Biological species monitoring method and system based on image preprocessing

Abstract

The invention discloses a biological species monitoring method and system based on image preprocessing, and relates to the field of biological species monitoring, and the method comprises the steps: obtaining a habit activity rule of organisms in a to-be-monitored region, and building a knowledge base of the habits of the organisms in the to-be-monitored region; obtaining a biological image of a to-be-monitored area, and recording biological monitoring visual information; according to the biological image of the to-be-monitored area, preprocessing the collected biological image of the to-be-monitored area; establishing a biological recognition model of the attention mechanism, and recognizing organisms in the to-be-monitored area; automatically adjusting a clipping area of the biological species image, and recording biological aggregation information; and establishing a biological species monitoring platform and a biological species monitoring information base for recording and displaying activity information of organisms in the to-be-monitored area. The method has the advantages that through a biological habit-biological monitoring visual information-biological image three-mode data fusion mode, the limitation of a single mode is effectively solved, and the comprehensiveness of biological species monitoring is dynamically improved.

Description

Technical Field

[0001] The present invention relates to the field of biological species monitoring, and in particular to a biological species monitoring method and system based on image preprocessing. Background Art

[0002] The composition and distribution of biological species are important components of the ecosystem. Different species are interdependent and mutually restrictive. By monitoring information such as the types, numbers, distribution and living habits of biological species, we can effectively understand the structure and function of the ecosystem. By monitoring changes in the habits and numbers of biological species, it helps to promote the evolution and development of the ecosystem, detect abnormal events in a timely manner, and improve the stability and ecological harmony of the ecosystem through intervention measures. Secondly, biological species monitoring helps to detect endangered species and threatened species in a timely manner, understand their survival status and endangering factors, and thus formulate targeted protection measures, improve the survival and reproduction ability of endangered species, protect the biodiversity of the ecosystem, and provide technical support for the Earth's life community.

[0003] Existing biological species monitoring technologies mostly obtain the morphological characteristics and behavioral patterns of monitored organisms through image analysis. However, traditional image processing relies on manual intervention, the feature extraction is single, and it is easily affected by the environment, resulting in insufficient robustness and feature redundancy in biological species monitoring in complex environments. Secondly, traditional methods such as histogram equalization and Gaussian filtering do not fully consider the impact of biological habits on image quality, such as the luminescence of the eyes and fur of nocturnal animals and the differences in habits of similar biological species, resulting in low feature extraction efficiency. In addition, the unified preprocessing process uses the same computing power for all scenes, and does not optimize the allocation of computing resources based on biological habits, resulting in low efficiency in biological species monitoring and poor adaptability to dynamic environments. Summary of the Invention

[0004] In order to solve the above technical problems, a biological species monitoring method and system based on image preprocessing are provided. This technical solution solves the problems raised in the above background technology, such as reliance on manual intervention, single feature extraction, and susceptibility to environmental influences, which lead to insufficient robustness and feature redundancy of biological species monitoring in complex environments. Secondly, traditional methods such as histogram equalization and Gaussian filtering do not fully consider the impact of biological habits on image quality, such as the luminescence of the eyes and fur of nocturnal animals and differences in the habits of similar biological species, resulting in low feature extraction efficiency. In addition, the unified preprocessing process uses the same computing power for all scenes, and does not optimize the allocation of computing resources based on biological habits, resulting in low efficiency of biological species monitoring and poor adaptability to dynamic environments.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] A biological species monitoring method based on image preprocessing, comprising:

[0007] Based on big data, obtain the habits and activity patterns of organisms in the monitored area and establish a knowledge base of the habits of organisms in the monitored area;

[0008] According to the activity trajectory of the organisms in the monitored area, biological species monitoring devices are set at different trajectory nodes to obtain images of the organisms in the monitored area and record the biological monitoring visual information;

[0009] Preprocessing the collected biological images of the area to be monitored is performed based on the biological images of the area to be monitored and combined with the biological monitoring visual information of the same sequence;

[0010] Based on the biological image feature data of the monitored area and the knowledge base of the biological habits in the monitored area, a biological recognition model with attention mechanism is established to identify the biological organisms in the monitored area;

[0011] Based on the biometric identification results of the monitored area, the cropping area of ​​the biological species image is automatically adjusted for biological species aggregation sites or sudden biological aggregation behaviors, and biological aggregation information is recorded;

[0012] Establish a biological species monitoring platform and a biological species monitoring information database to record and display the activity information of organisms in the monitored area.

[0013] Preferably, the preprocessing of the collected biological images of the area to be monitored in combination with the biological monitoring visual information of the same sequence specifically includes:

[0014] Perform grayscale conversion and Gaussian filtering on the collected biological images of the area to be monitored;

[0015] Kalman filtering and normalization are used to process the simultaneous biological monitoring visual information data;

[0016] According to the simultaneous biological monitoring visual information, the mathematical mapping relationship between biological monitoring vision and image processing parameters is obtained, and a dynamic multi-dimensional feature quantization convolution kernel is established;

[0017] Based on the convolutional neural network algorithm, the dynamic multi-dimensional feature quantization convolution kernel is used to extract features from the filtered biological image of the monitored area.

[0018] Preferably, the method of establishing a biometric recognition model with an attention mechanism based on the image feature data of the organisms in the monitored area in combination with the knowledge base of the habits of the organisms in the monitored area, and identifying the organisms in the monitored area specifically includes:

[0019] Construct a biological habit feature vector based on the knowledge base of biological habits in the area to be monitored;

[0020] By using the Bayesian formula, the biological habits of the monitored area are used as prior knowledge, and the network's focus area on biological morphological features is dynamically adjusted to generate Bayesian fusion attention.

[0021] Mapping Bayesian fusion attention to attention control parameters through a fully connected layer;

[0022] Based on convolutional neural networks, a biometric recognition model with attention mechanism is established, and the organisms in the monitored area are identified by controlling the attention parameters.

[0023] Preferably, the method of automatically adjusting the cropping area of ​​the biological species image based on the biometric identification results of the area to be monitored and targeting biological species gathering places or sudden biological gathering behaviors, and recording biological gathering information specifically includes:

[0024] Based on the multi-angle collection of images of organisms at each trajectory node in the monitored area, the three-dimensional posture data of the organisms in the monitored area are obtained;

[0025] Establish a spatial rectangular coordinate system and obtain the geometric center coordinates of the biological species through the three-dimensional posture data of the organisms in the monitored area;

[0026] Based on the three-dimensional posture data of the organisms in the monitored area, the discrete degree value of the individual distance between the species is obtained according to the Euclidean distance formula;

[0027] According to the standard deviation of the distance between individuals of biological species, the dynamic adjustment expansion coefficient of the biological species image clipping is determined;

[0028] Based on the dynamically adjusted expansion coefficient of the biological species image clipping, the clipping area of ​​the biological species image is automatically adjusted according to the biological species aggregation or sudden biological aggregation behavior;

[0029] Mapping the clipping area of ​​the acquired three-dimensional biological species image into a two-dimensional space, acquiring and outputting the two-dimensional clipping area of ​​the biological species image;

[0030] According to the two-dimensional cropping area of ​​the biological species image, the biological aggregation information is recorded and counted. By setting the threshold, when the number and frequency of aggregation exceed the preset threshold, an abnormal biological monitoring alarm is issued.

[0031] Furthermore, this solution proposes a biological species monitoring system based on image preprocessing, which is used to implement the above-mentioned biological species monitoring method based on image preprocessing, including:

[0032] A biological habit module, which is used to obtain the habit and activity patterns of organisms in the monitored area based on big data and establish a knowledge base of the habits of organisms in the monitored area;

[0033] A monitoring data module is used to set up biological species monitoring devices at different trajectory nodes according to the activity trajectory of the biological habits in the monitored area, obtain images of the biological organisms in the monitored area, and record biological monitoring visual information;

[0034] A biological monitoring module is configured to pre-process the collected biological images of the monitored area based on the biological images and the simultaneous biological monitoring visual information; establish a biological recognition model with an attention mechanism based on the characteristic data of the biological images of the monitored area and the knowledge base of the habits of the organisms in the monitored area to identify the organisms in the monitored area; and automatically adjust the cropping area of ​​the biological species images based on the biological species gathering places or sudden biological gathering behaviors, and record the biological gathering information according to the biological recognition results of the monitored area;

[0035] The monitoring platform module is used to establish a biological species monitoring platform and a biological species monitoring information database, and is used to record and display activity information of organisms in the area to be monitored.

[0036] Preferably, the biological monitoring module includes:

[0037] An image preprocessing unit, configured to preprocess the collected biological images of the area to be monitored based on the biological images of the area to be monitored and in combination with the biological monitoring visual information of the same sequence;

[0038] A biometric recognition unit, which is used to establish a biometric recognition model with an attention mechanism based on the image feature data of the organisms in the monitored area and a knowledge base of the habits of the organisms in the monitored area, so as to identify the organisms in the monitored area;

[0039] The biological aggregation unit is used to automatically adjust the cropping area of ​​the biological species image and record the biological aggregation information based on the biological recognition results of the area to be monitored and the biological species aggregation area or sudden biological aggregation behavior.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] Obtain the habits and activity patterns of organisms in the monitored area and establish a knowledge base of the habits of organisms in the monitored area. Secondly, based on the simultaneous biological monitoring visual information, obtain the mathematical mapping relationship between biological monitoring vision and image processing parameters, establish a dynamic multi-dimensional feature quantization convolution kernel, and based on the convolutional neural network algorithm, extract features from the filtered biological images of the monitored area through the dynamic multi-dimensional feature quantization convolution kernel. Furthermore, based on the knowledge base of the biological habits of the monitored area, introduce the biological habit feature vector, map the biological habit feature vector to the attention control parameter through the fully connected layer, and thus establish a biological recognition model of the attention mechanism. Through the attention control parameter, the organisms in the monitored area are identified. Finally, for biological species gathering places or sudden biological gathering behaviors, the cropping area of ​​the biological species image is automatically adjusted, and the biological gathering information is recorded. By setting a threshold, when the number and frequency of gatherings exceed the preset threshold, an abnormal biological monitoring alarm is issued. Therefore, through the three-modal data fusion method of biological habits, biological monitoring visual information, and biological images, the limitations of a single modality are effectively solved and the comprehensiveness of biological species monitoring is dynamically improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of the biological species monitoring method based on image preprocessing of the present invention;

[0043] Figure 2 This is a flow chart of preprocessing the collected biological images of the area to be monitored based on the biological images of the area to be monitored and combined with the biological monitoring visual information of the same sequence;

[0044] Figure 3 This is a flowchart of the present invention for identifying organisms in the monitored area by establishing a biometric recognition model with an attention mechanism based on the image feature data of the organisms in the monitored area, combined with the knowledge base of the habits of the organisms in the monitored area;

[0045] Figure 4 The present invention is a flow chart for automatically adjusting the cropping area of ​​biological species images and recording biological aggregation information based on the biological recognition results of the area to be monitored and targeting biological species gathering places or sudden biological aggregation behaviors;

[0046] Figure 5 The present invention establishes a biological species monitoring platform and a biological species monitoring information database for recording and displaying a flow chart of the activity information of the organisms in the area to be monitored. DETAILED DESCRIPTION

[0047] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0048] Reference Figure 1As shown, a biological species monitoring method based on image preprocessing includes:

[0049] Based on big data, obtain the habits and activity patterns of organisms in the monitored area and establish a knowledge base of the habits of organisms in the monitored area;

[0050] According to the activity trajectory of the organisms in the monitored area, biological species monitoring devices are set at different trajectory nodes to obtain images of the organisms in the monitored area and record the biological monitoring visual information;

[0051] Preprocessing the collected biological images of the area to be monitored is performed based on the biological images of the area to be monitored and combined with the biological monitoring visual information of the same sequence;

[0052] Based on the biological image feature data of the monitored area and the knowledge base of the biological habits in the monitored area, a biological recognition model with attention mechanism is established to identify the biological organisms in the monitored area;

[0053] Based on the biometric identification results of the monitored area, the cropping area of ​​the biological species image is automatically adjusted for biological species aggregation sites or sudden biological aggregation behaviors, and biological aggregation information is recorded;

[0054] Establish a biological species monitoring platform and a biological species monitoring information database to record and display the activity information of organisms in the monitored area.

[0055] It can be explained that this scheme establishes a knowledge base of the habits of organisms in the monitored area by obtaining the habits and activity patterns of the organisms in the monitored area. Secondly, based on the simultaneous biological monitoring visual information, the mathematical mapping relationship between biological monitoring vision and image processing parameters is obtained, and a dynamic multi-dimensional feature quantization convolution kernel is established. Based on the convolutional neural network algorithm, the dynamic multi-dimensional feature quantization convolution kernel is used to extract features from the filtered biological images of the monitored area. Furthermore, based on the knowledge base of the biological habits of the monitored area, the biological habit feature vector is introduced, and the biological habit feature vector is mapped into attention control parameters through the fully connected layer, thereby establishing a biological recognition model of the attention mechanism. The attention control parameters are used to identify the organisms in the monitored area. Finally, for biological species gathering places or sudden biological gathering behaviors, the cropping area of ​​the biological species image is automatically adjusted, and the biological gathering information is recorded. By setting a threshold, when the number and frequency of gatherings exceed the preset threshold, an abnormal biological monitoring alarm is issued. Therefore, through the three-modal data fusion method of biological habits, biological monitoring visual information, and biological images, the limitations of a single modality are effectively solved and the comprehensiveness of biological species monitoring is dynamically improved.

[0056] The method of obtaining the habits and activity patterns of organisms in the monitored area based on big data and establishing a knowledge base of the habits of organisms in the monitored area specifically includes:

[0057] The characteristics of biological habits and activities in the monitored area are divided into three parts: temporal activity characteristics, spatial distribution characteristics and behavioral pattern characteristics;

[0058] Among them, temporal activity characteristics include: diurnal activity, seasonal migration, spatial distribution characteristics include: population density, habitat preference, behavioral pattern characteristics include: foraging, courtship, camouflage and territorial awareness;

[0059] Based on big data, characteristic maps of biological habits and activities in the area to be monitored are obtained, and a knowledge base of biological habits in the area to be monitored is established.

[0060] It can be explained that there is a very close connection between the activities of biological species and their biological habits. Their temporal activity characteristics, spatial distribution characteristics and behavioral pattern characteristics have certain activity patterns, such as diurnal activities, territorial awareness, etc. Therefore, when monitoring biological species, by introducing biological habits and parameterizing them in an image-recognizable manner, the dynamic tracking effect of biological species monitoring can be effectively improved.

[0061] The method of setting up biological species monitoring devices at different trajectory nodes according to the activity trajectory of the biological habits in the monitored area, acquiring biological images of the monitored area, and recording biological monitoring visual information specifically includes:

[0062] According to the habitual activity trajectory of the organisms in the monitored area, biological species monitoring devices are set up at different trajectory nodes. Among them, multi-view image acquisition method is used at the places where biological species gather;

[0063] A synchronous sampling method is used to obtain biological images at each trajectory node in the monitored area, and a visual perception device is used to obtain simultaneous biological monitoring visual information at each trajectory node in the monitored area;

[0064] Among them, biological monitoring visual information includes: photosensitivity, spatiotemporal resolution, field of view, color discrimination, and light adaptation ability;

[0065] Based on the Internet of Things technology, biological images and biological monitoring visual information at each trajectory node in the monitored area are periodically obtained, and a biological species monitoring database for each trajectory node is established.

[0066] It can be explained that biological monitoring visual information is the most direct and abundant data source in biological monitoring. Core functions such as species identification, behavior analysis, and habitat assessment can be realized through images, videos and other data. This scheme adopts a synchronous sampling method to obtain biological images and biological monitoring visual information with the same time sequence, and establishes a biological species monitoring database for each trajectory node, thereby providing data support for subsequent feature extraction of biological images. Among them, the synchronous sampling method refers to the same sampling frequency.

[0067] Reference Figure 2 As shown, the pre-processing of the collected biological images of the area to be monitored in combination with the biological monitoring visual information of the same sequence specifically includes:

[0068] Perform grayscale conversion and Gaussian filtering on the collected biological images of the area to be monitored;

[0069] Kalman filtering and normalization are used to process the simultaneous biological monitoring visual information data;

[0070] According to the simultaneous biological monitoring visual information, the mathematical mapping relationship between biological monitoring vision and image processing parameters is obtained, and a dynamic multi-dimensional feature quantization convolution kernel is established;

[0071] Based on the convolutional neural network algorithm, the dynamic multi-dimensional feature quantization convolution kernel is used to extract features from the filtered biological image of the monitored area.

[0072] It can be explained that due to the different habits of different biological species, the image collection of different types of biological species at different times will be affected by the biological monitoring visual information, resulting in poor quality of biological feature extraction and thus errors in the judgment of the type of biological species. Therefore, this scheme obtains the mathematical mapping relationship between biological monitoring vision and image processing parameters based on the simultaneous biological monitoring visual information, establishes a dynamic multi-dimensional feature quantization convolution kernel, and based on the convolutional neural network algorithm, extracts features from the filtered biological image of the monitored area through the dynamic multi-dimensional feature quantization convolution kernel, thereby improving the feature quality of the biological image of the monitored area. Among them, the discrete two-dimensional convolution formula of the biological image and the dynamic multi-dimensional feature quantization convolution kernel is:

[0073] G(i, j) = ∑ m ∑ n I(i+m,j+n)·K(m,n)

[0074] Where G(i, j) is the value of the output feature map of the biological image at position (i, j), I(·) is the input feature matrix of the biological image, K(·) is the dynamic multidimensional feature quantization convolution kernel, m and n are the relative coordinate offsets within the convolution kernel, and (i, j) is the position coordinate of the pixel point.

[0075] Reference Figure 3 As shown, the method of establishing a biometric recognition model with an attention mechanism based on the image feature data of the biological organisms in the monitored area and combining it with the knowledge base of the habits of the biological organisms in the monitored area specifically includes:

[0076] Construct a biological habit feature vector based on the knowledge base of biological habits in the area to be monitored;

[0077] By using the Bayesian formula, the biological habits of the monitored area are used as prior knowledge, and the network's focus area on biological morphological features is dynamically adjusted to generate Bayesian fusion attention.

[0078] Mapping Bayesian fusion attention to attention control parameters through a fully connected layer;

[0079] Based on convolutional neural networks, a biometric recognition model with attention mechanism is established, and the organisms in the monitored area are identified by controlling the attention parameters.

[0080] It can be explained that due to the different types and habits of biological species in the monitored area, there is complexity and diversity in the process of biological species monitoring, which is not conducive to the stability of biological species monitoring. This scheme constructs a biological habit feature vector through the knowledge base of biological habits in the monitored area, generates a unique, discrete species ID, time activity characteristics, spatial distribution characteristics and a unified coding feature vector of behavioral pattern characteristics, and uses the Bayesian formula to take the biological habit feature vector of the monitored area as prior knowledge to generate the Bayesian posterior weight of dynamic attention with biological habits, thereby solving the problem of multi-source data fragmentation and the problem of ignoring biological habits in the traditional biological species monitoring process. Among them, the Bayesian formula is used to take the biological habit feature vector of the monitored area as prior knowledge as the expression:

[0081] P(V)=S(W P V S +b P )

[0082] Where P(V) is the prior probability value of extracting the biological habit parameters from the knowledge base of the biological habits of the area to be monitored, S(·) is the Sigmoid activation function, and W P is the Bayesian learnable parameter matrix, V S is the biological habit characteristic vector, b P is the Bayesian learnable bias;

[0083] The Bayesian posterior weight expression is:

[0084]

[0085] Where P(V|G) is the posterior probability value of the association between the biological habit parameters and the biological image features, P(G|V) is the characteristic likelihood value of the biological species, that is, the convolution value of the biological image output feature map G(i, j) and the Gaussian kernel, P(G|V i ) is the likelihood value of the biological species characteristics corresponding to the i-th biological habit, P(V i) is the prior probability value of the biological habit parameters extracted from the knowledge base of biological habits in the monitored area corresponding to the i-th biological habit, k is the number of biological habit types, and Θ is the element-by-element multiplication symbol.

[0086] The weight expression for mapping Bayesian fusion attention to attention control parameters is:

[0087] w=S[W w P(V|G)+B w ]

[0088] Where w is the weight that maps Bayesian fusion attention to attention control parameters, S(·) is the Sigmoid activation function, and W w is the attention learnable parameter matrix, B w is the learnable bias for attention.

[0089] Reference Figure 4 As shown, the method of automatically adjusting the cropping area of ​​the biological species image based on the biometric identification results of the monitored area for biological species gathering places or sudden biological gathering behaviors, and recording biological gathering information specifically includes:

[0090] Based on the multi-angle collection of images of organisms at each trajectory node in the monitored area, the three-dimensional posture data of the organisms in the monitored area are obtained;

[0091] Establish a spatial rectangular coordinate system and obtain the geometric center coordinates of the biological species through the three-dimensional posture data of the organisms in the monitored area;

[0092] Based on the three-dimensional posture data of the organisms in the monitored area, the discrete degree value of the individual distance between the species is obtained according to the Euclidean distance formula;

[0093] According to the standard deviation of the distance between individuals of biological species, the dynamic adjustment expansion coefficient of the biological species image clipping is determined;

[0094] Based on the dynamically adjusted expansion coefficient of the biological species image clipping, the clipping area of ​​the biological species image is automatically adjusted according to the biological species aggregation or sudden biological aggregation behavior;

[0095] Mapping the clipping area of ​​the acquired three-dimensional biological species image into a two-dimensional space, acquiring and outputting the two-dimensional clipping area of ​​the biological species image;

[0096] According to the two-dimensional cropping area of ​​the biological species image, the biological aggregation information is recorded and counted. By setting the threshold, when the number and frequency of aggregation exceed the preset threshold, an abnormal biological monitoring alarm is issued.

[0097] It can be explained that when conducting biological species monitoring, extra attention needs to be paid to biological species aggregation sites or sudden biological aggregation behaviors. Biological aggregation sites are the core areas for species reproduction, foraging and habitat. Biological species aggregation sites or sudden biological aggregation behaviors can effectively reflect the habits of biological species and the response of organisms to environmental mutations or resource fluctuations. Therefore, in the process of biological species monitoring, additional aggregation information of biological species needs to be obtained. By judging whether the number and frequency of aggregations exceed the preset threshold, if so, an abnormal biological monitoring alarm is issued. If not, it means that everything is normal. Among them, this solution adopts dynamic ROI clipping to automatically adjust the clipping area of ​​the biological species image for biological species aggregation sites or sudden biological aggregation behaviors, thereby achieving cross-species adaptive feature enhancement and dynamically adjusting to the optimal perception range of the target species. The specific implementation process is as follows:

[0098] The expression for obtaining the geometric center coordinates of the biological species is:

[0099]

[0100] Where R xyz is the geometric center coordinate of the biological species, (x i ,y i , z i ) is the coordinate of the i-th organism in the rectangular coordinate system, and N is the number of individuals in the species;

[0101] The expression for obtaining the discrete degree value of the individual spacing between biological species is:

[0102]

[0103] In the formula, q is the discrete degree value of the distance between individuals of biological species, r i is the coordinate of the i-th organism in the rectangular coordinate system, that is, r i =(x i ,y i , z i ),||r i -R xyz || is the Euclidean distance formula, which is used to reflect the degree of discreteness between individual distances between biological species;

[0104] The expression for the dynamic adjustment expansion coefficient for determining the cropping of biological species images is:

[0105] α=b+cln(1+q)

[0106] Where α is the dynamic adjustment expansion coefficient of the biological species image clipping, b and c are constant term coefficients;

[0107] The expression for automatically adjusting the clipping radius of the biological species image is:

[0108] R=αmax(||r i -R xyz ||)

[0109] Where R is the cropping radius for automatically adjusting the biological species image.

[0110] Reference Figure 5 As shown, the establishment of a biological species monitoring platform and a biological species monitoring information database for recording and displaying the activity information of organisms in the monitored area specifically includes:

[0111] Establish a biological species monitoring platform and use IoT technology to obtain biometric information and activity data of trajectory nodes in the monitored area;

[0112] Based on the biological species monitoring platform, according to the biological recognition results, the biological activity data files are automatically classified according to the biological type. The biological activity data files include: biological type, biological quantity, biological behavior, biological image capture time and biological monitoring visual information;

[0113] Establish a biological species monitoring information database to record and save the activity data of organisms in the monitored area for easy subsequent access;

[0114] Based on the biological species monitoring platform, the activity information of organisms in the monitored area is called and displayed through the human-computer interactive interface.

[0115] It can be explained that biological species monitoring is a long-term and continuous monitoring of biological species activities. In order to effectively preserve biological species activity data, this plan needs to use Internet of Things technology to establish a biological species monitoring platform for long-term acquisition of biometric information and activity data of trajectory nodes in the monitored area, and establish a biological species monitoring information database to record and save the activity data of organisms in the monitored area to facilitate subsequent calls and provide data support for the analysis of seasonal activities and habits of biological species.

[0116] Furthermore, based on the same inventive concept as the above-mentioned biological species monitoring method based on image preprocessing, this solution proposes a biological species monitoring system based on image preprocessing, comprising:

[0117] A biological habit module, which is used to obtain the habit and activity patterns of organisms in the monitored area based on big data and establish a knowledge base of the habits of organisms in the monitored area;

[0118] A monitoring data module is used to set up biological species monitoring devices at different trajectory nodes according to the activity trajectory of the biological habits in the monitored area, obtain images of the biological organisms in the monitored area, and record biological monitoring visual information;

[0119] A biological monitoring module is configured to pre-process the collected biological images of the monitored area based on the biological images and the simultaneous biological monitoring visual information; establish a biological recognition model with an attention mechanism based on the characteristic data of the biological images of the monitored area and the knowledge base of the habits of the organisms in the monitored area to identify the organisms in the monitored area; and automatically adjust the cropping area of ​​the biological species images based on the biological species gathering places or sudden biological gathering behaviors, and record the biological gathering information according to the biological recognition results of the monitored area;

[0120] A monitoring platform module, which is used to establish a biological species monitoring platform and a biological species monitoring information database, and is used to record and display activity information of organisms in the monitored area;

[0121] The biomonitoring module includes:

[0122] An image preprocessing unit, configured to preprocess the collected biological images of the area to be monitored based on the biological images of the area to be monitored and in combination with the biological monitoring visual information of the same sequence;

[0123] A biometric recognition unit, which is used to establish a biometric recognition model with an attention mechanism based on the image feature data of the organisms in the monitored area and a knowledge base of the habits of the organisms in the monitored area, so as to identify the organisms in the monitored area;

[0124] The biological aggregation unit is used to automatically adjust the cropping area of ​​the biological species image and record the biological aggregation information based on the biological recognition results of the area to be monitored and the biological species aggregation area or sudden biological aggregation behavior.

[0125] In summary, the advantages of the present invention are: through the three-modal data fusion method of biological habits, biological monitoring visual information and biological images, it effectively solves the limitations of a single modality and dynamically improves the comprehensiveness of biological species monitoring.

[0126] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A biological species monitoring method based on image preprocessing, characterized in that: include: Based on big data, obtain the habits and activity patterns of organisms in the monitored area and establish a knowledge base of the habits of organisms in the monitored area; According to the activity trajectory of the organisms in the monitored area, biological species monitoring devices are set at different trajectory nodes to obtain images of the organisms in the monitored area and record the biological monitoring visual information; Preprocessing the collected biological images of the area to be monitored is performed based on the biological images of the area to be monitored and combined with the biological monitoring visual information of the same sequence; Based on the biological image feature data of the monitored area and the knowledge base of the biological habits in the monitored area, a biological recognition model with attention mechanism is established to identify the biological organisms in the monitored area; Based on the biometric identification results of the monitored area, the cropping area of ​​the biological species image is automatically adjusted for biological species aggregation sites or sudden biological aggregation behaviors, and biological aggregation information is recorded; Establish a biological species monitoring platform and a biological species monitoring information database to record and display the activity information of organisms in the monitored area.

2. The biological species monitoring method based on image preprocessing according to claim 1, characterized in that: The method of obtaining the habits and activity patterns of organisms in the monitored area based on big data and establishing a knowledge base of the habits of organisms in the monitored area specifically includes: The characteristics of biological habits and activities in the monitored area are divided into three parts: temporal activity characteristics, spatial distribution characteristics and behavioral pattern characteristics; Among them, temporal activity characteristics include: diurnal activity, seasonal migration, spatial distribution characteristics include: population density, habitat preference, behavioral pattern characteristics include: foraging, courtship, camouflage and territorial awareness; Based on big data, characteristic maps of biological habits and activities in the area to be monitored are obtained, and a knowledge base of biological habits in the area to be monitored is established.

3. The biological species monitoring method based on image preprocessing according to claim 2, characterized in that: The method of setting up biological species monitoring devices at different trajectory nodes according to the activity trajectory of the biological habits in the monitored area, acquiring biological images of the monitored area, and recording biological monitoring visual information specifically includes: According to the habitual activity trajectory of the organisms in the monitored area, biological species monitoring devices are set up at different trajectory nodes. Among them, multi-view image acquisition method is used at the places where biological species gather; A synchronous sampling method is used to obtain biological images at each trajectory node in the monitored area, and a visual perception device is used to obtain simultaneous biological monitoring visual information at each trajectory node in the monitored area; Among them, biological monitoring visual information includes: photosensitivity, spatiotemporal resolution, field of view, color discrimination, and light adaptation ability; Based on the Internet of Things technology, biological images and biological monitoring visual information at each trajectory node in the monitored area are periodically obtained, and a biological species monitoring database for each trajectory node is established.

4. The biological species monitoring method based on image preprocessing according to claim 3, characterized in that: The preprocessing of the collected biological images of the area to be monitored in combination with the biological monitoring visual information of the same sequence specifically includes: Perform grayscale conversion and Gaussian filtering on the collected biological images of the area to be monitored; Kalman filtering and normalization are used to process the simultaneous biological monitoring visual information data; According to the simultaneous biological monitoring visual information, the mathematical mapping relationship between biological monitoring vision and image processing parameters is obtained, and a dynamic multi-dimensional feature quantization convolution kernel is established; Based on the convolutional neural network algorithm, the dynamic multi-dimensional feature quantization convolution kernel is used to extract features from the filtered biological image of the monitored area.

5. The biological species monitoring method based on image preprocessing according to claim 4, characterized in that: The method of establishing a biometric recognition model with an attention mechanism based on the image feature data of the biological organisms in the monitored area and combining it with the knowledge base of the habits of the biological organisms in the monitored area to identify the biological organisms in the monitored area specifically includes: Construct a biological habit feature vector based on the knowledge base of biological habits in the area to be monitored; By using the Bayesian formula, the biological habits of the monitored area are used as prior knowledge, and the network's focus area on biological morphological features is dynamically adjusted to generate Bayesian fusion attention. Mapping Bayesian fusion attention to attention control parameters through a fully connected layer; Based on convolutional neural networks, a biometric recognition model with attention mechanism is established, and the organisms in the monitored area are identified by controlling the attention parameters.

6. The biological species monitoring method based on image preprocessing according to claim 5, characterized in that: The method of automatically adjusting the cropping area of ​​the biological species image based on the biometric identification results of the monitored area and targeting biological species gathering places or sudden biological gathering behaviors, and recording biological gathering information specifically includes: Based on the multi-angle collection of images of organisms at each trajectory node in the monitored area, the three-dimensional posture data of the organisms in the monitored area are obtained; Establish a spatial rectangular coordinate system and obtain the geometric center coordinates of the biological species through the three-dimensional posture data of the organisms in the monitored area; Based on the three-dimensional posture data of the organisms in the monitored area, the discrete degree value of the individual distance between the species is obtained according to the Euclidean distance formula; According to the standard deviation of the distance between individuals of biological species, the dynamic adjustment expansion coefficient of the biological species image clipping is determined; Based on the dynamically adjusted expansion coefficient of the biological species image clipping, the clipping area of ​​the biological species image is automatically adjusted according to the biological species aggregation or sudden biological aggregation behavior; Mapping the clipping area of ​​the acquired three-dimensional biological species image into a two-dimensional space, acquiring and outputting the two-dimensional clipping area of ​​the biological species image; According to the two-dimensional cropping area of ​​the biological species image, the biological aggregation information is recorded and counted. By setting the threshold, when the number and frequency of aggregation exceed the preset threshold, an abnormal biological monitoring alarm is issued.

7. The biological species monitoring method based on image preprocessing according to claim 6, characterized in that: The establishment of a biological species monitoring platform and a biological species monitoring information database for recording and displaying the activity information of organisms in the monitored area specifically includes: Establish a biological species monitoring platform and use IoT technology to obtain biometric information and activity data of trajectory nodes in the monitored area; Based on the biological species monitoring platform, according to the biological recognition results, the biological activity data files are automatically classified according to the biological type. The biological activity data files include: biological type, biological quantity, biological behavior, biological image capture time and biological monitoring visual information; Establish a biological species monitoring information database to record and save the activity data of organisms in the monitored area for easy subsequent access; Based on the biological species monitoring platform, the activity information of organisms in the monitored area is called and displayed through the human-computer interactive interface.

8. A biological species monitoring system based on image preprocessing, characterized in that: A method for implementing a biological species monitoring method based on image preprocessing according to any one of claims 1 to 7, comprising: A biological habit module, which is used to obtain the habit and activity patterns of organisms in the monitored area based on big data and establish a knowledge base of the habits of organisms in the monitored area; A monitoring data module is used to set up biological species monitoring devices at different trajectory nodes according to the activity trajectory of the biological habits in the monitored area, obtain images of the biological organisms in the monitored area, and record biological monitoring visual information; A biological monitoring module is configured to pre-process the collected biological images of the monitored area based on the biological images and the simultaneous biological monitoring visual information; establish a biological recognition model with an attention mechanism based on the characteristic data of the biological images of the monitored area and the knowledge base of the habits of the organisms in the monitored area to identify the organisms in the monitored area; and automatically adjust the cropping area of ​​the biological species images based on the biological species gathering places or sudden biological gathering behaviors, and record the biological gathering information according to the biological recognition results of the monitored area; The monitoring platform module is used to establish a biological species monitoring platform and a biological species monitoring information database, and is used to record and display activity information of organisms in the area to be monitored.

9. The biological species monitoring system based on image preprocessing according to claim 8, characterized in that: The biological monitoring module includes: An image preprocessing unit, configured to preprocess the collected biological images of the area to be monitored based on the biological images of the area to be monitored and in combination with the biological monitoring visual information of the same sequence; A biometric recognition unit, which is used to establish a biometric recognition model with an attention mechanism based on the image feature data of the organisms in the monitored area and a knowledge base of the habits of the organisms in the monitored area, so as to identify the organisms in the monitored area; The biological aggregation unit is used to automatically adjust the cropping area of ​​the biological species image and record the biological aggregation information based on the biological recognition results of the area to be monitored and the biological species aggregation area or sudden biological aggregation behavior.

Images