A method for identifying objects in coastal wetland high-definition remote sensing images based on convolutional networks

Abstract

The invention relates to the technical field of remote sensing image processing and object recognition, and in particular to a method for identifying objects in coastal wetlands using high-definition remote sensing images based on a convolutional network. The method is used to solve the problems that the existing method for identifying objects in coastal wetlands using high-definition remote sensing images cannot perform data analysis on feature information of coastal wetland remote sensing images, cannot judge the ecological environment of coastal wetlands, cannot timely discover when coastal wetlands are threatened, and the damage is easy to expand. The method of the invention uses a convolutional neural network to perform object recognition on high-definition remote sensing images of coastal wetlands, can fully extract feature information in images, and improve recognition accuracy. At the same time, by performing data analysis on feature information in images, the ecological stability of coastal wetlands can be monitored in real time, and when coastal wetlands are threatened, they can be discovered and alarmed in time, thereby avoiding causing greater damage, and making an outstanding contribution to the protection of the ecological environment of coastal wetlands.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing and object recognition, and in particular to a method for recognizing objects in coastal wetland high-definition remote sensing images based on a convolutional network. Background Art

[0002] Coastal wetlands are one of the most important ecosystems on Earth, with rich biodiversity and ecological service functions. However, due to the impact of human activities and natural factors, the ecological environment of coastal wetlands is facing serious threats. Therefore, remote sensing monitoring and object identification of coastal wetlands are of great significance for the protection and restoration of wetland ecosystems.

[0003] The patent with application number CN202111465535.2 discloses a dual-stream encoding and decoding method for identifying objects in coastal wetland high-definition remote sensing images. The method combines the densely connected atrous convolutional network with the feature extraction process of the graph convolutional neural network to form a dual-stream hybrid convolutional neural network encoding model to extract the characteristic feature information of the objects in the coastal wetland remote sensing images. The decoding network is further used to decode and predict the accurate and precise identification type of objects in the coastal wetland images based on the feature information extracted by the dual streams. The invention first performs preprocessing operations such as normalization, data enhancement, and cropping on the coastal wetland remote sensing image data; then the preprocessed data is input into the dual-stream hybrid convolutional neural network encoding model for feature extraction and model training; finally, the test samples to be segmented in the data set are input into the trained model, and the model is used to achieve the purpose of accurate object recognition of the coastal wetland remote sensing images, but there are still the following shortcomings: it is impossible to perform data analysis on the feature information of the coastal wetland remote sensing images, it is impossible to judge the ecological environment of the coastal wetlands, and when the coastal wetlands are threatened, they cannot be discovered in time, and the infringement is easy to expand.

[0004] In recent years, convolutional neural networks have achieved great success in the field of image processing and are widely used in the field of image recognition. Therefore, the present invention proposes a method for identifying objects in coastal wetland high-definition remote sensing images based on convolutional networks to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned technical problems, the purpose of the present invention is to provide a method for identifying objects in coastal wetlands using high-definition remote sensing images based on a convolutional network: an image information acquisition model is established through a model building module, a high-definition remote sensing image is obtained through the image acquisition module, the high-definition remote sensing image is input into the image information acquisition model through the model building module, and an information acquisition instruction is generated at the same time. After the information acquisition module receives the information acquisition instruction, the image information acquisition model is used to obtain ecological stability information, an ecological stability coefficient is obtained according to the ecological stability information through the information analysis module, an impact alarm instruction is generated through the ecological stability coefficient of the object recognition platform, and an impact alarm is sounded after the image alarm module receives the impact alarm instruction. This solves the problem that the existing method for identifying objects in coastal wetlands using high-definition remote sensing images cannot perform data analysis on the characteristic information of coastal wetland remote sensing images, cannot judge the ecological environment of coastal wetlands, cannot discover threats to coastal wetlands in time, and the infringement is easy to expand.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for identifying objects in coastal wetland high-definition remote sensing images based on a convolutional network comprises the following steps:

[0008] Step 1: The model building module builds the image information acquisition model;

[0009] Step 2: The image acquisition module obtains high-definition remote sensing images and sends the high-definition remote sensing images to the model building module;

[0010] Step 3: The model building module inputs the high-definition remote sensing image into the image information acquisition model, generates information acquisition instructions, and sends the information acquisition instructions to the information acquisition module;

[0011] Step 4: After receiving the information collection instruction, the information collection module uses the image information collection model to obtain ecological stability information, and sends the ecological stability information to the information analysis module;

[0012] Step 5: The information analysis module obtains the ecological stability coefficient ST according to the ecological stability information, and sends the ecological stability coefficient ST to the ground feature recognition platform;

[0013] Step 6: The ecological stability coefficient ST of the ground feature recognition platform generates an impact alarm instruction and sends the impact alarm instruction to the image alarm module;

[0014] Step 7: After receiving the impact alarm command, the image alarm module sounds the impact alarm.

[0015] As a further solution of the present invention: the specific process of the model building module building the image information acquisition model is as follows:

[0016] The convolutional network model is established using convolutional network technology and marked as an image information acquisition model.

[0017] As a further solution of the present invention: the specific process of the image acquisition module obtaining high-definition remote sensing images is as follows:

[0018] Remote sensing technology is used to collect images of coastal wetlands, obtain high-definition remote sensing images, and send the high-definition remote sensing images to the model building module.

[0019] As a further solution of the present invention: the ecological stability information includes vegetation information ZB, land information TD and animal information DW.

[0020] As a further solution of the present invention: the specific process of the information acquisition module acquiring vegetation information ZB is as follows:

[0021] After receiving the information collection instruction, the image information collection model is used to obtain the vegetation area and average vegetation height in the high-definition remote sensing image, and they are marked as the planting surface value ZM and the planting height value ZG respectively. The planting surface value ZM and the planting height value ZG are quantified, and the values ​​of the planting surface value ZM and the planting height value ZG are extracted and substituted into the formula for calculation. According to the formula The vegetation information ZB is obtained, wherein b1 and b2 are preset proportional coefficients corresponding to the set planting surface value ZM and planting height value ZG respectively, and b1 and b2 satisfy b1+b2=1, 0<b2<b1<1, and b1=0.56 and b2=0.44 are taken.

[0022] As a further solution of the present invention: the specific process of the information acquisition module acquiring the land information TD is as follows:

[0023] After receiving the information collection instruction, the image information collection model is used to obtain the land area collected for the first time in the high-definition remote sensing image and the land area at the current moment, and the difference between the two is obtained, and it is marked as the ground change value DB. The average volume of all peaks collected for the first time in the high-definition remote sensing image and the average volume of all peaks at the current moment are obtained, and the difference between the two is obtained, and it is marked as the peak change value FB. All peak change values ​​FB are summed up to obtain the peak change total value FZ, and the ground change value DB and the peak change total value FZ are quantified. The values ​​of the ground change value DB and the peak change total value FZ are extracted and substituted into the formula for calculation. According to the formula The land information TD is obtained, where t1 and t2 are the preset proportional coefficients corresponding to the set ground change value DB and the peak change total value FZ respectively, and t1 and t2 satisfy t1+t2=1, 0<t1<t2<1, and t1=0.33 and t2=0.67.

[0024] As a further solution of the present invention: the specific process of the information acquisition module acquiring the animal information DW is as follows:

[0025] After receiving the information collection instruction, the image information collection model is used to randomly obtain an area in the high-definition remote sensing image, and it is marked as the monitoring area. The animals in the monitoring area whose volume reaches the preset volume are marked as monitored animals. The total number of initial monitored animals in the preset time period of the monitoring area is obtained, and it is marked as the initial value CS. When the monitored animal enters the monitoring area, the number of monitoring times of the monitored animal is increased by one. The total number of monitoring times of the monitored animals entering the monitoring area within the preset time period is obtained, and it is marked as the measurement value CC. The initial value CS and the measurement value CC are quantified, and the values ​​of the initial value CS and the measurement value CC are extracted and substituted into the formula for calculation. According to the formula The animal information DW is obtained, where d1 and d2 are the preset proportional coefficients corresponding to the set initial value CS and the measured value CC respectively, and d1 and d2 satisfy d1+d2=1, 0<d1<d2<1, and d1=0.19 and d2=0.81.

[0026] As a further solution of the present invention: the specific process of the information analysis module obtaining the ecological stability coefficient ST is as follows:

[0027] The vegetation information ZB, land information TD and animal information DW are quantified, the values ​​of vegetation information ZB, land information TD and animal information DW are extracted, and substituted into the formula for calculation. The ecological stability coefficient ST is obtained, where π is a mathematical constant, s1, s2 and s3 are the preset weight factors corresponding to the set vegetation information ZB, land information TD and animal information DW, s1, s2 and s3 satisfy s2>s1>s3>1.451, and s1=1.95, s2=2.34, s3=1.58;

[0028] The ecological stability coefficient ST is sent to the land feature identification platform.

[0029] As a further solution of the present invention: the specific process of the ground object recognition platform generating an impact alarm instruction is as follows:

[0030] Compare the ecological stability coefficient ST with the preset ecological stability threshold STy:

[0031] If the ecological stability coefficient ST is less than the ecological stability threshold STy, an impact alarm instruction is generated and sent to the image alarm module.

[0032] As a further solution of the present invention: the specific process of the image alarm module sounding the alarm is as follows:

[0033] Get the value of the ecological stability coefficient ST and mark it as the volume adjustment coefficient YT. Substitute the volume adjustment coefficient YT into the formula for calculation. A sound intensity value YQ is obtained, wherein δ is a preset adjustment factor, and the intensity of the sound affecting the alarm is adjusted according to the sound intensity value YQ.

[0034] Beneficial effects of the present invention:

[0035] The invention discloses a method for identifying objects in coastal wetlands using high-definition remote sensing images based on a convolutional network. The method establishes an image information acquisition model through a model establishment module, obtains high-definition remote sensing images through the image acquisition module, inputs the high-definition remote sensing images into the image information acquisition model through the model establishment module, generates information acquisition instructions at the same time, obtains ecological stability information by using the image information acquisition model after receiving the information acquisition instructions through the information acquisition module, obtains an ecological stability coefficient according to the ecological stability information through the information analysis module, generates an impact alarm instruction through the ecological stability coefficient of the object recognition platform, and sounds an impact alarm after receiving the impact alarm instruction through the image alarm module. The method for identifying objects in coastal wetlands using high-definition remote sensing images uses the image information acquisition model to collect and analyze data from high-definition remote sensing images to obtain ecological stability information. The ecological stability information includes vegetation information, land information and animal information. The vegetation information is used to measure the enrichment degree of vegetation in coastal wetlands, and the larger the vegetation information, the higher the enrichment degree. The land information is used to measure the degree of change of land in coastal wetlands, and the larger the land information, the higher the degree of change. The animal information is used to measure the degree of change of land in coastal wetlands. The enrichment degree of animals, and the greater the animal information, the higher the enrichment degree. Therefore, the ecological stability coefficient obtained according to the three can comprehensively measure the ecological environment stability of coastal wetlands, and the larger the ecological stability coefficient, the higher the ecological environment stability, indicating that the better the ecological environment protection, the smaller the threat faced. Finally, the ecological stability coefficient is determined, so as to realize the discrimination of the stability of the ecological environment of coastal wetlands, and alarm the threatened coastal wetlands, and sound alarms of different sound intensities according to the value of the ecological stability coefficient; the coastal wetland high-definition remote sensing image object recognition method of the present invention uses a convolutional neural network to perform object recognition on the high-definition remote sensing images of coastal wetlands, which can fully extract the feature information in the image, improve the recognition accuracy, have strong generalization ability, and can adapt to different types of coastal wetland remote sensing image object recognition tasks. At the same time, by performing data analysis on the feature information in the image, the ecological stability of the coastal wetlands can be monitored in real time, and when the coastal wetlands are threatened, they can be discovered and alarmed in time, thereby avoiding greater damage, and making outstanding contributions to the ecological environment protection of coastal wetlands. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention will be further described below in conjunction with the accompanying drawings.

[0037] Figure 1 It is a principle block diagram of a method for identifying objects in coastal wetland high-definition remote sensing images based on a convolutional network in the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] Embodiment 1:

[0040] See also Figure 1 As shown, this embodiment is a method for identifying objects in coastal wetland high-definition remote sensing images based on a convolutional network, comprising the following modules: a model building module, an image acquisition module, an information acquisition module, an information analysis module, an object recognition platform, and an image alarm module;

[0041] The model building module is used to build an image information acquisition model; it is also used to input high-definition remote sensing images into the image information acquisition model, generate information acquisition instructions, and send the information acquisition instructions to the information acquisition module;

[0042] The image acquisition module is used to obtain high-definition remote sensing images and send the high-definition remote sensing images to the model building module;

[0043] The information acquisition module is used to obtain ecological stability information using the image information acquisition model after receiving the information acquisition instruction, and send the ecological stability information to the information analysis module; wherein the ecological stability information includes vegetation information ZB, land information TD and animal information DW;

[0044] Wherein, the information analysis module is used to obtain the ecological stability coefficient ST according to the ecological stability information, and send the ecological stability coefficient ST to the ground feature recognition platform;

[0045] Wherein, the ground object recognition platform is used to generate an impact alarm instruction for the ecological stability coefficient ST, and send the impact alarm instruction to the image alarm module;

[0046] Wherein, the image alarm module is used to sound an impact alarm after receiving an impact alarm instruction.

[0047] Embodiment 2:

[0048] See also Figure 1 As shown, this embodiment is a method for identifying objects in coastal wetland high-definition remote sensing images based on a convolutional network, comprising the following steps:

[0049] Step 1: The model building module builds the image information acquisition model;

[0050] Step 2: The image acquisition module obtains high-definition remote sensing images and sends the high-definition remote sensing images to the model building module;

[0051] Step 3: The model building module inputs the high-definition remote sensing image into the image information acquisition model, generates information acquisition instructions, and sends the information acquisition instructions to the information acquisition module;

[0052] Step 4: After receiving the information collection instruction, the information collection module uses the image information collection model to obtain ecological stability information, and sends the ecological stability information to the information analysis module;

[0053] Step 5: The information analysis module obtains the ecological stability coefficient ST according to the ecological stability information, and sends the ecological stability coefficient ST to the ground feature recognition platform;

[0054] Step 6: The ecological stability coefficient ST of the ground feature recognition platform generates an impact alarm instruction and sends the impact alarm instruction to the image alarm module;

[0055] Step 7: After receiving the impact alarm command, the image alarm module sounds the impact alarm.

[0056] Embodiment 3:

[0057] Based on any of the above embodiments, embodiment 3 of the present invention is an information collection module, and the function of the information collection module is to obtain ecological stability information, which includes vegetation information ZB, land information TD and animal information DW; vegetation information can measure the enrichment degree of vegetation in coastal wetlands, and the larger the vegetation information, the higher the enrichment degree; land information can measure the degree of change of land in coastal wetlands, and the larger the land information, the higher the degree of change; animal information can measure the enrichment degree of animals in coastal wetlands, and the larger the animal information, the higher the enrichment degree;

[0058] The vegetation information ZB is: obtain the vegetation area and average vegetation height in the high-definition remote sensing image, and mark them as the planting surface value ZM and the planting height value ZG respectively, quantify the planting surface value ZM and the planting height value ZG, extract the values ​​of the planting surface value ZM and the planting height value ZG, and substitute them into the formula for calculation. Obtain vegetation information ZB, where b1 and b2 are preset proportional coefficients corresponding to the set planting surface value ZM and planting height value ZG, respectively, b1 and b2 satisfy b1+b2=1, 0<b2<b1<1, and b1=0.56 and b2=0.44;

[0059] The land information TD is as follows: obtain the land area collected for the first time in the high-definition remote sensing image and the land area at the current moment, obtain the difference between the two, and mark it as the ground change value DB; obtain the average volume of all peaks collected for the first time in the high-definition remote sensing image and the average volume of all peaks at the current moment, obtain the difference between the two, and mark it as the peak change value FB; sum up all the peak change values ​​FB to obtain the peak change total value FZ; quantify the ground change value DB and the peak change total value FZ, extract the values ​​of the ground change value DB and the peak change total value FZ, and substitute them into the formula for calculation; according to the formula Obtain land information TD, where t1 and t2 are respectively the preset proportional coefficients corresponding to the set ground change value DB and the peak change total value FZ, t1 and t2 satisfy t1+t2=1, 0<t1<t2<1, and t1=0.33, t2=0.67;

[0060] Among them, the animal information DW is: randomly obtain an area in the high-definition remote sensing image, and mark it as the monitoring area, mark the animals in the monitoring area that reach the preset volume as monitored animals, obtain the total number of initial monitored animals in the preset time period of the monitoring area, and mark it as the initial value CS, when the monitored animal enters the monitoring area, the number of monitoring times of the monitored animal is increased by one, obtain the total number of monitoring times of the monitored animals entering the monitoring area within the preset time period, and mark it as the measurement value CC, quantify the initial value CS and the measurement value CC, extract the values ​​of the initial value CS and the measurement value CC, and substitute them into the formula for calculation, according to the formula The animal information DW is obtained, where d1 and d2 are the preset proportional coefficients corresponding to the set initial value CS and the measured value CC respectively, and d1 and d2 satisfy d1+d2=1, 0<d1<d2<1, and d1=0.19 and d2=0.81.

[0061] Embodiment 4:

[0062] Based on any of the above embodiments, embodiment 4 of the present invention is an information analysis module. The function of the information analysis module is to obtain an ecological stability coefficient ST. The ecological stability coefficient ST can comprehensively measure the ecological environment stability of coastal wetlands. The larger the ecological stability coefficient, the higher the ecological environment stability, indicating that the better the ecological environment protection, the smaller the threat faced;

[0063] The specific process of obtaining the ecological stability coefficient ST is as follows:

[0064] The vegetation information ZB, land information TD and animal information DW are quantified, the values ​​of vegetation information ZB, land information TD and animal information DW are extracted, and substituted into the formula for calculation. The ecological stability coefficient ST is obtained, where π is a mathematical constant, s1, s2 and s3 are the preset weight factors corresponding to the set vegetation information ZB, land information TD and animal information DW, s1, s2 and s3 satisfy s2>s1>s3>1.451, and s1=1.95, s2=2.34, s3=1.58;

[0065] The ecological stability coefficient ST is sent to the land feature identification platform.

[0066] Embodiment 5:

[0067] Based on any of the above embodiments, Embodiment 5 of the present invention is a ground feature recognition platform. The function of the ground feature recognition platform is to determine the ecological stability coefficient ST. The specific process is as follows:

[0068] Compare the ecological stability coefficient ST with the preset ecological stability threshold STy:

[0069] If the ecological stability coefficient ST is less than the ecological stability threshold STy, an impact alarm instruction is generated and sent to the image alarm module.

[0070] Embodiment 6:

[0071] Based on any of the above embodiments, Embodiment 6 of the present invention is an image alarm module. The function of the image alarm module is to promptly sound an impact alarm when an abnormality occurs in the coastal wetland ecological environment. The details are as follows:

[0072] Get the value of the ecological stability coefficient ST and mark it as the volume adjustment coefficient YT. Substitute the volume adjustment coefficient YT into the formula for calculation. A sound intensity value YQ is obtained, wherein δ is a preset adjustment factor, δ>0, and the intensity of the sound affecting the alarm is adjusted according to the sound intensity value YQ.

[0073] Based on the above embodiments 1-6, the working principle of the present invention is as follows:

[0074] The coastal wetland high-definition remote sensing image feature recognition method uses an image information acquisition model to collect and analyze high-definition remote sensing images to obtain ecological stability information. The ecological stability information includes vegetation information, land information and animal information. Vegetation information is used to measure the enrichment of vegetation in coastal wetlands, and the larger the vegetation information, the higher the enrichment. Land information is used to measure the degree of change of land in coastal wetlands, and the larger the land information, the higher the degree of change. Animal information is used to measure the enrichment of animals in coastal wetlands, and the larger the animal information, the higher the enrichment. Therefore, the ecological stability coefficient obtained based on the three can comprehensively measure the ecological environment stability of coastal wetlands, and the larger the ecological stability coefficient, the higher the ecological environment stability, which means that the better the ecological environment protection is, the smaller the threat faced, and finally the ecological environment is protected. The stability coefficient is determined, thereby realizing the discrimination of the stability of the ecological environment of the coastal wetland, and sounding an alarm for the threatened coastal wetland, and sounding an alarm with different sound intensities according to the value of the ecological stability coefficient; the coastal wetland high-definition remote sensing image object recognition method of the present invention recognizes objects in high-definition remote sensing images of coastal wetlands through a convolutional neural network, can fully extract feature information in the image, improve recognition accuracy, has strong generalization ability, can adapt to different types of coastal wetland remote sensing image object recognition tasks, and at the same time, through data analysis of feature information in the image, can monitor the ecological stability of the coastal wetland in real time, and can timely discover and alarm when the coastal wetland is threatened, thereby avoiding causing greater damage, and making an outstanding contribution to the ecological environment protection of the coastal wetland.

[0075] It should be further explained that the above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The coefficients in the formula are set by technical personnel in this field according to actual conditions.

[0076] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0077] The above contents are merely examples and explanations of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the invention or exceed the scope defined in this application, they shall all fall within the protection scope of the present invention.

Claims

1. A method for identifying objects in coastal wetland high-definition remote sensing images based on convolutional networks, characterized in that: The following steps are involved: Step 1: The model building module builds the image information acquisition model; Step 2: The image acquisition module obtains high-definition remote sensing images and sends the high-definition remote sensing images to the model building module; Step 3: The model building module inputs the high-definition remote sensing image into the image information acquisition model, generates information acquisition instructions, and sends the information acquisition instructions to the information acquisition module; Step 4: After receiving the information collection instruction, the information collection module uses the image information collection model to obtain ecological stability information, and sends the ecological stability information to the information analysis module; the ecological stability information includes vegetation information ZB, land information TD and animal information DW; The specific process of the information acquisition module acquiring vegetation information ZB is as follows: After receiving the information collection instruction, the image information collection model is used to obtain the vegetation area and average vegetation height in the high-definition remote sensing image, and they are marked as the planting surface value ZM and the planting height value ZG respectively. The planting surface value ZM and the planting height value ZG are quantified according to the formula Obtain vegetation information ZB, wherein b1 and b2 are preset proportional coefficients corresponding to the set planting surface value ZM and planting height value ZG respectively; The specific process of the information acquisition module acquiring land information TD is as follows: After receiving the information collection instruction, the image information collection model is used to obtain the land area collected for the first time in the high-definition remote sensing image and the land area at the current moment, and the difference between the two is obtained and marked as the ground change value DB. The average volume of all peaks collected for the first time in the high-definition remote sensing image and the average volume of all peaks at the current moment are obtained, and the difference between the two is obtained and marked as the peak change value FB. All peak change values ​​FB are summed up to obtain the peak change total value FZ. The ground change value DB and the peak change total value FZ are quantified according to the formula The land information TD is obtained, wherein t1 and t2 are respectively the preset proportional coefficients corresponding to the set ground change value DB and the peak change total value FZ; The specific process of the information acquisition module acquiring the animal information DW is as follows: After receiving the information collection instruction, the image information collection model is used to randomly obtain an area in the high-definition remote sensing image and mark it as the monitoring area. The animals in the monitoring area that reach the preset volume are marked as monitored animals. The total number of initial monitored animals in the preset time period of the monitoring area is obtained and marked as the initial value CS. When the monitored animal enters the monitoring area, the number of monitoring times of the monitored animal is increased by one. The total number of monitoring times of the monitored animals entering the monitoring area within the preset time period is obtained and marked as the measurement value CC. The initial value CS and the measurement value CC are quantified according to the formula The animal information DW is obtained, where d1 and d2 are the preset proportional coefficients corresponding to the set initial value CS and the measured value CC respectively; Step 5: The information analysis module obtains the ecological stability coefficient ST according to the ecological stability information, and sends the ecological stability coefficient ST to the ground feature recognition platform; The specific process of the information analysis module obtaining the ecological stability coefficient ST is as follows: The vegetation information ZB, land information TD and animal information DW are quantified according to the formula The ecological stability coefficient ST is obtained, where π is a mathematical constant, s1, s2 and s3 are respectively the preset weight factors corresponding to the set vegetation information ZB, land information TD and animal information DW; Step 6: The ecological stability coefficient ST of the ground feature recognition platform generates an impact alarm instruction and sends the impact alarm instruction to the image alarm module; Step 7: After receiving the impact alarm command, the image alarm module sounds the impact alarm.

2. According to claim 1, a method for identifying objects in coastal wetland high-definition remote sensing images based on convolutional networks is characterized in that: The specific process of the model building module building the image information acquisition model is as follows: A convolutional network model is established using convolutional network technology and is labeled as an image information acquisition model.

3. The method for identifying objects in coastal wetland high-definition remote sensing images based on a convolutional network according to claim 1 is characterized in that: The specific process of the image acquisition module obtaining high-definition remote sensing images is as follows: Remote sensing technology is used to collect images of coastal wetlands, obtain high-definition remote sensing images, and send the high-definition remote sensing images to the model building module.

4. The method for identifying objects in coastal wetland high-definition remote sensing images based on a convolutional network according to claim 1 is characterized in that: The specific process of the object recognition platform generating an alarm-influencing instruction is as follows: Compare the ecological stability coefficient ST with the preset ecological stability threshold STy: If the ecological stability coefficient ST is less than the ecological stability threshold STy, an impact alarm instruction is generated and sent to the image alarm module.

5. The method for identifying objects in coastal wetland high-definition remote sensing images based on convolutional networks according to claim 1 is characterized in that: The specific process of the image alarm module sounding an alarm is as follows: Get the value of the ecological stability coefficient ST and mark it as the volume adjustment coefficient YT. Substitute the volume adjustment coefficient YT into the formula for calculation. A sound intensity value YQ is obtained, wherein δ is a preset adjustment factor, and the intensity of the sound affecting the alarm is adjusted according to the sound intensity value YQ.

Images