Protection area bird monitoring protection system based on machine learning

Abstract

The invention relates to the technical field of bird monitoring and protection, in particular to a protection area bird monitoring and protection system based on machine learning, and the system obtains a monitoring image through an acquisition module, and obtains a first bird image and corresponding first bird data through a default monitoring module based on a preset machine learning algorithm. A monitoring image is recognized through a rapid monitoring module based on a mode of pixel comparison with a first bird image, monitoring data are stored through a cache module, the cache monitoring module is used for judging whether the monitoring data in the cache module meet a preset triggering condition or not, if yes, the rapid monitoring module is called for monitoring, and if not, the rapid monitoring module is called for monitoring. If yes, calling a default monitoring module for monitoring. According to the invention, the preliminary analysis of the monitoring data is realized in a temporary storage mode of the cache module, so that the rapid monitoring module is called to carry out rapid monitoring in a pixel comparison mode according to the number condition of the currently monitored birds, and the problem that a large number of birds are difficult to identify at the same time in a short time in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of bird monitoring and protection, and particularly to a bird monitoring and protection system in a protected area based on machine learning. Background Art

[0002] As a key species in the ecosystem, the changes in the quantity and species of birds can reflect the environmental quality and the health status of the ecosystem, and are important indicators for biodiversity monitoring. By monitoring birds, not only can endangered birds and their habitats be effectively protected, but also the spread of infectious diseases can be warned and prevented in advance, and the impacts of birds on power facilities and aviation safety can be reduced, etc.

[0003] The bird monitoring method based on machine learning is the current mainstream research direction, especially the method based on image recognition. Due to its efficient real-time detection ability, this method has been widely used in bird monitoring. For example, the existing bird monitoring technologies based on image recognition include the bird recognition and classification technology using convolutional neural network (CNN) in deep learning, the bird recognition technology based on YOLO series algorithms, etc.

[0004] However, when the existing bird monitoring methods based on machine learning image recognition face the scenario where a large number of birds appear in a short-time picture, the effect is often poor. For example, in scenarios such as birds wandering back and forth or flocks of birds gathering in wetlands, due to the large number of bird individuals, large pose changes, and fast movement speeds, the existing algorithms often have difficulty accurately identifying and distinguishing each individual in the picture, and problems such as target loss or incorrect association are likely to occur, thus affecting the accuracy of the monitoring results. Therefore, people need a monitoring method that can identify a large number of birds simultaneously in a short time. Summary of the Invention

[0005] Therefore, the present invention provides a bird monitoring and protection system in a protected area based on machine learning to solve the problem in the prior art that it is difficult to identify a large number of birds simultaneously in a short time.

[0006] The present invention provides a bird monitoring and protection system in a protected area based on machine learning, including a cache module, a collection module, a default monitoring module, a fast monitoring module, and a cache monitoring module, wherein:

[0007] The collection module is used to obtain monitoring images;

[0008] The default monitoring module is used to identify the monitoring images based on a preset machine learning algorithm, obtain the first bird image and the corresponding first bird data, and save the first bird image and the first bird data as monitoring data in the cache module;

[0009] A fast monitoring module, which is used to identify a monitoring image by comparing pixels with a first bird image, obtain a second bird image and corresponding second bird data, and save the second bird image and the second bird data as monitoring data in a cache module;

[0010] A cache monitoring module, which is used to determine whether the monitoring data in the cache module meets a preset trigger condition. If so, it calls the fast monitoring module for monitoring; if not, it calls the default monitoring module for monitoring. The preset trigger condition is used to judge the quantity level of the birds to be monitored currently according to the monitoring data.

[0011] In a preferred manner, identifying the monitoring image by comparing pixels with the first bird image to obtain a second bird image and corresponding second bird data includes:

[0012] Normalize multiple first bird images to obtain multiple sample images with the same size;

[0013] According to the size distribution of multiple first bird images, adjust the size of the monitoring image multiple times to obtain target images with multiple size specifications;

[0014] Set a detection window with the same size as the sample image in the target image, move the detection window, and compare the content in the detection window with the sample image pixel by pixel to obtain a second bird image and corresponding second bird data.

[0015] In a preferred manner, setting a detection window with the same size as the sample image in the target image, moving the detection window, and comparing the content in the detection window with the sample image pixel by pixel to obtain a second bird image and corresponding second bird data includes:

[0016] Obtain a preset mapping model, where the preset mapping model is used to map different images to unique feature values;

[0017] Input the sample image into the preset mapping model to obtain a sample label value corresponding to each sample image;

[0018] Set a detection window with the same size as the sample image in the target image, move the detection window, and input the content in the detection window into the preset mapping model each time to obtain a set of output feature values;

[0019] Match the sample label value with the feature values in the set of output feature values, and obtain the position information of the detection window corresponding to the successfully matched feature value;

[0020] Intercept the local image content in the monitoring image according to the position information as the initial bird image corresponding to the sample label value;

[0021] Based on the correspondence relationships among the initial bird images, sample label values, sample images, first bird images, and first bird data, all the initial bird images are aggregated to obtain a second bird image and the corresponding second bird data.

[0022] In a preferred manner, normalizing multiple first bird images to obtain multiple sample images with the same size, including:

[0023] Performing feature-preserving size adjustment on the first bird images to obtain standardized images;

[0024] Performing pixel value transformation on the standardized images to obtain sample images.

[0025] In a preferred manner, performing feature-preserving size adjustment on the first bird images to obtain standardized images, including:

[0026] Based on the target adjustment size, obtaining a first pixel and the neighboring pixels of the first pixel;

[0027] Performing weighted average calculation on the pixel values of the first pixel and the neighboring pixels to obtain standardized pixels, where multiple standardized pixels form a standardized image.

[0028] In a preferred manner, performing pixel value transformation on the standardized images to obtain sample images, including:

[0029] Obtaining a preset mapping pixel interval;

[0030] Transforming the pixel values within the preset mapping pixel interval in the standardized images to 1, and the remaining pixel values to 0, to obtain sample images.

[0031] In a preferred manner, the preset mapping model includes:

[0032] Calculating the feature value through the following formula:

[0033] V = ∑c × p;

[0034] where V represents the feature value corresponding to the input image of a preset mapping model, p represents a pixel value in the input image, c represents the preset coefficient corresponding to each pixel value, and the preset coefficients are respectively different powers of two.

[0035] In a preferred manner, based on the correspondence relationships among the initial bird images, sample label values, sample images, first bird images, and first bird data, aggregating all the initial bird images to obtain a second bird image and the corresponding second bird data, including:

[0036] Judging the coverage relationship of multiple initial bird images;

[0037] If the number of other initial bird images covered by an initial bird image exceeds a preset threshold, the covered initial bird image will be used as the second bird image;

[0038] If the number of other initial bird images covered by an initial bird image does not exceed a preset threshold, the covering initial bird image will be used as the second bird image;

[0039] Based on the correspondence between the second bird image, the initial bird image, the sample marking value, the sample image, the first bird image, and the first bird data, the first bird data corresponding to the second bird image will be used as its corresponding second bird data.

[0040] In a preferred manner, each piece of monitoring data is temporarily stored in the cache module and then deleted regularly; determining whether the monitoring data in the cache module meets a preset trigger condition includes:

[0041] Determining whether the cache module has reached its capacity limit;

[0042] If so, call the fast monitoring module for monitoring;

[0043] If not, call the default monitoring module for monitoring.

[0044] In a preferred manner, determining whether the monitoring data in the cache module meets a preset trigger condition includes:

[0045] Determining whether the number of monitoring data of the same type of bird saved in the cache module within a preset time reaches a preset quantity threshold;

[0046] If so, call the fast monitoring module for monitoring;

[0047] If not, call the default monitoring module for monitoring.

[0048] The beneficial effects of adopting the above embodiments are:

[0049] The present invention provides a machine learning-based bird monitoring and protection system for a protected area, including a cache module, a collection module, a default monitoring module, a quick monitoring module, and a cache monitoring module. The collection module is used to obtain monitoring images. The default monitoring module is used to identify the monitoring images based on a preset machine learning algorithm to obtain a first bird image and corresponding first bird data, and save the first bird image and the first bird data as monitoring data in the cache module. The quick monitoring module is used to identify the monitoring images by comparing pixels with the first bird image to obtain a second bird image and corresponding second bird data, and save the second bird image and the second bird data as monitoring data in the cache module. The cache monitoring module is used to determine whether the monitoring data in the cache module meets a preset trigger condition. If so, it calls the quick monitoring module for monitoring. If not, it calls the default monitoring module for monitoring. The preset trigger condition is used to judge the quantity level of the birds to be monitored according to the monitoring data. Through the way of temporarily storing in the cache module, the present invention cleverly uses a small improvement to perform a preliminary analysis of the monitoring data through the preset trigger condition, so as to judge the quantity of the currently monitored birds. Usually, the default monitoring module is used to perform accurate monitoring through machine learning, and when the number of birds is large, the quick monitoring module is called to perform quick monitoring by pixel comparison, solving the problem in the prior art that it is difficult to identify a large number of birds simultaneously in a short time. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 FIG. is a system architecture diagram of an embodiment of the machine learning-based bird monitoring and protection system for a protected area provided by the present invention;

[0051] Figure 2 is Figure 1 a method step diagram executed by the quick monitoring module in;

[0052] Figure 3 FIG. is a specific step diagram of step S203 in. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0054] Combined with Figure 1As shown in the figure, a specific embodiment of the present invention discloses a bird monitoring and protection system based on machine learning, including a cache module 110, a collection module 120, a default monitoring module 130, a fast monitoring module 140, and a cache monitoring module 150, where:

[0055] The collection module 120 is used to obtain monitoring images;

[0056] The default monitoring module 130 is used to identify the monitoring images based on a preset machine learning algorithm to obtain the first bird image and the corresponding first bird data, and save the first bird image and the first bird data as monitoring data in the cache module;

[0057] The fast monitoring module 140 is used to identify the monitoring images by comparing pixels with the first bird image to obtain the second bird image and the corresponding second bird data, and save the second bird image and the second bird data as monitoring data in the cache module;

[0058] The cache monitoring module 150 is used to determine whether the monitoring data in the cache module meets a preset trigger condition. If so, it calls the fast monitoring module for monitoring; if not, it calls the default monitoring module for monitoring. The preset trigger condition is used to judge the quantity level of the birds to be monitored according to the monitoring data.

[0059] In the above process, the monitoring image refers to the unprocessed image collected, and the monitoring data represents the data required for bird monitoring. Specifically, what kind of data can be flexibly set according to the actual specific situation, generally including the data of the image and the birds of the corresponding species of the image. In this embodiment, it specifically refers to the first bird image, the second bird image, the first bird data, and the second bird data. Here, the first and second are only the distinguishing prefixes of different sources of the same kind of data. The first bird image and the second bird data are the monitoring data obtained by the default monitoring module, while the second bird image and the second bird data are the monitoring data obtained by the fast detection module.

[0060] Normally, this system uses the default monitoring module to monitor birds using any existing image recognition algorithm. The fast monitoring module uses the pixel comparison method to detect birds. Compared with the prior art, the pixel comparison method does not require a large amount of calculations. Therefore, theoretically, its recognition speed is higher than that of the default monitoring module. However, it can be imagined that if you want to achieve a fast effect, then there must be enough and accurate data for pixel comparison.

[0061] In this embodiment, accurate image data for pixel comparison can be obtained through the default monitoring module, and these monitoring data are temporarily stored in the cache module, which can also be used to monitor the number of birds in the current monitoring environment, thereby ensuring the accuracy of the operation of the fast monitoring module. On the one hand, when there are a large number of birds in the environment, it is exactly the situation where the existing deep learning-based image recognition algorithm fails. On the other hand, the situation of a large number of birds means that a large amount of monitoring data is obtained. At this time, the monitoring data in the cache module exactly meets the conditions for accurately operating the fast detection module, that is, the data is accurate enough and the data volume is sufficient. In this way, the present invention can flexibly adopt a suitable method for bird monitoring and recognition according to the actual situation, thereby overcoming the defects of the prior art, taking into account both recognition accuracy and recognition speed, effectively improving the efficiency and accuracy of bird monitoring. Especially in the scenario of a large number of birds, it can quickly and accurately complete the monitoring task, providing strong support for the protection of birds in the reserve.

[0062] It can be seen that the preset trigger condition represents the condition for determining the number of birds and can also be regarded as the condition for judging whether the current monitoring data meets the accurate pixel comparison. Similarly, it can be understood that the specific definition of the preset trigger condition can be flexibly set according to the actual situation.

[0063] For example, in a preferred embodiment, the above preset trigger condition specifically includes:

[0064] Judge whether the number of monitoring data of the same species of birds saved in the cache module reaches a preset quantity threshold within a preset time;

[0065] If so, call the fast monitoring module for monitoring;

[0066] If not, call the default monitoring module for monitoring.

[0067] The advantage of this method is that it can accurately judge the activity level and population scale of this species of birds in the current monitoring area according to the accumulated quantity of monitoring data of the same species of birds within a preset time, thereby realizing the intelligent switching of the monitoring mode. In addition, this setting of the trigger condition based on the quantity of monitoring data of the same species of birds can also flexibly adjust the preset quantity threshold according to the actual monitoring requirements, making it applicable to diverse monitoring scenarios of different nature reserves, different bird species, and different monitoring periods, further enhancing the adaptability and practicality of the system.

[0068] For another example, in a preferred embodiment, each piece of monitoring data is temporarily stored in the cache module and then deleted regularly; the above preset trigger condition can specifically be:

[0069] Judge whether the cache module reaches the capacity limit;

[0070] If so, call the fast monitoring module for monitoring;

[0071] If not, then call the default monitoring module for monitoring.

[0072] The advantage of this method is that it can intelligently select a suitable monitoring module for bird monitoring according to the capacity status of the cache module, thus achieving an effective balance between monitoring efficiency and accuracy. Specifically, when the monitoring data in the cache module does not reach the capacity limit, it indicates that the number of birds in the current monitoring environment is relatively small. At this time, calling the default monitoring module for monitoring can give full play to its high-accuracy advantage based on machine learning algorithms, ensuring accurate identification and data recording of each bird, and providing detailed and reliable data support for the protection of birds in the reserve. When the cache module reaches the capacity limit, it means that there are a large number of birds in the current monitoring area and a large amount of monitoring data. At this time, calling the fast monitoring module and using its fast recognition method of pixel comparison can complete a large number of bird monitoring tasks in a short time, effectively avoiding problems such as monitoring delay or system crash caused by excessive data volume, and ensuring the continuity and timeliness of the monitoring work. More importantly, compared with the method of directly monitoring the quantity, this trigger condition setting based on the capacity of the cache module also has the advantages of simplicity and easy implementation. Without complex algorithms and a large amount of computing resources, it can achieve intelligent switching of the monitoring mode, reduce the operation cost and maintenance difficulty of the system, and improve the stability and reliability of the system.

[0073] Furthermore, the steps executed by the above-mentioned default monitoring module 120: identifying the monitoring image based on a preset machine learning algorithm to obtain a first bird image and corresponding first bird data, where the preset machine learning algorithm can be classic and widely used machine learning algorithms such as convolutional neural network (CNN), support vector machine (SVM), and random forest (Random Forest). It can be understood that the above algorithms are all existing technologies that can be understood by those skilled in the art. Specifically, which preset machine learning algorithm to adopt and how to implement bird monitoring are all conceivable by those skilled in the art, so this article will not elaborate too much.

[0074] Furthermore, in combination with Figure 2 As shown, in a preferred embodiment, on the premise that the monitoring data in the cache module meets the preset trigger condition, the steps executed by the above-mentioned fast monitoring module 130: identifying the monitoring image based on pixel comparison with the first bird image to obtain a second bird image and corresponding second bird data, specifically including:

[0075] S201. Normalize multiple first bird images to obtain multiple sample images of the same size;

[0076] S202. Adjust the size of the monitoring image multiple times according to the size distribution of multiple first bird images to obtain target images of multiple size specifications;

[0077] S203. Set a detection window with the same size as the sample image in the target image, move the detection window, and perform pixel comparison between the content in the detection window and the sample image to obtain a second bird image and corresponding second bird data.

[0078] In the above process, since the first bird images are cropped from the monitoring image and only show partial images of the birds, their sizes may not be uniform, so normalization processing is required. In this embodiment, normalization means adjusting first bird images of different sizes to a unified size to ensure that the sizes of the sample images are the same, facilitating subsequent pixel comparison operations. For example, all first bird images can be adjusted to a sample image of 64x64 pixels. In this way, when performing pixel comparison, it can be ensured that the number and arrangement of pixel points of each sample image are the same, thereby improving the accuracy and efficiency of the comparison.

[0079] The target image refers to the monitoring image to be analyzed currently. In the actual monitoring process, due to the influence of factors such as shooting angle and distance, the sizes of birds in the same monitoring image may vary. Therefore, to ensure accurate identification of birds of different sizes, the monitoring image needs to be adjusted multiple times to generate target images of multiple size specifications. For example, if the sizes of the first bird images are mainly concentrated in a relatively small range, such as between 32x32 pixels and 64x64 pixels, then the monitoring image can be adjusted to multiple sizes close to this range, such as 32x32, 48x48, 64x64, etc., to cover bird images of different sizes and ensure the integrity of identification.

[0080] The detection window refers to a region set in the target image with the same size as the sample image. When performing pixel comparison, by moving the detection window in the target image, the target image can be scanned region by region. In each detection window, the image content within the window is compared with the sample image in terms of pixels. By comparing the similarity of the pixels, it is determined whether the region contains a bird image similar to the sample image. If the similarity reaches a preset threshold, it is considered that the region contains a bird image, thereby obtaining a second bird image and corresponding second bird data. For example, a pixel difference algorithm can be used to calculate the pixel difference value between the image in the detection window and the sample image. If the difference value is lower than a certain threshold, it is considered that the two are similar. This way of pixel comparison does not require complex feature extraction and machine learning operations, so the recognition speed is relatively fast and is suitable for rapid recognition and counting in scenarios with a large number of birds.

[0081] In addition, the detection window of this embodiment is similar to the convolution technology in the prior art, but the difference is that convolution is only responsible for feature extraction and cannot directly complete the recognition work. It requires subsequent further processing of the extracted features (such as the fully connected layer in the neural network, etc.), while the detection window of this embodiment can directly obtain the result without subsequent data analysis. This means that by moving the window and the preset mapping model, birds can be directly identified without the need for complex feature extraction and classification operations like convolutional neural networks, making the entire system more efficient.

[0082] Specifically, in a preferred embodiment, the above step S201, normalizing the plurality of first bird images to obtain a plurality of sample images of the same size, specifically includes:

[0083] Performing feature-preserving size adjustment on the first bird image to obtain a standardized image;

[0084] The pixel values ​​of the standardized image are transformed to obtain a sample image.

[0085] In the above process, feature-preserving resizing refers to the use of a method that can retain the key features of the image when adjusting the image size, ensuring that the adjusted image retains the significant features of the bird, such as contours and textures, while maintaining a uniform size, to facilitate subsequent identification and comparison. Pixel value transformation refers to adjusting the pixel values ​​of the standardized image to meet the requirements of subsequent processing, such as normalizing the pixel values ​​to a specific interval, or performing grayscale processing to convert color images into grayscale images, reducing the amount of data and improving processing efficiency. Through feature-preserving resizing and pixel value transformation, sample images with uniform size, feature preservation, and pixel value adaptation can be obtained, which facilitates subsequent pixel comparison operations and improves the accuracy and efficiency of comparison.

[0086] Specifically, in a preferred embodiment, the above step of performing feature-preserving size adjustment on the first bird image to obtain a standardized image specifically includes:

[0087] Adjust the size based on the target to obtain a first pixel and neighboring pixels of the first pixel;

[0088] A weighted average calculation is performed on the pixel values ​​of the first pixel and the neighboring pixels to obtain a standardized pixel, wherein a plurality of standardized pixels constitute a standardized image.

[0089] In the above process, the target adjusted size is the size of the sample image to which it is currently to be adjusted. The first pixel is the pixel to be adjusted currently. The specific selection range of the neighboring pixels can be flexibly set based on the target adjusted size according to specific circumstances. For example, when it is necessary to adjust to a smaller size, a larger range of neighboring pixels can be selected for weighted averaging to better retain features. The advantage of this adjustment method is that even after the size of the sample image is reduced, it still has a high recognition rate, which is convenient for subsequent precise comparison with the content in the detection window.

[0090] Further, the above step: performing pixel value transformation on the standardized image to obtain a sample image, includes:

[0091] Obtaining a preset mapping pixel interval;

[0092] Transforming the pixel values in the standardized image that are within the preset mapping pixel interval into 1, and the remaining pixel values into 0, to obtain a sample image.

[0093] In the above process, the preset pixel interval refers to a pixel value range preset according to the characteristics of the birds and the monitoring requirements. For example, if the birds in a certain area's habitat are black, then the color range of black or close to black can be set as the preset mapping pixel interval. The advantage of doing this is that the bird part in the image can be effectively distinguished from the background or other non-bird parts, improving the contrast and feature salience of the sample image, thereby enhancing the accuracy and efficiency of subsequent pixel comparison. By transforming the pixel values within the preset mapping pixel interval into 1 and the remaining pixel values into 0, a binary sample image can be generated, making the contour and main features of the birds clearer and facilitating quick identification and comparison.

[0094] Further, as shown in Figure 3 In a preferred embodiment, the above step S203: setting a detection window with the same size as the sample image in the target image, moving the detection window and performing pixel comparison between the content in the detection window and the sample image to obtain a second bird image and corresponding second bird data, specifically includes:

[0095] S301: Obtaining a preset mapping model, where the preset mapping model is used to map different images to unique feature values;

[0096] S302: Inputting the sample image into the preset mapping model to obtain a sample marker value corresponding to each sample image;

[0097] S303: Setting a detection window with the same size as the sample image in the target image, moving the detection window and inputting the content in the detection window into the preset mapping model each time it is moved to obtain a set of output feature values;

[0098] S304. Match the sample marker value with the characteristic values in the set of output characteristic values, and obtain the position information of the detection window corresponding to the successfully matched characteristic value;

[0099] S305. Intercept the local image content in the monitoring image according to the position information as the initial bird image corresponding to the sample marker value;

[0100] S306. Based on the corresponding relationships among the initial bird image, the sample marker value, the sample image, the first bird image, and the first bird data, summarize all the initial bird images to obtain the second bird image and the corresponding second bird data.

[0101] In the above process, the only meaning of the preset mapping model output is to ensure that each sample image can be accurately identified and distinguished. Specifically, the preset mapping model maps different images to unique characteristic values, so that each sample image has a unique identifier, that is, the sample marker value. In this way, when performing characteristic value matching, it can be accurately judged whether the content in the detection window matches the sample image. The sample marker value is the characteristic value corresponding to the sample image.

[0102] It can be understood that if conditional judgment is directly used for mechanical pixel comparison, the pixel comparison process is slow, because this method will perform a large number of conditional judgments cyclically, with a large amount of calculation and a lot of computing resources consumed (mainly the time delay caused by the computer reading data). In this embodiment, the method of calculating characteristic values using the preset mapping model is used for comparison, which can convert complex pixel comparison into simple numerical comparison, greatly reducing conditional judgment, thereby improving the comparison speed. Especially when dealing with a large amount of data, this embodiment can complete the simultaneous comparison of a large number of data, without comparing one by one according to the moving order of the detection window, and its fast advantage is more obvious.

[0103] Specifically, in a preferred embodiment, the preset mapping model includes:

[0104] Calculate the characteristic value through the following formula:

[0105] V = ∑c × p;

[0106] Where, V represents the characteristic value corresponding to the input image of a preset mapping model, p represents a pixel value in the input image, c represents the preset coefficient corresponding to each pixel value, and the preset coefficients are different powers of two.

[0107] This model can cooperate with the pixel value transformation method that maps pixels to 0 and 1 in the previous text, so as to ensure the uniqueness of the output of the preset mapping model.

[0108] In addition, as can be seen from the foregoing, there is a one-to-one correspondence among the initial bird image, the sample marking value, the sample image, the first bird image, and the first bird data. Therefore, once the initial bird image is determined, the bird species in the image can be directly determined. It can be understood that when performing feature value matching, the feature values in the feature data set are obtained based on target images of different sizes. Then, when mapping the successfully matched feature values back to the monitoring image, the size ranges of the corresponding different detection windows are different, which will result in the situation that multiple initial bird images also have different sizes and may very likely have an overlapping and covering relationship. Therefore, it is necessary to perform summarization again to obtain the accurate second bird image and data. By summarizing all the initial bird images, duplicate counting can be avoided, thereby improving the accuracy and reliability of the monitoring data.

[0109] Specifically, in a preferred embodiment, the above step S306, based on the correspondence among the initial bird image, the sample marking value, the sample image, the first bird image, and the first bird data, summarizes all the initial bird images to obtain the second bird image and the corresponding second bird data, specifically including:

[0110] Judge the covering relationship of multiple initial bird images;

[0111] If the number of other initial bird images covered by an initial bird image exceeds the preset threshold, then use the covered initial bird image as the second bird image;

[0112] If the number of other initial bird images covered by an initial bird image does not exceed the preset threshold, then use the covering initial bird image as the second bird image;

[0113] Based on the correspondence among the second bird image, the initial bird image, the sample marking value, the sample image, the first bird image, and the first bird data, use the first bird data corresponding to the second bird image as its corresponding second bird data.

[0114] By judging the covering relationship of the initial bird images and determining the second bird image according to whether the covering quantity exceeds the preset threshold, the overlapping and covering problem among the initial bird images can be effectively solved. Specifically:

[0115] When the number of other initial bird images covered by an initial bird image exceeds a preset threshold, the covered initial bird image is taken as the second bird image, so that a more complete and clearer bird image can be retained, avoiding information loss or double counting caused by overlapping. When the number of other initial bird images covered by an initial bird image does not exceed the preset threshold, the covered initial bird image may only be a misidentified feature, then the covering initial bird image is taken as the second bird image, so as to ensure that the selected second bird image has high representativeness and accuracy, reducing the error caused by image overlapping.

[0116] The present invention provides a machine learning-based protected area bird monitoring and protection system, including a cache module, a collection module, a default monitoring module, a fast monitoring module and a cache monitoring module. The collection module is used to obtain monitoring images, and the default monitoring module is used to identify the monitoring images based on a preset machine learning algorithm to obtain the first bird image and the corresponding first bird data, and save the first bird image and the first bird data as monitoring data in the cache module. The fast monitoring module is used to identify the monitoring images by comparing pixels with the first bird image to obtain the second bird image and the corresponding second bird data, and save the second bird image and the second bird data as monitoring data in the cache module. The cache monitoring module is used to determine whether the monitoring data in the cache module meets a preset trigger condition. If so, the fast monitoring module is called for monitoring. If not, the default monitoring module is called for monitoring. The preset trigger condition is used to judge the quantity degree of the birds to be monitored currently according to the monitoring data. Through the way of temporarily storing in the cache module, the present invention cleverly uses a small improvement to perform a preliminary analysis of the monitoring data through the preset trigger condition, so as to judge the quantity situation of the currently monitored birds. Usually, the default monitoring module is used to perform accurate monitoring through machine learning, while when the number of birds is large, the fast monitoring module is called to perform fast monitoring by pixel comparison, solving the problem in the prior art that it is difficult to identify a large number of birds simultaneously in a short time.

[0117] It should be noted that the embodiments in this specification are all described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0118] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A bird monitoring and protection system in a protected area based on machine learning, characterized in that, It includes a cache module, a collection module, a default monitoring module, a quick monitoring module, and a cache monitoring module, where: The collection module is used to obtain monitoring images; The default monitoring module is used to identify the monitoring images based on a preset machine learning algorithm to obtain the first bird images and the corresponding first bird data, and save the first bird images and the first bird data as monitoring data in the cache module; The quick monitoring module is used to identify the monitoring images by comparing pixels with the first bird images to obtain the second bird images and the corresponding second bird data, and save the second bird images and the second bird data as monitoring data in the cache module; The cache monitoring module is used to determine whether the monitoring data in the cache module meets a preset trigger condition. If so, it calls the quick monitoring module for monitoring. If not, it calls the default monitoring module for monitoring. The preset trigger condition is used to judge the quantity level of the birds to be monitored currently according to the monitoring data.

2. The bird monitoring and protection system in the protected area based on machine learning according to claim 1, characterized in that Identifying the monitoring images by comparing pixels with the first bird images to obtain the second bird images and the corresponding second bird data includes: Normalizing multiple first bird images to obtain multiple sample images of the same size; Adjusting the size of the monitoring image multiple times according to the size distribution of the multiple first bird images to obtain target images of multiple size specifications; Setting a detection window with the same size as the sample image in the target image, moving the detection window, and comparing the content in the detection window with the sample image in terms of pixels to obtain the second bird images and the corresponding second bird data.

3. The bird monitoring and protection system in the protected area based on machine learning according to claim 2, characterized in that, Setting a detection window with the same size as the sample image in the target image, moving the detection window, and comparing the content in the detection window with the sample image in terms of pixels to obtain the second bird images and the corresponding second bird data includes: Obtaining a preset mapping model, where the preset mapping model is used to map different images to unique feature values; Inputting the sample images into the preset mapping model to obtain the sample marker values corresponding to each sample image; Setting a detection window with the same size as the sample image in the target image, moving the detection window, and inputting the content in the detection window into the preset mapping model each time it moves to obtain a set of output feature values; Matching the sample marker values with the feature values in the set of output feature values, and obtaining the position information of the detection window corresponding to the successfully matched feature values; Intercepting the local image content in the monitoring image according to the position information as the initial bird image corresponding to the sample marker value; Based on the corresponding relationships among the initial bird images, the sample marker values, the sample images, the first bird images, and the first bird data, summarizing all the initial bird images to obtain the second bird images and the corresponding second bird data.

4. The machine learning-based protected area bird monitoring and protection system according to claim 3, wherein Normalizing multiple first bird images to obtain multiple sample images of the same size includes: Adjusting the size of the first bird image while retaining features to obtain a standardized image; Performing pixel value transformation on the standardized image to obtain the sample images.

5. The machine learning-based protected area bird monitoring and protection system according to claim 4, characterized in that, Adjusting the size of the first bird image while retaining features to obtain a standardized image includes: Adjust the size based on the target to obtain the first pixel and the neighboring pixels of the first pixel; Perform weighted average calculation on the pixel values of the first pixel and the neighboring pixels to obtain normalized pixels, where multiple normalized pixels form a normalized image.

6. The machine learning-based protected area bird monitoring and protection system according to claim 4, characterized in that, Perform pixel value transformation on the normalized image to obtain a sample image, including: Obtain a preset mapped pixel interval; Transform the pixel values in the normalized image that are within the preset mapped pixel interval into 1, and transform the remaining pixel values into 0 to obtain a sample image.

7. The machine learning-based protected area bird monitoring and protection system according to claim 6, characterized in that It is characterized in that The preset mapping model includes: Calculate the characteristic value through the following formula: V = ∑c × p; Where, V represents the characteristic value corresponding to the input image of a preset mapping model, p represents a pixel value in the input image, c represents the preset coefficient corresponding to each pixel value, and among them, the preset coefficients are respectively different powers of two.

8. The machine learning-based protected area bird monitoring and protection system according to claim 3, characterized in that, Based on the correspondence relationship among the initial bird images, sample marker values, sample images, first bird images, and first bird data, summarize all the initial bird images to obtain a second bird image and the corresponding second bird data, including: Judge the coverage relationship of multiple initial bird images; If the number of other initial bird images covered by an initial bird image exceeds a preset threshold, then use the covered initial bird image as the second bird image; If the number of other initial bird images covered by an initial bird image does not exceed the preset threshold, then use the covering initial bird image as the second bird image; Based on the correspondence relationship among the second bird image, the initial bird image, the sample marker value, the sample image, the first bird image, and the first bird data, use the first bird data corresponding to the second bird image as its corresponding second bird data.

9. The machine learning-based protected area bird monitoring and protection system according to claim 1, wherein, Each piece of monitoring data is temporarily stored in the cache module and then deleted regularly; Judge whether the monitoring data in the cache module meets the preset trigger conditions, including: Judge whether the cache module has reached the capacity limit; If so, call the fast monitoring module for monitoring; If not, call the default monitoring module for monitoring.

10. The machine learning-based protected area bird monitoring and protection system according to claim 1, characterized in that Judge whether the monitoring data in the cache module meets the preset trigger conditions, including: Judge whether the number of monitoring data of the same kind of bird saved in the cache module within a preset time reaches a preset quantity threshold; If so, call the fast monitoring module for monitoring; If not, call the default monitoring module for monitoring.

Images