A smoke alarm method

Abstract

The present application relates to the field of smoke alarm, specifically to a smoke alarm method, comprising laser irradiation, image acquisition, image processing, feature extraction, light column recognition, path and diffusion recognition to identify smoke concentration and alarm steps, laser irradiation on the detection area, image acquisition device shooting the laser light path, using active laser light source to actively create visible reference light path in weak light or no light environment, thereby overcoming the problem of detection benchmark loss and recognition accuracy decline caused by insufficient ambient light in low illumination conditions of traditional image recognition method; through linearity, white pixel ratio and white pixel ratio change in shape, area and dynamic three dimensions to capture smoke characteristics, and double path determination, using light column integrity recognition of laser light path to determine low concentration smoke state, using diffusion degree recognition of laser light path to determine high concentration smoke state, to ensure reliability in the full concentration range.

Description

Technical Field

[0001] This invention relates to the field of smoke alarms, and more specifically to a smoke alarm method. Background Technology

[0002] Smoke alarms are primarily implemented through smoke detectors or image-processing-based smoke alarm systems. Smoke detectors have built-in sensors that monitor the concentration of smoke particles in the air in real time. When abnormal smoke or combustion products are detected, a high-decibel alarm is immediately sounded, alerting people to evacuate or take firefighting measures, thus effectively reducing loss of life and property caused by fire. Their core components typically include photoelectric sensors or ionization sensors. Photoelectric sensors work by detecting changes in the light scattered by smoke and are suitable for detecting larger smoke particles; ionization sensors utilize radioactive elements to ionize the air and are more sensitive to smaller particles. Image-processing-based smoke alarm systems are an advanced fire early warning technology that combines computer vision and intelligent algorithms. They use cameras or infrared sensors to acquire images of the monitored area in real time and use image processing algorithms to extract and analyze smoke features, thereby achieving early warning of fires.

[0003] Traditional smoke detectors experience a significant performance drop or even malfunction in low-light or dark environments, and are susceptible to interference from dust, moisture, and other contaminants, resulting in a high false alarm rate. Image recognition-based solutions also largely rely on ambient light or infrared illumination, rendering them ineffective or significantly reducing their accuracy in the dark. Summary of the Invention

[0004] The purpose of this invention is to provide a smoke alarm method that aims to improve the poor accuracy of existing smoke alarms and image recognition-based smoke alarm systems in low-light or no-light environments.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A smoke alarm method includes: Laser irradiation and image acquisition: The detection area is set, the image acquisition device acquires the background image, the laser is used as a light source to irradiate the detection area, and the image acquisition device acquires the laser light path image within the detection area in real time. Image processing involves differential processing of the laser path image and the background image, followed by HSV color space conversion to extract the laser color region. Morphological processing is then performed on the laser color region to remove noise and connect broken areas, resulting in a binary mask image representing the laser scattering region, where the laser color region is represented by white pixels. Feature extraction: Based on the binary mask image, linearity L_n is calculated for the laser scattering region, white pixel ratio w_r_n is calculated for the laser optical path image in the current frame, and white pixel ratio change rate w_r_c_n is calculated according to the white pixel ratio of consecutive frames. The beam recognition and diffusion recognition paths are used to determine the smoke concentration. Light pillar recognition determination path: When the calculated value of linearity L_n is greater than the preset linearity threshold L_th, and the proportion of white pixels w_r_n is greater than the preset first pixel proportion threshold w_r_th1, it is determined that there is smoke. Diffusion identification and judgment path: When the proportion of white pixels w_r_n is greater than the preset second pixel proportion threshold w_r_th2, and the rate of change of the proportion of white pixels w_r_c_n is greater than the preset rate of change threshold w_r_c_th, it is judged that there is smoke; An alarm will be triggered if either the light beam recognition path or the diffusion recognition path meets the specified conditions.

[0006] Furthermore, the image processing specifically includes the following steps: The current frame image is subtracted from the background image to eliminate static interference. The RGB image is converted to HSV space, and the hue range (H), saturation range (S), and brightness range (V) are set. The laser color region is extracted, and opening operations are performed to remove noise. Closing operations are performed to connect broken regions and generate the laser scattering region.

[0007] Furthermore, the laser is a red laser, satisfying the following condition: H∈[0,10]∪[170,180]; S>50; V>50.

[0008] Furthermore, the linearity L_n calculation for the laser scattering region includes the following steps: Input the binary mask image, extract the coordinates of all white pixels within the binary mask image, and determine if the number of coordinate points is greater than or equal to 'a'. If not, output a linearity of 0.0; otherwise, use the RANSAC algorithm to generate a fitted line, calculate the average distance between the fitted line and all white pixels, and obtain the farthest distance between the fitted line and the farthest white pixel. Output the linearity L_n. L_n = 1 - (average distance / farthest distance).

[0009] Furthermore, the percentage of white pixels w_r_n is the number of white pixels in the binary mask image / the total number of pixels in the binary mask image.

[0010] Furthermore, calculating the rate of change of white pixel proportion w_r_c_n based on the proportion of white pixels in consecutive frames specifically includes the following steps: Calculate the arithmetic mean of the percentage of white pixels in the 20 consecutive frames preceding the current frame, white_ratio_20. , where w_r_ni The percentage of white pixels in the i-th frame; Calculate the rate of change of the proportion of white pixels, w_r_c_n. .

[0011] Furthermore, the light beam recognition and determination specifically includes the following steps: Set the linearity threshold L_th to 0.9 and the first pixel proportion threshold w_r_th1 to 0.0001; When the linearity L_n is greater than the linearity threshold L_th and the proportion of white pixels w_r_n is greater than the first pixel proportion threshold w_r_th1, the condition is met and the low-concentration smoke is identified.

[0012] Furthermore, the diffusion identification and determination specifically includes the following steps: Set the second pixel proportion threshold w_r_th2 to 0.0005 and the change rate threshold w_r_c_th to 0.1; If the proportion of white pixels w_r_n is greater than the second pixel proportion threshold w_r_th2 and the rate of change of the proportion of white pixels w_r_c_n is greater than the rate of change threshold w_r_c_th, then the condition is met and the output is "Identify high concentration of smoke".

[0013] Furthermore, the alarm includes the following steps: An alarm is triggered if any path in the light beam recognition or diffusion recognition path meets the specified conditions. An alarm is triggered when three consecutive frames of images meet the alarm conditions.

[0014] Furthermore, the laser is positioned on one side of the detection area, and the laser irradiation direction forms an angle α with the acquisition direction of the image acquisition device, satisfying the following condition: 60° < α < 120°.

[0015] By adopting the above technical solution, the present invention has the following advantages compared with the prior art: The laser illuminates the detection area, and the image acquisition device captures the laser beam path. By actively creating a visible reference beam path in low-light or no-light environments using an active laser light source, the traditional image recognition method overcomes the problems of insufficient ambient light leading to a lack of detection benchmark and decreased recognition accuracy under low-light conditions. The smoke characteristics are captured in three dimensions—shape, region, and dynamics—by linearity, white pixel ratio, and the change in white pixel ratio. Dual-path determination is performed: the integrity of the laser beam path is used to determine the low-concentration smoke state, and the diffusion of the laser beam path is used to determine the high-concentration smoke state, ensuring reliability across the entire concentration range. Attached Figure Description

[0016] Figure 1 This is a flowchart of the smoke alarm method described in this invention; Figure 2 A summary diagram of the correct smoke detection in scenario 1 of the smoke alarm method described in this invention; Figure 3 This is a summary diagram of the missed frame determination in scenario 1 of the smoke alarm method described in this invention; Figure 4 A summary diagram showing the correct smoke detection in scenario 2 of the smoke alarm method described in this invention; Figure 5 This is a summary diagram of the missed frame determination in scenario 2 of the smoke alarm method described in this invention; Figure 6 A summary diagram of the correct smoke detection in scenario 3 of the smoke alarm method described in this invention; Figure 7 This is a summary diagram of the missed frame determination in scenario 3 of the smoke alarm method described in this invention; Figure 8 This is a schematic diagram of the hardware setup for the smoke alarm method described in this invention.

[0017] Explanation of reference numerals in the attached figures: 1. Detection area; 2. Image acquisition device; 3. Laser generator. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Additionally, it should be noted that the terms "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer" are all based on the orientation or positional relationship shown in the accompanying drawings. They are merely for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element of the present invention must have a specific orientation. Therefore, they should not be construed as limitations on the present invention.

[0020] When an element is referred to as being "fixed to," "set on," or "contained on" another element, it can be directly on or indirectly on that other element. When an element is referred to as being "connected to," it can be directly connected to or indirectly connected to that other element.

[0021] Unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. Example

[0022] Please refer to Figure 1 As shown, this embodiment provides a smoke alarm method, including laser irradiation and image acquisition, image processing, feature extraction, light beam recognition path determination, and diffusion recognition path determination to identify smoke concentration and trigger an alarm. Specifically, the laser irradiation and image acquisition step involves setting a detection area 1, an image acquisition device 2 acquiring a background image, using a laser as a light source to irradiate the detection area 1, and the image acquisition device 2 acquiring real-time images of the laser light path within the detection area 1. In this embodiment, a laser generator 3 emits a laser beam to irradiate the detection area 1.

[0023] The image processing steps are as follows: after performing differential processing on the laser path image and the background image, HSV color space conversion is performed to extract the laser color region, morphological processing is performed on the laser color region to remove noise and connect the broken areas to obtain a binary mask image representing the laser scattering region, where the laser color region is represented by white pixels.

[0024] The feature extraction steps are as follows: based on the binary mask image, the linearity L_n of the laser scattering region is calculated, the proportion of white pixels w_r_n of the current frame is calculated for the laser optical path image, and the rate of change of white pixel proportion w_r_c_n is calculated according to the proportion of white pixels in consecutive frames.

[0025] The specific steps for identifying smoke concentration using the light beam recognition path and diffusion recognition path are as follows: Light pillar recognition determination path: When the calculated value of linearity L_n is greater than the preset linearity threshold L_th, and the proportion of white pixels w_r_n is greater than the preset first pixel proportion threshold w_r_th1, it is determined that there is smoke; Diffusion recognition determination path: When the proportion of white pixels w_r_n is greater than the preset second pixel proportion threshold w_r_th2, and the rate of change of white pixel proportion w_r_c_n is greater than the preset rate of change threshold w_r_c_th, it is determined that there is smoke.

[0026] The alarm procedure is as follows: if either the light beam recognition path or the diffusion recognition path meets the conditions, an alarm will be triggered.

[0027] A laser illuminates the detection area 1, and the image acquisition device 2 captures images of the laser beam path. This utilizes the laser to address the issue of poor reference accuracy in low-light or no-light environments. Smoke characteristics are captured in three dimensions—shape, region, and dynamics—by analyzing linearity, white pixel ratio, and changes in white pixel ratio. A dual-path determination is performed: the integrity of the laser beam is used to determine low-concentration smoke, while the diffusion of the laser beam is used to determine high-concentration smoke, ensuring reliability across the entire concentration range. Through multi-parameter fusion and a dual-path determination mechanism, smoke is effectively distinguished from interfering substances such as dust and moisture. In this embodiment, a conventional laser pointer is used, and the image acquisition device 2 is a mobile phone camera or surveillance camera. No specialized wiring is required, allowing for rapid deployment in warehouses, garages, basements, and other similar locations. This approach offers low hardware costs and high deployment flexibility.

[0028] Specifically, image processing includes the following steps: The current frame image is subtracted from the background image to eliminate static interference and obtain the laser RGB image. The RGB image is then converted to HSV color space, and the hue (H), saturation (S), and brightness (V) ranges are set. The laser color region is extracted, and opening operations are performed to remove noise. Closing operations are then performed to connect broken regions, generating a binary mask image containing the laser scattering region. Background subtraction, HSV color space conversion, and morphological processing ensure the accuracy of the generated laser scattering region.

[0029] In this embodiment, the laser is a red laser; specifically, the laser pointer is a red laser pointer with a wavelength of 650nm and a power of 5mW, which meets human eye safety standards. The HSV color space conversion process satisfies the following conditional expression: H∈[0,10]∪[170,180];S>50;V>50, accurately extracting the red laser region.

[0030] Specifically, the linearity L_n calculation for the laser scattering region includes the following steps: Input a binary mask image, extract the coordinates of all white pixels within the image, and determine if the number of coordinate points is greater than or equal to 'a'. If not, output a linearity of 0.0; otherwise, use the RANSAC algorithm to generate a fitted line, calculate the average distance between the fitted line and all white pixels, and obtain the farthest distance between the fitted line and the farthest white pixel. Output the linearity L_n, where L_n = 1 - (average distance / farthest distance). The linearity L_n is used to determine the laser path shape, quantify the distribution characteristics of the laser region, and calculate the degree of fit between the laser pixels and the optimal fitted line model using the RANSAC algorithm. This effectively identifies straight light column shapes and provides a basis for judging laser scattering in low-concentration smoke environments.

[0031] Specifically, the white pixel percentage w_r_n is the number of white pixels in the binary mask image divided by the total number of pixels in the binary mask image. The white pixel percentage w_r_n serves as a regional judgment criterion to assess the significance of laser scattering effects in the acquired image.

[0032] Specifically, calculating the rate of change of white pixel proportion w_r_c_n based on the proportion of white pixels in consecutive frames includes the following steps: Calculate the arithmetic mean of the percentage of white pixels in the 20 consecutive frames preceding the current frame, white_ratio_20. , where w_r_n i The percentage of white pixels in the i-th frame; Calculate the rate of change of the proportion of white pixels, w_r_c_n. The acquisition of the white pixel ratio change rate w_r_c_n provides a basis for dynamic judgment to distinguish between static and dynamic scenes. In static scene judgment, the white pixel ratio change rate w_r_c_n approaches 0, indicating that the current laser scattering area area has no significant change from the recent historical average. In dynamic smoke judgment, when smoke begins to be generated and spread, the white pixel ratio w_r_n of the current frame is either significantly higher or lower than the recent historical average, resulting in the calculated white pixel ratio change rate w_r_c_n being significantly greater than 0. This effective jump in value is used as the key basis for judging the scene as dynamic and containing smoke. The dynamic change of detection area 1 is obtained by calculating the ratio of the white pixel ratio of the current frame to the historical average of the air intake. In this embodiment, the arithmetic mean of the white pixel ratio of the previous 20 consecutive images is selected to establish a reliable estimate of the recent background area, while also being able to quickly respond to changes in laser diffusion.

[0033] In this embodiment, the light column recognition and determination specifically includes the following steps: Set the linearity threshold L_th to 0.9 and the first pixel proportion threshold w_r_th1 to 0.0001; When the linearity L_n is greater than the linearity threshold L_th and the proportion of white pixels w_r_n is greater than the first pixel proportion threshold w_r_th1, the condition is met and the low-concentration smoke is identified.

[0034] The diffusion identification and determination process specifically includes the following steps: Set the second pixel proportion threshold w_r_th2 to 0.0005 and the change rate threshold w_r_c_th to 0.1; If the proportion of white pixels w_r_n is greater than the second pixel proportion threshold w_r_th2 and the rate of change of the proportion of white pixels w_r_c_n is greater than the rate of change threshold w_r_c_th, then the condition is met and the output is "Identify high concentration of smoke".

[0035] Furthermore, the alarm includes the following steps: An alarm is triggered if any path in the light beam recognition or diffusion recognition path meets the specified conditions. An alarm is triggered when three consecutive frames of images meet the alarm conditions. Performing a three-frame check before triggering an alarm effectively avoids momentary interference.

[0036] Please refer to the appendix. Figure 8 Specifically, the laser is positioned on one side of the detection area 1, and the laser irradiation direction forms an angle α with the acquisition direction of the image acquisition device 2, satisfying the following conditions: 60° < α < 120°. In this embodiment, the included angle α is 90°. Similarly, the included angle α can also be set to 60° or 120°. Limiting the included angle α ensures that the laser irradiation path can be better captured by the image acquisition device 2, avoiding the situation where the laser light path in the acquired image is too small due to an excessively large or small angle, which would affect the accuracy of smoke detection.

[0037] Three sets of scene experiments were conducted with the following premises: linearity threshold L_th = 0.9, first pixel proportion threshold w_r_th1 = 0.0001, second pixel proportion threshold w_r_th2 = 0.0005, and rate of change threshold w_r_c_th = 0.1.

[0038] Scene 1: Initially, the laser strikes a rough wall surface; Scenario 2: Initially, the laser hits a smooth wall surface; Scenario 3: Initially, the laser point is not in the camera's view, but the laser path will pass through the camera's view.

[0039] The above three scenarios were tested in both smoky and smoke-free states. In the smoke-free experiments, no alarms or false alarms were generated. In the smoky experiments, the detection data statistics for Scenario 1 are shown in Table 1. Since no smoke-free scenes were detected as having smoke, the detection accuracy was 100%. However, due to low smoke concentration or linearity not reaching the threshold, some scenes with smoke were not identified.

[0040] Table 1: Statistics of Smoke Test Results in Scenario 1 Total frames 5824 Analyze the total number of frames in the video. Frames with smoke 3666 The smoke begins and the smoke stops. Smokeless frame rate 2158 The period from the start of video recording to the generation of smoke is the smoke-free phase. Correct alarm frame count (TP) 3622 The image showing smoke was determined to contain smoke. Missed Frames (FN) 44 The image showing smoke was determined to be smoke-free. False positives (FP) 0 The smoke-free scene was identified as having smoke. Precision 100.0% TP / (TP+FP) Recall rate 98.8% TP / (TP+FN) Example of correctly detecting smoke: Figure 2 As shown, where Figure 2The sub-images with the numbers (1), (2), (8), (9), and (10) are the results determined to be smoke-containing based on the detection of light pillars. Figure 2 The sub-images labeled (3), (4), (5), (6), and (7) were determined to be smoke-containing based on the smoke's movement. No smoke-free scenes were identified as smoke-containing throughout the process, therefore there were no false alarms. Examples of some missed frames in Scene 1 are shown below. Figure 3 As shown, where Figure 3 The sub-image with the number (1) has thin smoke, but the linear detection score is 0.78, which is less than 0.9, and the ratio=0.0001 does not reach the w_r_th2 threshold, so dynamic detection is not performed. Figure 3 The reason why no smoke was detected in the sub-images with the numbers (3), (4) and (5) is that the smoke diffused to a certain extent, the fluidity became poor, and no straight line of light column was detected.

[0041] The detection data statistics for Scenario 2 are shown in Table 2. Since no smoke-free scenes were detected as smoke-containing, the detection accuracy was 100%. Examples of correctly detected smoke in Scenario 2 are shown below. Figure 4 As shown, where, Figure 4 The sub-images with the numbers (1), (2), (7), (8), (9), and (10) are the results determined to be smoke based on the detection of light pillars. Figure 4 Sub-images numbered (3), (4), (5), and (6) were determined to be smoke-containing based on the movement of smoke. No smoke-free scenes were identified as smoke-containing throughout the process, therefore there were no false alarms. Examples of some missed frames in Scene 2 are shown below. Figure 5 As shown, where, Figure 4 The sub-image with the central label (1) has thin smoke, but the linear detection score is 0.57, which is less than 0.9, and the ratio=0.0003, which is less than the w_r_th2 threshold, so dynamic detection is not performed. Figure 5 The reason why smoke was not detected in the sub-images with the numbers (3), (4) and (5) in these three frames is that the smoke diffused to a certain extent, the fluidity became poor, and the straight line of the light column was not detected.

[0042] Table 2: Statistics of Smoke Test Results in Scenario 2 Total frames 3511 Analyze the total number of frames in the video. Frames with smoke 2100 The smoke begins and the smoke stops. Smokeless frame rate 1411 The period from the start of video recording to the generation of smoke is the smoke-free phase. Correct alarm frame count (TP) 2010 The image showing smoke was determined to contain smoke. Missed Frames (FN) 90 The image showing smoke was determined to be smoke-free. False positives (FP) 0 The smoke-free scene was identified as having smoke. Precision 100.0% TP / (TP+FP) Recall rate 95.71% TP / (TP+FN) The statistical results of the detection data for Scenario 3 are shown in Table 3. Examples of correctly detected smoke in Scenario 3 are shown below. Figure 6 As shown, smoke was detected in all scenes using a linear detection method. Examples where smoke was present but not detected include... Figure 7 As shown, the reason why no smoke was detected in this test was because the smoke concentration was low, and the linearity did not reach the threshold.

[0043] Table 3: Statistics of Smoke Test Results in Scenario 3 Total frames 4198 Analyze the total number of frames in the video. Frames with smoke 2302 The smoke begins and the smoke stops. Smokeless frame rate 1896 The period from the start of video recording to the generation of smoke is the smoke-free phase. Correct alarm frame count (TP) 2158 The image showing smoke was determined to contain smoke. Missed Frames (FN) 144 The image showing smoke was determined to be smoke-free. False positives (FP) 0 The smoke-free scene was identified as having smoke. Precision 100.0% TP / (TP+FP) Recall rate 93.74% TP / (TP+FN) The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method of smoke alarm comprising, include: Laser irradiation and image acquisition: The detection area is set, the image acquisition device acquires the background image, the laser is used as a light source to irradiate the detection area, and the image acquisition device acquires the laser light path image within the detection area in real time. Image processing involves differential processing of the laser path image and the background image, followed by HSV color space conversion to extract the laser color region. Morphological processing is then performed on the laser color region to remove noise and connect broken areas, resulting in a binary mask image representing the laser scattering region, where the laser color region is represented by white pixels. Feature extraction: Based on the binary mask image, linearity L_n is calculated for the laser scattering region, white pixel ratio w_r_n is calculated for the laser optical path image in the current frame, and white pixel ratio change rate w_r_c_n is calculated according to the white pixel ratio of consecutive frames. The beam recognition and diffusion recognition paths are used to determine the smoke concentration. Light pillar recognition determination path: When the calculated value of linearity L_n is greater than the preset linearity threshold L_th, and the proportion of white pixels w_r_n is greater than the preset first pixel proportion threshold w_r_th1, it is determined that there is smoke. Diffusion identification and judgment path: When the proportion of white pixels w_r_n is greater than the preset second pixel proportion threshold w_r_th2, and the rate of change of the proportion of white pixels w_r_c_n is greater than the preset rate of change threshold w_r_c_th, it is judged that there is smoke; An alarm will be triggered if either the light beam recognition path or the diffusion recognition path meets the specified conditions.

2. The smoke alarm method of claim 1, wherein, The image processing specifically includes the following steps: The current frame image is subtracted from the background image to eliminate static interference. The RGB image is converted to HSV space, and the hue range (H), saturation range (S), and brightness range (V) are set. The laser color region is extracted, and opening operations are performed to remove noise. Closing operations are performed to connect broken regions and generate the laser scattering region.

3. The smoke alarm method according to claim 1, characterized in that, The laser is a red laser, and satisfies the following condition: H∈[0,10]∪[170,180]； S>50； V>50。 4. The smoke alarm method according to claim 1, characterized in that, The linearity L_n calculation for the laser scattering region includes the following steps: Input the binary mask image, extract the coordinates of all white pixels within the binary mask image, and determine if the number of coordinate points is greater than or equal to 'a'. If not, output a linearity of 0.0; otherwise, use the RANSAC algorithm to generate a fitted line, calculate the average distance between the fitted line and all white pixels, and obtain the farthest distance between the fitted line and the farthest white pixel. Output the linearity L_n. L_n = 1 - (average distance / farthest distance).

5. The smoke alarm method according to claim 1, characterized in that, The percentage of white pixels, w_r_n, is the number of white pixels in the binary mask image divided by the total number of pixels in the binary mask image.

6. The smoke alarm method according to claim 1, characterized in that, The calculation of the rate of change of white pixel proportion w_r_c_n based on the proportion of white pixels in consecutive frames specifically includes the following steps: Calculate the arithmetic mean of the percentage of white pixels in the 20 consecutive frames preceding the current frame, white_ratio_20. , where w_r_n i The percentage of white pixels in the i-th frame; Calculate the rate of change of the proportion of white pixels, w_r_c_n. 。 7. The smoke alarm method according to claim 1, characterized in that, The specific steps for light beam recognition and determination are as follows: Set the linearity threshold L_th to 0.9 and the first pixel proportion threshold w_r_th1 to 0.0001; When the linearity L_n is greater than the linearity threshold L_th and the proportion of white pixels w_r_n is greater than the first pixel proportion threshold w_r_th1, the condition is met and the low-concentration smoke is identified.

8. The smoke alarm method according to claim 1, characterized in that, The diffusion identification and determination process includes the following steps: Set the second pixel proportion threshold w_r_th2 to 0.0005 and the change rate threshold w_r_c_th to 0.1; If the proportion of white pixels w_r_n is greater than the second pixel proportion threshold w_r_th2 and the rate of change of the proportion of white pixels w_r_c_n is greater than the rate of change threshold w_r_c_th, then the condition is met and the output is "Identify high concentration of smoke".

9. The smoke alarm method according to claim 1, characterized in that, The alarm includes the following steps: An alarm is triggered if any path in the light beam recognition or diffusion recognition path meets the specified conditions. An alarm is triggered when three consecutive frames of images meet the alarm conditions.

10. The smoke alarm method according to claim 1, characterized in that, The laser is positioned on one side of the detection area, and the laser irradiation direction forms an angle α with the acquisition direction of the image acquisition device, satisfying the following conditions: 60°<α<120°。

Images