confidence level),... For any value, the decision module outputs a safety signal. ); The aforementioned hazardous state collaborative confirmation rule is used when the train status indication is a moving hazardous state (current train status). If the intrusion status of the potential intrusion area is not specified (visual detection confidence level), then the intrusion status of the potential intrusion area is further determined; if the intrusion status of the potential intrusion area is not specified (visual detection confidence level), then the intrusion status of the potential intrusion area is further determined. Less than or equal to the minimum alarm threshold If ), then the decision module outputs a safety signal ( ); then when the intrusion status of the potential intrusion area is an intrusion status (visual detection confidence level) Greater than the minimum alarm threshold When this occurs, the decision module outputs a danger intrusion number. ).

[0034] Preferably, based on the decision-making rules of the decision module, an alarm probability model is constructed using a conditional probability model; the probability of an alarm occurring when an intrusion occurs is as follows:

[0035] In the formula, This represents the probability that the system will correctly trigger an alarm in the event of a real intrusion. This represents the probability that the pressure sensor correctly identifies the train as being in a moving state when the train is actually in a dangerous moving state. This is the probability of correctly detecting an intrusion based on video data from the platform monitoring area when an intrusion actually occurs. Among them, based on the pressure sensor data obtained from laboratory tests and field tests, the following was obtained: , expressed as: ; Based on the station intrusion data obtained from the simulated intrusion test experiment, we obtain The expression is: ; The probability of false alarms occurring during non-intrusion events is as follows:

[0036] In the formula, Indicates the probability of a system error alarm; This indicates the probability that the pressure sensor incorrectly identifies the train as moving; This indicates the probability of erroneously detecting an intrusion based on video data from the platform monitoring area. Among them, based on the false alarm data of the pressure sensor obtained when the train is stopped, the following was obtained: The expression is: ; Based on the error detection data of the platform monitoring area video data under non-intrusion conditions, the following results were obtained: The expression is: ; By suppressing the probability of false alarms through the alarm probability model, real threats can be identified more accurately, interference information can be reduced, and operators can focus on real alarm events. This helps to build a balanced system with "high detection rate and low false alarm rate", which meets actual security needs.

[0037] Preferably, when the alarm decision signal is a dangerous intrusion signal, the alarm strength is obtained; the expression for the alarm strength is:

[0038] In the formula, The overall alarm intensity for that day is normalized to a range of 0 to 1. This is the alarm response coefficient, and ; For the alarm decision at the current moment, and (0 = no alarm, 1 = alarm), output from the decision module; This represents the overall alarm intensity at the previous moment.

[0039] Specifically, the alarm intensity It is a quantitative risk level indicator that intuitively conveys the urgency of the risk by controlling the volume, tone, flashing frequency, and color of the sound and light alarm. The intensity is linked to the risk level. For low-intensity (a target has just entered the edge of the warning zone), consider using intermittent, low-volume beeping and a single yellow indicator light flashing slowly; monitoring personnel should combine real-time video to quickly determine the location, number, and behavior of the intruding target; if it is confirmed to be a false alarm, no intervention is required; if it is confirmed to be a real but low-risk event (a passenger approaches but does not cross the boundary), maintain close monitoring or issue a warning via broadcast; For medium intensity (the target has penetrated deep into the warning area but is moving slowly), consider using a continuous, medium-volume alarm sound and multiple yellow or orange indicator lights flashing rapidly; the monitoring personnel should initiate the emergency procedure: according to the operating procedures, enter the warning state; immediately issue a voice warning to the intrusion area through the platform broadcast system, or briefly notify nearby station staff or security personnel through walkie-talkies or internal communication systems, informing them of the scene and situation; For high-intensity (target rapidly approaching the track) scenarios, consider using high-frequency, high-volume, rapid pulse alarm sounds and high-frequency flashing or constant-on red warning lights, which can be combined with flashing red light across the entire screen frame; if the system is linked with the signal system, it can automatically or be triggered by monitoring personnel with a single button press to issue an emergency braking command to nearby trains; play emergency evacuation or warning broadcasts at maximum volume, and immediately order all available security and station staff to rush to the scene of the incident to intervene.

[0040] In this embodiment, alarm delay is used to measure and ensure the key performance indicators and safety baseline of the system's effectiveness. The total delay from data acquisition to alarm occurrence is expressed by the following formula:

[0041] In the formula, For the total delay, and , The maximum permissible response time as specified in railway safety standards; This refers to the time from the occurrence of a physical event to the completion of data acquisition by the sensor. This refers to the time from acquiring the raw data to completing the algorithm processing. This refers to the time from the generation of a decision to the execution of an action by the implementing agency. The alarm delay reflects the overall response speed from detecting a dangerous event to issuing an alarm or activating protective measures, and is lower than the maximum permissible response time specified in railway safety standards, thus meeting safety requirements; by analyzing the alarm delay, the overall performance can also be improved in a targeted manner.

[0042] In a specific embodiment, the track pressure sensor is deployed in the track area and is a piezoresistive sensor, which has high sensitivity and strong anti-interference ability; the dedicated signal conditioning circuit in the piezoresistive sensor includes a filtering module, an amplification module and a shaping module to ensure the accuracy and stability of the pressure signal; the piezoresistive sensor is connected to a status processing unit for processing the pressure sensing signal. The image sensor uses an OV5640 camera and is connected to the image processing module for grayscale conversion and median filtering. The decision module uses an FPGA / processor to perform decision processing on pressure sensing signals and visual signals; The monitoring personnel's terminal includes an HDMI display module for displaying processing results and status in real time.

[0043] This invention offers the following advantages: It provides an intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion. By introducing a decision-making method combining pressure sensors and a vision system, this invention uses the direct and reliable "train status" signal provided by the pressure sensor as the main decision-making thread, constructing a "state-driven" alarm logic. This logic explicitly stipulates that when the pressure sensor indicates a "safe stopping state," the system unconditionally suppresses the alarm. This mechanism, in principle, completely eliminates false alarms caused by normal boarding and alighting, theoretically reducing such false alarms to near zero. Furthermore, the visual detection result can only trigger an alarm if the train is in a dangerous moving state. This architecture ensures that the system's final decision is based on the most reliable physical signals, greatly improving the overall robustness of the system. Pressure sensing and vision technology complement each other, with pressure sensors providing stable state perception unaffected by light or weather, while the vision system provides accurate spatial intrusion detection.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent intrusion detection in a platform area based on pressure sensing and vision fusion, characterized in that, include: S1. Collect video data of the platform monitoring area through an image sensor, and simultaneously collect track pressure values ​​through a track pressure sensor; S2. Preprocess the video data of the platform monitoring area and obtain the visual detection confidence score through the target detection algorithm; The intrusion status of the potential intrusion area is obtained based on the visual detection confidence level and the preset minimum alarm threshold; the intrusion status of the potential intrusion area includes intrusion status and non-intrusion status. S3. Filter and denoise the track pressure value to obtain the filtered pressure value; determine the train status based on the filtered pressure value and the preset pressure threshold; the train status includes a safe stopping status and a dangerous moving status. S4. Based on the train status and the intrusion status of the potential intrusion area, obtain the alarm decision signal output by the decision module: If the train is in a safe stopping state, then ignore the intrusion status of the potential intrusion area and output a safety signal; If the train is in a moving danger state and the potential intrusion area is in an intrusion non-intrusion state, then a safety signal is output. If the train is in a moving danger state and the potential intrusion area is in an intrusion state, then a danger intrusion signal is output. S5. When the alarm decision signal is a dangerous intrusion signal, the alarm intensity is obtained; based on the alarm intensity, the audible and visual alarm is triggered to output the corresponding alarm signal and send the alarm signal to the monitoring center and the terminal of the guard personnel.

2. The intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion as described in claim 1, characterized in that, The expression used for filtering and denoising the track pressure values ​​is: In the formula, This is the filtered track pressure value at the current moment; This represents the original orbital pressure value at the current moment. This is the filtered track pressure value from the previous moment; These are the filter coefficients; Sampling time.

3. The intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion as described in claim 2, characterized in that, The formula used to determine the train status based on the filtered pressure value and the preset pressure threshold is as follows: In the formula, Decision parameters representing the current train status; This is the preset pressure threshold.

4. The intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion as described in claim 1, characterized in that, The decision-making rules of the decision module can be expressed by the following formula: In the formula, For alarm decision signals; Train status; For visual detection confidence; This is the minimum alarm threshold.

5. The intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion as described in claim 4, characterized in that, Based on the decision-making rules of the decision module, an alarm probability model is constructed using a conditional probability model; the probability of an alarm occurring when an intrusion occurs is as follows: In the formula, This represents the probability that the system will correctly trigger an alarm in the event of a real intrusion. This represents the probability that the pressure sensor correctly identifies the train as being in a moving state when the train is actually in a dangerous moving state. This is the probability of correctly detecting an intrusion based on video data from the platform monitoring area when an intrusion actually occurs. Among them, based on the pressure sensor data obtained from laboratory tests and field tests, the following was obtained: , expressed as: ； Based on the station intrusion data obtained from the simulated intrusion test experiment, we obtain The expression is: ； The probability of false alarms occurring during non-intrusion events is as follows: In the formula, Indicates the probability of a system error alarm; This indicates the probability that the pressure sensor incorrectly identifies the train as moving; This indicates the probability of erroneously detecting an intrusion based on video data from the platform monitoring area. Among them, based on the false alarm data of the pressure sensor obtained when the train is stopped, the following was obtained: The expression is: ； Based on the error detection data of the platform monitoring area video data under non-intrusion conditions, the following results were obtained: The expression is: 。 6. The intelligent intrusion detection method for platform areas based on pressure sensing and vision fusion as described in claim 1, characterized in that, When the alarm decision signal is a dangerous intrusion signal, the alarm strength is obtained; the expression for the alarm strength is: In the formula, The overall alarm intensity for that day; This is the alarm response coefficient; This serves as the alarm decision signal for the current moment; This represents the overall alarm intensity at the previous moment.

Images