Threshold and angle self-adaptive infrared flame detection system and control method

Abstract

The invention relates to the technical field of flame target detection, in particular to a threshold and angle self-adaptive infrared flame detection system and a control method, and the system comprises an infrared detection unit, a holder mechanism, a data processing unit and a fire alarm unit, the data processing unit divides each frame of real-time image into a to-be-identified area and a background area according to the temperature variation among the continuous multiple frames of real-time images of the monitoring area, and dynamically adjusts an abnormal temperature threshold according to the temperature of the background area near the to-be-identified area; extracting a suspected flame region from the to-be-identified region according to the abnormal temperature threshold value, preliminarily eliminating the influence of an interference heat source according to whether the displacement of the suspected flame region is continuous or not, calculating the position of the suspected flame region when the displacement of the suspected flame region is discontinuous, and judging whether the suspected flame region is a normal heat source of a fixed point location or not; therefore, interference of normal heat sources such as industrial fire to fire is eliminated, and the method is suitable for identification of flames which are long in distance, small in scale and unobvious in characteristics.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flame target detection, in particular to an infrared flame detection system and control method with adaptive threshold and angle. BACKGROUND

[0002] In the industrial field and forest fire prevention, infrared flame detectors have been widely used. Among them, infrared flame detectors based on image recognition technology are applied more and more because they can obtain images of the monitored area, making it convenient for patrol personnel to further analyze the fire situation according to the images. The infrared flame detector based on image recognition technology obtains the image of the scene by using an infrared thermal imaging camera, and identifies the flame in the monitoring area according to the temperature threshold combined with feature extraction. The infrared radiation received by the flame detector and other background will be affected by many interference factors, such as scattering of water and other impurities in the atmosphere, infrared radiation from the sun and other heat sources, etc. The traditional method of correcting the temperature calculation result based on the ideal atmospheric model will be affected by the actual atmospheric environment, resulting in a large error between the calculated temperature and the actual temperature. When a fixed temperature threshold is used, the miss detection rate for small-scale flames at a long distance is high.

[0003] The prior art CN114639041B proposes a flame monitoring method based on motion and color features, which takes the flickering feature of the flame as the theoretical basis to obtain the area with high frequency change of pixel value from the video stream frame image, and locates the real-time motion target area according to the segmented time length image pixel value change frequency statistics. For the motion target area, the motion amplitude (area change amplitude), color feature, etc. are combined to realize the detection of the flame, overcoming the problems of poor anti-interference ability and sample quantity constraint of traditional deep learning algorithms.

[0004] However, in the early stage of a fire, the flame is small in size, and especially when the flame is far away from the detector, its shape feature is not obvious, and its flickering characteristics are particularly difficult to capture, which causes a high miss detection rate in the identification of small-scale flames at a long distance. This is difficult to meet the monitoring needs in places where fire protection requirements are strict, or in order to improve the reliability of fire monitoring, a large density of monitoring equipment is arranged, which undoubtedly increases the construction and use cost of the monitoring system. SUMMARY

[0005] In order to solve the technical problem of high miss detection rate in the identification of small-scale flames at a long distance in the prior art, the present application provides an infrared flame detection system and control method with adaptive threshold and angle optimization function, wherein the system comprises: an infrared detection unit, a cloud platform mechanism, and a data processing unit. The infrared detection unit is used to receive the infrared light intensity radiated by the monitored area and convert the infrared light intensity into corresponding electrical signals, which are transmitted to the data processing unit. The data processing unit includes: Infrared image generator: used to generate a real-time image of the monitored area based on the electrical signals transmitted by the infrared detection unit; Region segmenter: Used to calculate the temperature change of corresponding pixels between consecutive real-time images based on multiple real-time images, and to divide each real-time image into the region to be identified and the background region based on the temperature change. Suspected flame region extractor: used to select the background region near the region to be identified in each frame of real-time image, dynamically adjust the abnormal temperature threshold according to the average temperature and / or temperature standard deviation of the nearby background region, and extract the suspected flame region from the region to be identified according to the abnormal temperature threshold. Flame detector: Used to determine whether the movement trajectory of a suspected flame area is continuous. If so, the suspected flame area is excluded from the fire situation. If not, the pan-tilt mechanism is controlled to drive the infrared detection unit to move according to the position of the suspected flame area in the last frame of real-time image, so that the viewing angle is aimed at the suspected flame area. The position of the suspected flame area in the monitoring area is calculated according to the angle of the infrared detection unit. It is determined whether the suspected flame area is located at the position of a normal heat source. If so, the fire situation is excluded. If not, a fire warning is issued.

[0006] The control method disclosed in this invention includes the following steps: Acquire real-time images of the monitored area in multiple consecutive frames, and divide each real-time image frame into the area to be identified and the background area based on the temperature change. For each frame of real-time image, the nearby background regions adjacent to the region to be identified are obtained, and the abnormal temperature threshold is adaptively adjusted based on the average temperature and / or temperature standard deviation of the nearby background regions. For each frame of real-time image, a suspected flame area is extracted from the area to be identified based on the abnormal temperature threshold. The movement trajectory of the suspected flame area is determined based on multiple consecutive frames of real-time images. If it is continuous, the suspected flame area is excluded from the fire situation. If not, the angle of the detector is adjusted according to the position of the suspected flame area in the last frame of real-time image so that the suspected flame area is located in the center of the field of view of the infrared detection unit. The location of the suspected flame area within the monitoring area is calculated based on the angle of the infrared detection unit. It is then determined whether the suspected flame area is located at the location of a normal heat source. If so, the fire is ruled out; otherwise, a fire alert is issued.

[0007] The technical effects and advantages of this invention are as follows: Real-time images are divided into a region to be identified and a background region by measuring the rate of temperature change. The abnormal temperature threshold is adjusted based on the average temperature and / or temperature standard deviation of the nearby background region, achieving adaptive threshold adjustment. This ensures that the determination of abnormal temperatures is not affected by the distance between the monitored area and the detection system, making it more scientific and reasonable and significantly reducing the probability of misjudgment. Simultaneously, by analyzing multiple consecutive frames of real-time images, the continuity of the movement trajectory of suspected flame areas is determined, eliminating interference from normal moving heat sources such as workers and vehicles. The position of suspected flame areas is calculated by the angle of the detection unit, determining whether the suspected flame area is located at a normal heat source, thereby eliminating interference from fixed artificial heat sources and further improving the accuracy of fire assessment, providing more effective protection for the safety of the monitored area. The method of this invention can identify flames without extracting and analyzing the shape and flickering features of suspected flame areas, making it applicable to the identification of flames that are far away, small in scale, and have unclear features, improving the sensitivity of flame identification and making fire warnings more timely. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the overall structure of the system disclosed in this invention.

[0009] Figure 2 This is a flowchart of the flame recognition process disclosed in this invention.

[0010] Figure 3 This is a flowchart illustrating the process of dividing the region to be identified and the background region in this invention.

[0011] Figure 4 This is a flowchart illustrating the process of determining the nearby background regions adjacent to the region to be identified in this invention.

[0012] Figure 5 This is a diagram illustrating the effectiveness of the invention in identifying small-scale flames at a distance. Detailed Implementation

[0013] 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, and 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.

[0014] When using infrared radiation emitted by a flame for flame identification, changes in atmospheric conditions and the background environment not only affect the intensity of infrared radiation emitted by the flame received by the detector, but also affect the intensity of infrared radiation emitted by the background environment received by the detector. The infrared intensity radiated by the flame and the area near its location is affected by atmospheric conditions in the same way for the part received by the detector. The core point of flame identification is not the absolute temperature of a certain point, but the significant difference between that point and the ambient temperature, and these differences are consistent with the characteristics of a flame.

[0015] Example 1 Based on the above analysis, this invention provides an infrared flame detection system with adaptive threshold and angle optimization functions, aiming to identify small-scale flames at a distance. (Reference) Figure 1 The system includes: an infrared detection unit, a pan-tilt mechanism, a data processing unit, and a fire alarm unit.

[0016] The infrared detection unit is used to receive the infrared light intensity radiated by the monitored area and convert the infrared light intensity into a corresponding electrical signal, which is then transmitted to the data processing unit. The gimbal mechanism is used to adjust the viewing angle of the infrared detection unit; The fire alarm unit is used to issue or not issue a fire alert based on the flame recognition results of the real-time image from the data processing unit. The data processing unit is used to generate a real-time image of the monitored area based on the electrical signal transmitted by the infrared detection unit, and to perform flame identification on the real-time image.

[0017] Specifically, the data processing unit includes: Infrared image generator: used to generate a real-time image of the monitored area based on the electrical signals transmitted by the infrared detection unit; Region segmenter: Used to calculate the temperature change of corresponding pixels between consecutive real-time images based on multiple real-time images, and to divide each real-time image into the region to be identified and the background region based on the temperature change. Suspected Flame Region Extractor: Used to select the background region near the region to be identified in each frame of real-time image, dynamically adjust the abnormal temperature threshold according to the average temperature of the nearby background region, and extract the suspected flame region from the region to be identified based on the abnormal temperature threshold. Flame detector: Used to determine whether the movement trajectory of a suspected flame area is continuous. If so, the suspected flame area is excluded from the fire situation. If not, the pan-tilt mechanism is controlled according to the position of the suspected flame area in the last frame of real-time image. The pan-tilt mechanism drives the infrared detection unit to move and point the viewpoint at the suspected flame area. The position of the suspected flame area in the monitoring area is calculated according to the angle of the infrared detection unit. It is determined whether the suspected flame area is located at the location of a normal heat source. If so, the fire situation is excluded. If not, a fire warning is issued.

[0018] refer to Figure 2 The flame recognition process of the above system is as follows: S100: The infrared detection unit continuously receives infrared light intensity from the monitored area and converts the infrared light intensity into an electrical signal through a photoelectric conversion element, which is then transmitted to the data processing unit. S200, the infrared image generator of the data processing unit generates a real-time image of the monitoring area based on the electrical signal transmitted by the infrared detection unit, and transmits the real-time image to the area divider. Specifically, the principle of infrared thermal imaging is based on the electrical signal generated by photoelectric conversion elements (such as a CCD sensor with a red filter, which only accepts infrared light in a specific wavelength range), and the temperature is calculated inversely based on the principle of thermal radiation (Stephan-Boltzmann law) and an atmospheric transmittance model (for example).

[0019] Typically, an infrared thermal imaging camera's image sensor chip integrates tens of thousands of photoelectric conversion elements, each of which is a pixel. Infrared light radiated from the monitored area shines onto the image sensor chip through a lens assembly. Each pixel on the image sensor chip generates a photoelectric signal based on the intensity of the received infrared light. After processing by the data processing unit, each pixel can reflect the temperature value of the corresponding point in the monitored area, thus forming an infrared image reflecting the temperature distribution of the monitored area. Some cameras can generate pseudo-color images based on temperature to intuitively perceive the temperature distribution of the monitored area, while others can generate grayscale images based on temperature, displaying the temperature distribution of the monitored area using different grayscale values.

[0020] S300: After receiving multiple consecutive real-time images of a predetermined number of monitoring frames, the region divider calculates the temperature change of each pixel in each real-time image relative to the corresponding pixel in the previous real-time image, and marks the region to be identified and the background region in each real-time image based on the temperature change. When an abnormal heat source appears in the monitored area, whether it is a flame or other interfering heat source, it will cause the temperature of some pixels to change rapidly. Therefore, by calculating the temperature change between two real-time images, the area with an abnormal heat source or abnormal temperature rise can be identified, and flame recognition can be performed on this area.

[0021] S400, the region divider passes the image with the marked region to be identified and background region to the suspected flame region extractor. The suspected flame region extractor calculates the average temperature of the background region near the region to be identified in each real-time image frame, and dynamically adjusts the abnormal temperature threshold according to the average temperature of the background region. Then, it extracts the suspected flame region from the region to be identified in each real-time image frame according to the abnormal temperature threshold. The farther the infrared detection unit is from the heat source, the less infrared light intensity it can receive. Although the data processing unit compensates for the distance when calculating the temperature based on the infrared light intensity, the distance to the heat source is estimated, so there is still a certain deviation. The magnitude of this deviation depends only on the distance at any given moment. Therefore, the infrared light intensity radiated by the heat source and the nearby background undergoes the same loss when it reaches the infrared detection unit, allowing the temperature difference between the heat source and the nearby background area to still be recorded by the infrared detection unit. Therefore, by calculating the average temperature of the background area near the area to be identified and dynamically adjusting the abnormal temperature threshold for extracting suspected flame areas, the influence of distance changes on the infrared light intensity can be eliminated, thereby improving the accuracy of flame identification.

[0022] The S500 flame detector analyzes whether the movement trajectory of a suspected flame area in multiple consecutive real-time images is continuous. If so, the suspected flame area is excluded from the fire situation. If not, the pan-tilt mechanism is controlled according to the position of the suspected flame area in the last real-time image. The pan-tilt mechanism drives the infrared detection unit to move and point the viewpoint at the suspected flame area. The position of the suspected flame area in the monitoring area is calculated according to the angle of the infrared detection unit. It is determined whether the suspected flame area is located at the position of a normal heat source. If so, the fire situation is excluded. If not, a fire warning is issued. For abnormal combustion heat sources like fires, their occurrence is often spatially random and temporally sudden. However, some interfering heat sources in everyday use scenarios are continuous in both space and time. For example, when a worker or vehicle moves from off-screen to in-screen, after extraction in steps S300 and S400, the location of the area to be identified or the suspected flame area in each frame of the real-time image will inevitably move from the edge of the real-time image into the frame, or remain at the edge, without any sudden changes. Therefore, motion trajectory analysis can initially eliminate some interfering heat sources.

[0023] S600: Calculate the location of the suspected flame area in the monitoring area based on the angle of the infrared detection unit, and determine whether the suspected flame area is located at the location of a normal heat source. If so, the fire is ruled out; otherwise, a fire warning is issued.

[0024] When the displacement of the suspected flame area is discontinuous, it is necessary to determine whether it is another type of interfering heat source. For example, in some application scenarios, there are artificial heat sources that need to be used daily, such as flares, boilers, and gas stoves in oil fields, coal mines, or refineries. These interfering heat sources may also suddenly appear in the real-time image of the detection system during use, and they also have discontinuous displacement. However, the location and size of these interfering heat sources are fixed, and they may not change even on a long time scale of years.

[0025] Therefore, when the system is started, these known artificial heat sources can be calibrated in advance to determine the pitch and horizontal angles of the infrared detection unit (or the pitch and horizontal rotation mechanisms in the pan-tilt mechanism) when the infrared detection unit is aligned with the artificial heat source. This can be used to determine whether the flame area is located at the position of a normal heat source.

[0026] Through the above steps, the system provided by this invention can adaptively adjust the abnormal temperature threshold according to the distance to the heat source, thereby reducing the impact of distance on infrared light intensity. Simultaneously, by analyzing the dynamic characteristics of suspected flame areas and their actual locations to determine abnormal fire conditions, it can effectively identify small-scale flames at a distance with indistinct geometric and spectral characteristics, reducing the false negative rate.

[0027] Example 2 Based on the system provided in Embodiment 1, this invention provides an infrared flame detection method. Combined with this system, it can achieve accurate identification of small-scale flames at a distance. (Refer to...) Figure 2 This includes the following steps: S1. Acquire real-time images of the monitoring area in multiple consecutive frames, and divide each real-time image frame into the area to be identified and the background area according to the temperature change. Once the number of frames of the real-time images obtained is sufficient to identify the movement characteristics of the heat source, the flame recognition program can be started. For example, the program can be set to sample real-time images in the video stream at a frequency of 25Hz. Based on experience, the movement characteristics of the heat source can be accurately determined when 6 to 20 seconds of video are obtained. Therefore, the flame recognition program can be set to start executing after 150 or 500 frames of real-time images are obtained, that is, the flame recognition start threshold is set to 150 frames or more.

[0028] Specifically, the temperature of the background area does not change significantly in a short period of time, and the sampling frequency of the infrared detection unit is very high, while the image frame update frequency of the video stream formed by the data processing unit is very fast. Therefore, the temperature change of the pixels contained in the background area between each real-time image frame should be small. Figure 3 The regions to be identified and the background regions in each frame of a real-time image can be marked using the following steps: S11. For the first real-time image frame, select one pixel at a time and calculate the temperature change of the pixel compared to the corresponding pixel in the previous real-time image frame. If the temperature change of the pixel is greater than the first threshold, then the pixel is included in the area to be identified. The previous frame of the implementation image is the last real-time image that participated in the flame recognition in the previous flame recognition program. If it is the first flame recognition program executed after the system starts, the previous real-time image is the calibration image, that is, the real-time image of the monitoring area used for calibration obtained when the system starts.

[0029] S12. For each subsequent real-time image frame, map the region to be identified from the previous real-time image frame adjacent to it into this real-time image. S13. Select one pixel at a time and calculate the temperature change of the pixel relative to the corresponding pixel in the previous real-time image. If the temperature change of the pixel is greater than the first threshold, the pixel is included in the area to be identified. If the temperature change of the pixel is less than the second threshold, the pixel is included in the background area. If the temperature change of the pixel is greater than or equal to the second threshold and less than or equal to the first threshold, the pixel's label is not adjusted.

[0030] By introducing an exit mechanism to record the movement of heat sources, when the temperature of a pixel suddenly rises, it indicates that the heat source has moved to the area where the pixel is located, and the pixel is included in the area to be identified. When the temperature of the pixel suddenly drops, it indicates that the heat source has left the area where the pixel is located, and the pixel is adjusted to the background area, thus realizing the recording of the movement of heat sources (i.e., the area to be identified) in multiple frames of real-time images.

[0031] Furthermore, considering the potential impact of isolated high-temperature points on the background temperature, such as welding flashes and hot liquid splashes (common in steel mills), and the spatial randomness and temporal discontinuity of these heat sources, which can affect subsequent flame identification, these isolated high-temperature points should be eliminated when delineating the area to be identified. This can be achieved through the following steps: S14. If the number of pixels in the area to be identified is less than the minimum identification threshold, then the area to be identified shall be included in the background area. S15. If the number of real-time images containing the area to be identified in multiple real-time images is less than the minimum number of identification frames, it indicates that the isolated high-temperature point is noise with abnormal signal. Therefore, the fire can be ruled out, the flame identification program can be terminated, and the last real-time image that does not contain the area to be identified can be used in the next flame identification program. S16. Reacquire multiple real-time images. When the total number of frames of the newly acquired real-time images and the real-time images transmitted by the previous flame recognition program is equal to the flame recognition start threshold, start the flame recognition program.

[0032] Considering that interfering heat sources such as welding light and high-temperature spatter have very small areas, limiting the minimum number of pixels in the area to be identified to eliminate these interfering heat sources can effectively improve the accuracy of flame recognition.

[0033] After dividing the real-time image into the region to be identified and the background region, the abnormal temperature threshold can be adaptively adjusted according to the temperature of the background region: S2. For each frame of real-time image, obtain the nearby background areas adjacent to the area to be identified, and adaptively adjust the abnormal temperature threshold according to the average temperature and / or temperature standard deviation of the nearby background areas. Specifically, refer to Figure 4 The range of the adjacent background region to be identified can be determined by the following steps: S21. Calculate the total number of pixels in the region to be identified, determine the initial neighborhood width according to a certain proportion based on the total number of pixels, obtain the nearby background region of the region to be identified based on the initial neighborhood width, and determine a minimum range of nearby background region. Generally, the initial neighborhood width W_org can be determined by the following formula: W_org = max(5, min(50, 0.3*sqrt(S_check))) In the formula, S_check is the total number of pixels in the region to be identified.

[0034] S22. Calculate the temperature gradient of each pixel on the boundary of the nearby background region in the normal direction of the boundary. When the temperature gradient of the pixel is greater than the abnormal gradient threshold, include the pixels in the boundary normal direction adjacent to the pixel into the nearby background region and dynamically adjust the size of the nearby background region.

[0035] The background area should have a relatively uniform temperature distribution. By using a temperature gradient to search for areas with uniform temperature distribution along the boundary of the area to be identified, a relatively small nearby background area can be obtained. This can effectively reduce the impact of distance differences on the recognition results. Furthermore, its range should vary with the size of the area to be identified; the larger the area to be identified, the larger the range of the adjacent background area should be.

[0036] After obtaining a clear nearby background region, the abnormal temperature threshold is adaptively adjusted based on the temperature characteristics of the nearby background region. Specifically, the adjustment can be performed in the following ways: Th_tem_abn = Tem_bg_mean + k_safe * Tem_bg_Std In the formula, Th_tem_abn is the abnormal temperature threshold, Tem_bg_mean is the average temperature of the nearby background area, k is the safety factor, which is generally taken as 10-15, and Tem_bg_Std is the standard deviation of the temperature of the nearby background area.

[0037] Furthermore, the temperature change of the background area between each real-time image frame can be considered. In the early stages of a fire, the temperature of the heat source changes rapidly, and the number of pixels in the area to be identified also increases rapidly. Therefore, it is necessary to introduce the temperature change rate to adjust the abnormal temperature threshold. Dy_tem_abn =(Factor_Sm * Th_tem_abn + (1 - Factor_Sm) * (Th_tem_abn *Factor_Grad)) In the formula, Dy_tem_abn is the dynamic abnormal temperature threshold, Factor_Sm is the time smoothing factor used to suppress frequent fluctuations in the threshold, and Factor_Grad is the temperature gradient correction factor, calculated by the following formula: Factor_Grad = 1.0 + 0.15 * Tem_bg_Grad In the formula, Tem_bg_Grad is the average temperature change of the corresponding background area in two adjacent real-time images.

[0038] S3. For each frame of real-time image, extract the suspected flame area from the area to be identified according to the abnormal temperature threshold, and determine whether the movement trajectory of the suspected flame area is continuous according to multiple consecutive frames of real-time images. If so, exclude the suspected flame area from the fire situation. If not, adjust the angle of the detector according to the position of the suspected flame area in the last frame of real-time image so that the suspected flame area is located in the center of the detector's field of view. Specifically, the suspected flame area is extracted from the area to be identified through the following steps: S31. For each pixel in the region to be identified in each frame of the image, if the temperature of the pixel is greater than the abnormal temperature threshold, the pixel is counted as a suspected flame region.

[0039] After identifying the suspected flame area, analyze its movement trajectory. There are several methods for this; a relatively simple calculation method is the centroid method, which includes the following steps: S32. For each frame of real-time image, calculate the centroid coordinates of the suspected flame region: C_x =int( sum(x_sus+i) / Area_sus) C_y =int( sum(y_sus+i) / Area_sus) In the formula, x_sus and y_sun are the number of pixels in the width and length directions of the suspected flame region, respectively. Area_sus is the area of ​​the suspected flame region, that is, the x-coordinate of the centroid is the sum of the pixel coordinates in the width direction divided by the area of ​​the suspected flame region, and the y-coordinate of the centroid is the sum of the pixel coordinates in the length direction divided by the area of ​​the suspected flame region.

[0040] S33. Calculate the displacement trajectory of the suspected flame region in multiple consecutive real-time images based on the centroid coordinates. When the movement trajectory of the suspected flame region is connected to the edge of the real-time image and the displacement of the suspected flame region between two adjacent real-time images is less than the continuity threshold, it is determined that the movement trajectory of the suspected flame region is continuous.

[0041] Specifically, whether the movement trajectory is connected to the edge of the real-time image can be determined based on whether the suspected flame area in the first frame of the real-time image is located at the edge of the real-time image, or based on whether the centroid of the suspected flame area in the first frame of the real-time image is less than a preset threshold from the edge of the real-time image.

[0042] When the movement trajectory of the suspected flame area is continuous, it indicates that the heat source moved from outside into the monitored area, so a fire can be ruled out directly, and there is no need to adjust the camera angle. When the movement is discontinuous, it indicates that there may be a fire. The field of view angle of the infrared detection unit should be adjusted to focus on monitoring the suspected flame area and identify whether it is an interfering heat source or a fire.

[0043] S4. Calculate the location of the suspected flame area in the monitoring area based on the angle of the infrared detection unit, and determine whether the suspected flame area is located at the location of a normal heat source. If so, rule out a fire; otherwise, issue a fire alert.

[0044] In normal application scenarios, some commonly used artificial heat sources, such as boilers, heaters, and industrial fires (flare fires), are usually located in fixed positions. When the infrared detection unit's viewing angle is aligned with these heat sources, the elevation and horizontal angles of the infrared detection unit can be used to determine whether the suspected flame area is located in the position of a normal heat source. If so, it means that the suspected flame area is a normal heat source and the fire can be ruled out. If it is located in other positions, it is an abnormal heat source, and a fire warning will be issued.

[0045] The advantage of the above method lies in its ability to accurately locate suspected flame areas from multiple angles, without needing to analyze the shape and flickering characteristics of the flame. It also demonstrates excellent identification performance for flames that are far away or whose features are unclear. For example... Figure 5The demonstrated recognition effect shows that the system provided by this invention can monitor distant villages from a high position and successfully identify flames on the ground. It can effectively distinguish between normal and abnormal heat sources, greatly improving the accuracy and reliability of flame detection. In practical applications, various complex interference factors may exist in different usage scenarios, and this method can effectively cope with these interferences by adjusting the angle and determining the position of the infrared detection unit.

[0046] For example, in industrial production environments, in addition to common artificial heat sources, there may be temporary and irregular heat sources, such as welding sparks during maintenance. This method can accurately identify whether these temporary heat sources are abnormal by continuously monitoring and dynamically judging the location of suspected flame areas. Furthermore, for some special scenarios, such as outdoor forests and grasslands, where interference may occur due to factors like sunlight reflection and animal body temperature, this method can also quickly eliminate these interfering factors and accurately locate the true flame source through adaptive angle adjustment and position judgment mechanisms.

[0047] Furthermore, this method is highly adaptable and scalable. With continuous technological advancements and evolving application scenarios, the parameters of the infrared detection unit can be adjusted and optimized according to actual needs, further improving system performance. Moreover, this method can be combined with other monitoring technologies, such as smoke sensors and image recognition technology, to form a more comprehensive fire monitoring system, providing stronger protection for people's lives and property.

[0048] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A threshold and angle adaptive infrared flame detection system, characterized in that, include: Infrared detection unit, gimbal mechanism, data processing unit; The infrared detection unit is used to receive the infrared light intensity radiated by the monitored area and convert the infrared light intensity into a corresponding electrical signal, which is then transmitted to the data processing unit. The data processing unit includes: Infrared image generator: used to generate a real-time image of the monitored area based on the electrical signals transmitted by the infrared detection unit; Region segmenter: Used to calculate the temperature change of corresponding pixels between consecutive real-time images based on multiple real-time images, and to divide each real-time image into the region to be identified and the background region based on the temperature change. Suspected flame region extractor: used to select the background region near the region to be identified in each frame of real-time image, dynamically adjust the abnormal temperature threshold according to the average temperature and / or temperature standard deviation of the nearby background region, and extract the suspected flame region from the region to be identified according to the abnormal temperature threshold. Flame detector: Used to determine whether the movement trajectory of a suspected flame area is continuous. If so, the suspected flame area is excluded from the fire situation. If not, the pan-tilt mechanism is controlled to drive the infrared detection unit to move according to the position of the suspected flame area in the last frame of real-time image, so that the viewing angle is aimed at the suspected flame area. The position of the suspected flame area in the monitoring area is calculated according to the angle of the infrared detection unit. It is determined whether the suspected flame area is located at the position of a normal heat source. If so, the fire situation is excluded. If not, a fire warning is issued.

2. The system according to claim 1, characterized in that, The region segmenter divides the region to be identified and the background region through the following steps: For each frame of real-time image, the region to be identified in the previous frame of real-time image that is adjacent to it is mapped into this real-time image; Select a pixel one by one and calculate the temperature change of that pixel compared to the corresponding pixel in the previous real-time image. If the temperature change of that pixel is greater than a first threshold, then the pixel is included in the area to be identified. If the temperature change of that pixel is less than a second threshold, then the pixel is included in the background area. If the temperature change of that pixel is greater than or equal to the second threshold and less than or equal to the first threshold, then the label of that pixel is not adjusted.

3. The system according to claim 1, characterized in that, The region divider also performs the following steps: If the number of pixels in the area to be identified is less than the minimum identification threshold, then the area to be identified is included in the background area. If the number of real-time images containing the area to be identified in multiple real-time images is less than the minimum number of identification frames, then the fire is excluded, the flame identification process is terminated, and the last real-time image that does not contain the area to be identified is used in the next flame identification process. Reacquire multiple real-time images. When the total number of frames of the newly acquired real-time images and the real-time images transmitted by the previous flame recognition program equals the flame recognition start threshold, start the flame recognition program.

4. The system according to claim 1, characterized in that, The suspected flame region extractor determines the adjacent background region to the region to be identified through the following steps: S21. For each frame of real-time image, calculate the total number of pixels in the region to be identified, determine the preliminary neighborhood width based on the total number of pixels, and obtain the background region near the region to be identified based on the preliminary neighborhood width. S22. Calculate the temperature gradient of each pixel on the boundary of the nearby background region in the normal direction of the boundary. When the temperature gradient of the pixel is greater than the abnormal gradient threshold, include the pixels in the boundary normal direction adjacent to the pixel in the nearby background region.

5. The system according to claim 1, characterized in that, The abnormal temperature threshold is adaptively adjusted through the following steps: For each frame of real-time image, calculate the average temperature and temperature standard deviation of the nearby background area; Calculate the change in average temperature of the nearby background region in the current real-time image compared to the previous real-time image; The abnormal temperature threshold is adjusted based on the average temperature, temperature standard deviation, and the amount of change in the average temperature of the nearby background area.

6. The system according to claim 1, characterized in that, Extract suspected flame areas from the area to be identified using the following steps: For each pixel in the region to be identified in each frame of the image, if the temperature of the pixel is greater than the abnormal temperature threshold, the pixel is counted as a suspected flame region.

7. The system according to claim 1, characterized in that, Determine whether the movement trajectory of a suspected flame area is continuous using the following steps: S32. For each frame of real-time image, calculate the centroid coordinates of the suspected flame region: S33. Calculate the displacement trajectory of the suspected flame region in multiple consecutive real-time images based on the centroid coordinates. When the movement trajectory of the suspected flame region is connected to the edge of the real-time image and the displacement of the suspected flame region between two adjacent real-time images is less than the continuity threshold, it is determined that the movement trajectory of the suspected flame region is continuous.

8. A control method employing the system according to any one of claims 1-7, characterized in that, Includes the following steps: Acquire real-time images of the monitored area in multiple consecutive frames, and divide each real-time image frame into the area to be identified and the background area based on the temperature change. For each frame of real-time image, the nearby background regions adjacent to the region to be identified are obtained, and the abnormal temperature threshold is adaptively adjusted based on the average temperature and / or temperature standard deviation of the nearby background regions. For each frame of real-time image, a suspected flame area is extracted from the area to be identified based on the abnormal temperature threshold. The movement trajectory of the suspected flame area is determined based on multiple consecutive frames of real-time images. If it is continuous, the suspected flame area is excluded from the fire situation. If not, the angle of the detector is adjusted according to the position of the suspected flame area in the last frame of real-time image so that the suspected flame area is located in the center of the field of view of the infrared detection unit. The location of the suspected flame area within the monitoring area is calculated based on the angle of the infrared detection unit. It is then determined whether the suspected flame area is located at the location of a normal heat source. If so, the fire is ruled out; otherwise, a fire alert is issued.

9. The method according to claim 8, characterized in that, The following steps are used to label the region to be identified and the background region in each frame of real-time image: For each frame of real-time image, the region to be identified in the previous frame of real-time image that is adjacent to it is mapped into this real-time image; Select a pixel one by one and calculate the temperature change of that pixel compared to the corresponding pixel in the previous real-time image. If the temperature change of that pixel is greater than a first threshold, then the pixel is included in the area to be identified. If the temperature change of that pixel is less than a second threshold, then the pixel is included in the background area. If the temperature change of that pixel is greater than or equal to the second threshold and less than or equal to the first threshold, then the label of that pixel is not adjusted.

10. The method according to claim 8, characterized in that, The following steps are used to determine the nearby background regions adjacent to the region to be identified: Calculate the total number of pixels in the region to be identified, determine the initial neighborhood width based on the total number of pixels, obtain the nearby background region of the region to be identified based on the initial neighborhood width, and determine a minimum range of nearby background region; Calculate the temperature gradient of each pixel on the boundary of the nearby background region along the normal direction of the boundary. When the temperature gradient of a pixel is greater than the abnormal gradient threshold, include the pixels adjacent to that pixel along the boundary normal direction into the nearby background region and dynamically adjust the size of the nearby background region.

Images