Forest fire smoke automatic identification system and method based on adaptive target detection

Abstract

The invention discloses a forest fire smoke automatic identification system and method based on adaptive target detection, and belongs to the technical field of forestry engineering. Comprising the following steps that a plurality of fire smoke information monitoring units are arranged in a forest area to be monitored, forest state information is collected, and the forest state information comprises a visible light image and a thermal infrared image; performing adaptive adjustment on the initial feature threshold value based on the average brightness in the acquired visible light image to obtain an adaptive feature threshold value, and calculating image feature data at the same time; making a smoke identification standard, marking the pixel points meeting the smoke identification standard, marking an image area formed by the marked pixel points as a smoke key area, extracting the smoke key area, and obtaining an area temperature value through a thermal infrared image; and generating a fire smoke evaluation value based on the texture features, the shape features and the corresponding temperature values of the smoke key areas, and determining whether the smoke key areas are forest fire smoke based on the fire smoke evaluation value.

Description

Technical Field

[0001] The invention relates to the technical field of forestry engineering, and in particular to an automatic identification system and method for forest fire smoke based on adaptive target detection. Background Art

[0002] In the face of frequent forest fires, timely and accurate detection and assessment of smoke presence has become an important task in forest fire prevention. Traditional fire monitoring methods mainly rely on manual inspections and fixed monitoring points. This method is not only time-consuming and labor-intensive, but also difficult to achieve real-time comprehensive coverage due to the vastness and complexity of the forest, which can easily lead to failure in early detection of fires. Especially in the case of changing weather conditions and complex terrain, the limitations of manual inspections are becoming more and more apparent, and cannot meet the needs of modern forest fire prevention. Therefore, the development of an efficient and automated monitoring system has become an urgent need.

[0003] Existing forest fire monitoring technologies mainly rely on single sensor data, such as visible light cameras or infrared sensors, which leads to instability in detection results under complex meteorological conditions (such as strong light, haze, rain and fog, etc.). Visible light images are easily interfered with under strong sunlight or insufficient ambient light, resulting in false alarms or missed alarms. On the other hand, although thermal infrared sensors can effectively capture temperature changes, their ability to detect smoke formed under low temperature conditions is limited. Therefore, existing technologies are insufficient in data fusion and comprehensive analysis capabilities, making it difficult to achieve efficient smoke detection and fire warning.

[0004] In the prior art, the publication number CN112686190A discloses a method for automatic identification of forest fire smoke based on adaptive target detection, which specifically includes obtaining calibrated image data from a pre-established forest fire smoke image database as a target training image; inputting the target training image into a pre-defined fire smoke detection and recognition model, and extracting image features through a region candidate network in the model; selecting indicators in a pre-established indicator model library, and evaluating the recognition accuracy of image feature extraction. The method is based on an improved adaptive target detection network, which effectively improves the ability to extract forest fire smoke features, thereby realizing real-time and high-precision detection and identification of forest fire smoke. However, this method relies on a pre-established forest fire smoke image database. If the image samples in the database are not sufficient to cover different environmental conditions, lighting changes, smoke types and background conditions, the model may perform poorly in certain situations. Low diversity and representativeness will affect the generalization ability of the model, resulting in false positives or negative negatives in actual monitoring. At the same time, the model cannot continue to learn and optimize based on newly collected data, which may cause its performance to decline over time.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a forest fire smoke automatic identification system and method based on adaptive target detection to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An automatic forest fire smoke recognition system based on adaptive target detection, specifically comprising:

[0009] An information collection module is used to monitor the entire forest area to be monitored, and is composed of a number of fire smoke information monitoring units for collecting forest status information, wherein the fire smoke information monitoring units include visual sensors and thermal infrared sensors, and the forest status information includes visible light images and thermal infrared images;

[0010] An adaptive detection adjustment module is used to adaptively adjust the initial feature threshold based on the average brightness in the visible light image to obtain an adaptive feature threshold, and calculate the image feature data of each pixel based on the visible light image, wherein the image feature data includes the hue, saturation and brightness of the pixel, and the initial feature threshold includes a hue range threshold, a saturation threshold and a brightness threshold;

[0011] The key area identification and extraction module is used to formulate smoke identification standards based on adaptive feature thresholds, identify pixels in the visible light image whose image feature data meets the smoke identification standards, record the image area composed of the identified pixels as the smoke key area, extract the smoke key area through the contour detection algorithm, and obtain the temperature value of the smoke key area through the thermal infrared image;

[0012] The smoke identification and determination module is used to generate a fire smoke assessment value based on the texture features, shape features and corresponding temperature values ​​of the smoke key area, and determine whether the smoke key area is forest fire smoke based on the fire smoke assessment value, the texture features include contrast and energy, and the shape features include contour simplicity and roundness.

[0013] Furthermore, the visual sensor and thermal infrared sensor are a high-definition camera and a thermal infrared camera, which collect real-time video of the forest area to be monitored through the high-definition camera and the thermal infrared camera, and obtain forest status information through the real-time video, including visible light images and thermal infrared images, wherein the method for obtaining visible light images and thermal infrared images is to collect visible light and thermal infrared frame images at the same time from the real-time video, and the collection interval is 30 frames.

[0014] Further, the initial feature threshold is adaptively adjusted according to the average brightness to obtain the adaptive feature threshold, which includes the following steps: calculating the average brightness L of the current visible light image according to the brightness of each pixel in the collected visible light image; converting the visible light image from the RGB color space to the HSV color space, and calculating the image feature data, wherein the image feature data includes the hue H, saturation S and lightness V of the pixel;

[0015] The initial feature threshold is adaptively adjusted according to the average brightness, and the formula for the adaptive feature threshold is obtained as follows:

[0016] H final =[H base (0)+α H *L,H base (1)-β H *L]

[0017] S final =S base -γ S *L

[0018] V final =V base +δ V *L

[0019] In the formula, H final Represents the adaptive hue range threshold, S final Represents the adaptive saturation threshold, V final Represents the adaptive brightness threshold, and the adaptive feature threshold includes H final , S final and V final , H base (0) and H base (1) represents the lower and upper limits of the hue range in the initial feature threshold, S base is the saturation threshold in the initial feature threshold, V base is the brightness threshold in the initial feature threshold, where α H , β H , γ S and δ V They represent the lower and upper limits of the hue range, the saturation threshold, and the adjustment coefficients of the brightness threshold respectively. L is the average brightness of the current visible light image.

[0020] Furthermore, the average brightness L in the collected visible light image is calculated according to the formula:

[0021]

[0022] Where M represents the total number of pixels in the visible light image, Ip Represents the brightness value of the pth pixel. The first pixel in the upper left corner of the visible light image is denoted as the first pixel. The numbers are arranged in order of row priority to obtain a total of M pixels, where p is the index of the pixel, p = 1, 2, ..., M;

[0023] The brightness value of the pth pixel I p Calculated by the following formula:

[0024] I p =0.299R p +0.587G p +0.114B p

[0025] In the formula, R p Represents the intensity value of the red channel of the pth pixel, G p Indicates the intensity value of the green channel of the pth pixel, B p Represents the intensity value of the blue channel of the p-th pixel.

[0026] Furthermore, converting the visible light image from the RGB color space to the HSV color space and calculating the image feature data includes the following steps:

[0027] The image feature data is calculated based on the intensity values ​​of the red, green, and blue channels of each pixel. The specific calculation formula is as follows:

[0028]

[0029] In the formula, atan2 is the inverse tangent function, which is used to calculate the inverse tangent value of two variables, H p , S p and V p are the hue, saturation, and brightness of the pth pixel in the visible light image, respectively.

[0030] Furthermore, a smoke recognition standard is formulated according to the adaptive feature threshold, and the logic of the smoke recognition standard is as follows:

[0031] When the pixel meets the following conditions:

[0032] H p ∈H final

[0033] S p final

[0034] V p >V final

[0035] ​Mark the pixel point, and denote the image area composed of all pixel points that fully meet the smoke recognition standard as the key smoke area. Then, based on the Canny edge detection algorithm, extract the key smoke area, and obtain the temperature value corresponding to the key smoke area in the thermal infrared image. Among them, calculate the average temperature through the temperature values of each pixel point in the key smoke area corresponding to the thermal infrared image, and use this average temperature as the temperature value of the key smoke area.

[0036] Furthermore, the formula for generating the fire smoke evaluation value based on the texture features, shape features, and corresponding temperature values of the key smoke area is as follows:

[0037]

[0038] In the formula, HY represents the fire smoke evaluation value, T is the temperature value of the key smoke area, Ct and En respectively represent the contrast and energy in the texture features, Cm and Cl respectively represent the contour simplicity and roundness in the shape features, and ω1, ω2, ω3, ω4, and ω5 are the weight coefficients of T, En, Ct, Cm, and Cl respectively;

[0039] The logic for determining whether the key smoke area is forest fire smoke based on the fire smoke evaluation value is: set the fire smoke threshold as yz;

[0040] When HY < yz, it is determined that the current key smoke area is not forest fire smoke;

[0041] When HY ≥ yz, it is determined that the current key smoke area is forest fire smoke, and an alarm is issued.

[0042] Furthermore, the steps for obtaining the shape features of the texture features of the key smoke area include: calculating the texture features of the key smoke area including contrast and energy through the gray-level co-occurrence matrix, and calculating the shape features including contour simplicity and roundness based on the contour length and area of the key smoke area.

[0043] The present invention also provides a method for automatically identifying forest fire smoke based on adaptive target detection. The method for automatically identifying forest fire smoke based on adaptive target detection is used to control the above-mentioned system for automatically identifying forest fire smoke based on adaptive target detection. The specific steps include:

[0044] Set several fire smoke information monitoring units in the forest area to be monitored to ensure that the entire forest area to be monitored can be monitored, and collect forest status information. The fire smoke information monitoring unit includes a visual sensor and a thermal infrared sensor. The forest status information includes visible light images and thermal infrared images;

[0045] Based on the average brightness in the visible light image, the initial feature threshold is adaptively adjusted to obtain an adaptive feature threshold, and image feature data of each pixel is calculated based on the visible light image, wherein the image feature data includes the hue, saturation and brightness of the pixel, and the initial feature threshold includes a hue range threshold, a saturation threshold and a brightness threshold;

[0046] The smoke recognition standard is formulated based on the adaptive feature threshold, and the pixel points whose image feature data in the visible light image meets the smoke recognition standard are marked. The image area composed of the marked pixel points is recorded as the smoke key area. The smoke key area is extracted through the contour detection algorithm, and the temperature value of the smoke key area is obtained through the thermal infrared image.

[0047] A fire smoke assessment value is generated based on the texture features, shape features and corresponding temperature values ​​of the smoke key area, and based on the fire smoke assessment value, it is determined whether the smoke key area is forest fire smoke, the texture features include contrast and energy, and the shape features include contour simplicity and roundness.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The combination of visual sensors and thermal infrared sensors ensures coverage of the forest area to be monitored under various meteorological conditions, greatly improving the comprehensiveness and real-time nature of monitoring. The recognition threshold of key smoke areas is adaptively adjusted according to actual brightness conditions, which improves the applicability of the system under complex meteorological conditions (such as strong light, haze, rain and fog, etc.), and determines the key smoke areas at the same time, reducing the computational burden and improving the accuracy of recognition. The temperature information of thermal infrared images and the texture and shape characteristics of key smoke areas are combined to achieve deep fusion of multimodal data, overcoming the limitations of a single sensor under specific environmental conditions, thereby improving the stability and reliability of smoke detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the overall system structure of the present invention;

[0051] Figure 2 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0053] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0054] Example:

[0055] See also Figure 1 , the present invention provides a system structure:

[0056] An automatic forest fire smoke recognition system based on adaptive target detection, specifically comprising:

[0057] An information collection module is used to monitor the entire forest area to be monitored, and is composed of a number of fire smoke information monitoring units for collecting forest status information, wherein the fire smoke information monitoring units include visual sensors and thermal infrared sensors, and the forest status information includes visible light images and thermal infrared images;

[0058] The visual sensor and thermal infrared sensor are a high-definition camera and a thermal infrared camera. Real-time video of the forest area to be monitored is collected by the high-definition camera and the thermal infrared camera. Forest status information including visible light images and thermal infrared images are obtained through the real-time video. The method for obtaining the visible light image and the thermal infrared image is to collect visible light and thermal infrared frame images at the same time from the real-time video, and the collection interval is 30 frames.

[0059] An adaptive detection adjustment module is used to adaptively adjust the initial feature threshold based on the average brightness in the visible light image to obtain an adaptive feature threshold, and simultaneously calculate image feature data of each pixel of the visible light image, wherein the image feature data includes the hue, saturation and brightness of the pixel, and the initial feature threshold includes a hue range threshold, a saturation threshold and a brightness threshold;

[0060] Adaptively adjusting the initial feature threshold according to the average brightness to obtain the adaptive feature threshold includes the following steps: calculating the average brightness L of the current visible light image according to the brightness of each pixel in the collected visible light image; converting the visible light image from the RGB color space to the HSV color space, and calculating image feature data, wherein the image feature data includes the hue H, saturation S and lightness V of the pixel;

[0061] The initial feature threshold is adaptively adjusted according to the average brightness, and the formula for the adaptive feature threshold is obtained as follows:

[0062] H final =[H base (0)+α H *L,H base (1)-β H *L]

[0063] S final =S base -γ S *L

[0064] V final =V base +δ V *L

[0065] In the formula, H final Represents the adaptive hue range threshold, S final Represents the adaptive saturation threshold, V final Represents the adaptive brightness threshold, and the adaptive feature threshold includes H final , S final and V final , H base (0) and H base (1) represents the lower and upper limits of the hue range in the initial feature threshold, S base is the saturation threshold in the initial feature threshold, V base is the brightness threshold in the initial feature threshold, where α H , β H , γ S and δ V They represent the lower and upper limits of the hue range, the saturation threshold, and the adjustment coefficient of the brightness threshold, respectively. L is the average brightness of the current visible light image. The initial feature threshold can be set according to the actual environmental conditions of the monitored forest area and expert experience.

[0066] where α H , β H , γ S and δ V It can be set by expert experience, and the approximate range is generally between 0.01 and 0.3, and can be adjusted according to actual conditions;

[0067] The formula for calculating the average brightness L in the collected visible light image is:

[0068]

[0069] Where M represents the total number of pixels in the visible light image, I p Represents the brightness value of the pth pixel. The first pixel in the upper left corner of the visible light image is denoted as the first pixel. The numbers are arranged in order of row priority to obtain a total of M pixels, where p is the index of the pixel, p = 1, 2, ..., M;

[0070] The brightness value of the pth pixel I p Calculated by the following formula:

[0071] I p =0.299R p +0.587G p +0.114B p

[0072] In the formula, R p Represents the intensity value of the red channel of the pth pixel, G p Indicates the intensity value of the green channel of the pth pixel, B p Represents the intensity value of the blue channel of the p-th pixel.

[0073] Converting a visible light image from RGB color space to HSV color space and calculating image feature data includes the following steps:

[0074] The image feature data is calculated based on the intensity values ​​of the red, green, and blue channels of each pixel. The specific calculation formula is as follows:

[0075]

[0076] In the formula, atan2 is the inverse tangent function, which is used to calculate the inverse tangent value of two variables, H p , S p and V p are the hue, saturation, and brightness of the pth pixel in the visible light image, respectively.

[0077] The key area identification and extraction module is used to formulate smoke identification standards based on adaptive feature thresholds, identify pixels in the visible light image whose image feature data meets the smoke identification standards, record the image area composed of the identified pixels as the smoke key area, extract the smoke key area through the contour detection algorithm, and obtain the temperature value of the smoke key area through the thermal infrared image;

[0078] A smoke recognition standard is formulated according to an adaptive feature threshold, and the logic on which the smoke recognition standard is based is:

[0079] When the pixel meets the following conditions:

[0080] H p ∈H final

[0081] S p final

[0082] V p >V final

[0083] The pixel is marked, and the image area composed of all the pixels that meet the smoke recognition standard is recorded as the smoke key area; the saturation of smoke is usually low. Because the light is scattered by the smoke, the colors in the image are not bright enough and the saturation is reduced. The colors in the smoke area often look pale, blurred, and lack vividness; smoke usually has a high brightness, especially under sunlight, the smoke scatters light and increases the brightness of the area. Therefore, the brightness value of the smoke area is often higher than the surrounding environment.

[0084] Based on the Canny edge detection algorithm, the key smoke area is extracted to obtain the temperature value of the key smoke area in the thermal infrared image. The average temperature is calculated by the temperature value of each pixel in the key smoke area corresponding to the thermal infrared image. The average temperature is used as the temperature value of the key smoke area. The formula for calculating the average temperature is:

[0085]

[0086] In the formula, T represents the mean temperature, Q is the total number of pixels in the key smoke area in the infrared image, and T d is the temperature value of the dth pixel, d is the index of the pixel point in the key smoke area in the infrared image, d=1,2,…,Q.

[0087] The smoke identification and determination module is used to generate a fire smoke assessment value based on the texture features, shape features and corresponding temperature values ​​of the smoke key area, and determine whether the smoke key area is forest fire smoke based on the fire smoke assessment value, the texture features include contrast and energy, and the shape features include contour simplicity and roundness.

[0088] The formula for generating the fire smoke assessment value based on the texture features, shape features and corresponding temperature values ​​of the smoke key areas is:

[0089]

[0090] ​In the formula, HY represents the fire smoke evaluation value, T is the temperature value of the key smoke area, Ct and En respectively represent the contrast and energy in the texture features, Cm and Cl respectively represent the contour simplicity and roundness in the shape features, and ω1, ω2, ω3, ω4, and ω5 are the weight coefficients of T, En, Ct, Cm, and Cl respectively; among them, the temperature value T is directly proportional to HY. The higher the temperature, the greater the probability that the key smoke area is forest fire smoke. Ct and En are inversely proportional to HY. Lower contrast values and energy values are usually associated with smoke areas. Cm and Cl are also inversely proportional to HY. Generally, the smaller the values of Cm and Cl, the more irregular the shape, indicating a greater possibility that the key smoke area is forest fire smoke. And because the temperature value T has the greatest impact on HY, and Ct and En are smaller compared to the temperature value T, so ω1>ω2≥ω3>ω4≥ω5 and ω1, ω2, ω3, ω4, and ω5 are greater than 0.

[0091] The logic for determining whether the key smoke area is forest fire smoke based on the fire smoke evaluation value is: set the fire smoke threshold as yz;

[0092] When HY < yz, it is judged that the current key smoke area is not forest fire smoke;

[0093] When HY ≥ yz, it is judged that the current key smoke area is determined to be forest fire smoke and an alarm is issued.

[0094] The steps for obtaining the shape features of the texture features of the key smoke area include: calculating the texture features of the key smoke area including contrast and energy through the gray-level co-occurrence matrix, and calculating the shape features including contour simplicity and roundness based on the contour length and area of the key smoke area;

[0095] Among them, calculating the texture features of the key smoke area including contrast and energy through the gray-level co-occurrence matrix is a mature existing technology, and the specific calculation method will not be elaborated here; calculating the shape features including contour simplicity and roundness based on the contour length and area of the key smoke area, the specific formulas are respectively:

[0096]

[0097] In the formula, A is the area of the key smoke area, and L represents the contour length of the key smoke area; a smaller value of Cm may indicate a more irregular shape, and a value of Cl close to 1 indicates a more regular shape, and the smaller the value, the more irregular the shape.

[0098] Please refer to Figure 2The present invention also provides a method for automatically identifying forest fire smoke based on adaptive target detection. The method for automatically identifying forest fire smoke based on adaptive target detection is used to control the above-mentioned automatic forest fire smoke identification system based on adaptive target detection. The specific steps include:

[0099] Step 1: several fire smoke information monitoring units are set up in the forest area to be monitored to ensure that the entire forest area to be monitored can be monitored and forest status information is collected. The fire smoke information monitoring units include visual sensors and thermal infrared sensors, and the forest status information includes visible light images and thermal infrared images;

[0100] Step 2: Based on the average brightness in the visible light image, the initial feature threshold is adaptively adjusted to obtain an adaptive feature threshold, and image feature data of each pixel is calculated based on the visible light image, wherein the image feature data includes the hue, saturation and brightness of the pixel, and the initial feature threshold includes a hue range threshold, a saturation threshold and a brightness threshold;

[0101] Step 3: Formulate smoke recognition standards based on adaptive feature thresholds, identify pixels in the visible light image whose image feature data meets the smoke recognition standards, record the image area composed of the identified pixels as the smoke key area, extract the smoke key area through the contour detection algorithm, and obtain the temperature value of the smoke key area through the thermal infrared image;

[0102] Step 4: Generate a fire smoke assessment value based on the texture features, shape features and corresponding temperature values ​​of the smoke key area, and determine whether the smoke key area is forest fire smoke based on the fire smoke assessment value, the texture features include contrast and energy, and the shape features include contour simplicity and roundness.

[0103] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0104] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0105] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0106] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An automatic forest fire smoke recognition system based on adaptive target detection, characterized in that: Specifically include: An information collection module is used to monitor the entire forest area to be monitored, and is composed of a number of fire smoke information monitoring units for collecting forest status information, wherein the fire smoke information monitoring units include visual sensors and thermal infrared sensors, and the forest status information includes visible light images and thermal infrared images; An adaptive detection adjustment module is used to adaptively adjust the initial feature threshold based on the average brightness in the visible light image to obtain an adaptive feature threshold, and simultaneously calculate image feature data of each pixel of the visible light image, wherein the image feature data includes the hue, saturation and brightness of the pixel, and the initial feature threshold includes a hue range threshold, a saturation threshold and a brightness threshold; The key area identification and extraction module is used to formulate smoke identification standards based on adaptive feature thresholds, identify pixels in the visible light image whose image feature data meets the smoke identification standards, record the image area composed of the identified pixels as the smoke key area, extract the smoke key area through the contour detection algorithm, and obtain the temperature value of the smoke key area through the thermal infrared image; The smoke identification and determination module is used to generate a fire smoke assessment value based on the texture features, shape features and corresponding temperature values ​​of the smoke key area, and determine whether the smoke key area is forest fire smoke based on the fire smoke assessment value, the texture features include contrast and energy, and the shape features include contour simplicity and roundness.

2. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 1 is characterized in that: The visual sensor and thermal infrared sensor are a high-definition camera and a thermal infrared camera. Real-time video of the forest area to be monitored is collected by the high-definition camera and the thermal infrared camera. Forest status information including visible light images and thermal infrared images are obtained through the real-time video. The method for obtaining the visible light image and the thermal infrared image is to collect visible light and thermal infrared frame images at the same time from the real-time video, and the collection interval is 30 frames.

3. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 1 is characterized in that: Adaptively adjusting the initial feature threshold according to the average brightness to obtain the adaptive feature threshold includes the following steps: calculating the average brightness L of the current visible light image according to the brightness of each pixel in the collected visible light image; converting the visible light image from the RGB color space to the HSV color space, and calculating image feature data, wherein the image feature data includes the hue H, saturation S and lightness V of the pixel; The initial feature threshold is adaptively adjusted according to the average brightness, and the formula for the adaptive feature threshold is obtained as follows: H final =[H base (0)+a H *L, H base (1)-b H *L] S final =S base -c S *L V final =V base +δ V *L In the formula, H final Represents the adaptive hue range threshold, S final Represents the adaptive saturation threshold, V final Represents the adaptive brightness threshold, and the adaptive feature threshold includes H final , S final and V final , H base (0) and H base (1) represents the lower and upper limits of the hue range in the initial feature threshold, S base is the saturation threshold in the initial feature threshold, V base is the brightness threshold in the initial feature threshold, where α H , β H , γ S and δ V They represent the lower and upper limits of the hue range, the saturation threshold, and the adjustment coefficients of the brightness threshold respectively. L is the average brightness of the current visible light image.

4. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 3 is characterized in that: The formula for calculating the average brightness L in the collected visible light image is: Where M represents the total number of pixels in the visible light image, I p Represents the brightness value of the pth pixel. The first pixel in the upper left corner of the visible light image is denoted as the first pixel. The numbers are arranged in order of row priority to obtain a total of M pixels, where p is the index of the pixel, p = 1, 2, ..., M; The brightness value of the pth pixel I p Calculated by the following formula: I p =0.299R p +0.587G p +0.114B p In the formula, R p Represents the intensity value of the red channel of the pth pixel, G p Indicates the intensity value of the green channel of the pth pixel, B p Represents the intensity value of the blue channel of the p-th pixel.

5. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 4 is characterized in that: Converting a visible light image from RGB color space to HSV color space and calculating image feature data includes the following steps: The image feature data is calculated based on the intensity values ​​of the red, green, and blue channels of each pixel. The specific calculation formula is as follows: In the formula, atan2 is the inverse tangent function, which is used to calculate the inverse tangent value of two variables, H p , S p and V p are the hue, saturation, and brightness of the pth pixel in the visible light image, respectively.

6. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 5 is characterized in that: A smoke recognition standard is formulated according to an adaptive feature threshold, and the logic on which the smoke recognition standard is based is: When the pixel meets the following conditions at the same time: H p ∈H final S p <S final V p >V final Mark the pixel point, and denote the image area composed of all pixel points that fully meet the smoke recognition criteria as the key smoke area. Based on the Canny edge detection algorithm, extract the key smoke area, and obtain the temperature value corresponding to the key smoke area in the thermal infrared image. Among them, calculate the average temperature through the temperature values of each pixel point in the key smoke area corresponding to the thermal infrared image, and use this average temperature as the temperature value of the key smoke area.

7. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 1 is characterized in that: The formula for generating the fire smoke evaluation value based on the texture feature, shape feature, and corresponding temperature value of the key smoke area is: In the formula, HY represents the fire smoke evaluation value, T is the temperature value of the key smoke area, Ct and En respectively represent the contrast and energy in the texture feature, Cm and Cl respectively represent the contour simplicity and roundness in the shape feature, and ω1, ω2, ω3, ω4, and ω5 are the weight coefficients of T, En, Ct, Cm, and Cl respectively, where ω1 > ω2 ≥ ω3 > ω4 ≥ ω5 and ω1, ω2, ω3, ω4, and ω5 are all greater than 0; The logic for determining whether the key smoke area is forest fire smoke based on the fire smoke evaluation value is: Set the fire smoke threshold as yz; When HY < yz, it is determined that the current key smoke area is not forest fire smoke; When HY ≥ yz, it is determined that the current key smoke area is forest fire smoke, and an alarm is issued.

8. The automatic forest fire smoke recognition system based on adaptive target detection according to claim 7 is characterized in that: The steps for obtaining the shape features of the texture features of the key smoke area include: Calculate the texture features of the key smoke area including contrast and energy through the gray-level co-occurrence matrix, and calculate the shape features including contour simplicity and roundness based on the contour length and area of the key smoke area.

9. A method for automatic identification of forest fire smoke based on adaptive target detection, characterized in that: The automatic forest fire smoke recognition method based on adaptive object detection is used to control the automatic forest fire smoke recognition system according to any one of claims 1-8. The specific steps include: Set several fire smoke information monitoring units in the forest area to be monitored to ensure that the entire forest area to be monitored can be monitored, and collect forest status information. The fire smoke information monitoring unit includes a visual sensor and a thermal infrared sensor, and the forest status information includes visible light images and thermal infrared images; Based on the average brightness in the visible light image, adaptively adjust the initial feature threshold to obtain the adaptive feature threshold. At the same time, calculate the image feature data of each pixel point based on the visible light image. The image feature data includes the hue, saturation, and lightness of the pixel point. The initial feature threshold includes the hue range threshold, saturation threshold, and lightness threshold; Based on the adaptive feature threshold, formulate the smoke recognition criteria, mark the pixel points in the visible light image whose image feature data meet the smoke recognition criteria, denote the image area composed of the marked pixel points as the key smoke area, extract the key smoke area through the contour detection algorithm, and obtain the temperature value of the key smoke area through the thermal infrared image; A fire smoke assessment value is generated based on the texture features, shape features and corresponding temperature values ​​of the smoke key area, and based on the fire smoke assessment value, it is determined whether the smoke key area is forest fire smoke, the texture features include contrast and energy, and the shape features include contour simplicity and roundness.

Images