A bird-repelling method and system based on radar laser technology

Abstract

The present invention provides a bird-repelling method and system based on radar laser technology, which relates to the field of image processing technology. The method collects and segments regional images to obtain multiple bird images, combines historical bird identification with a bird-repelling database, marks new birds and adds them to the database, retrieves historical laser parameters of historical birds and performs light resistance analysis, adjusts and optimizes current laser parameters to perform bird-repelling operations. The method solves the technical problem in the prior art that birds have adaptability to the same type of laser, and when the same laser is used to repel the same bird multiple times, the tolerance to the laser leads to a decrease in the bird-repelling effect, and the problem of poor bird-repelling effect due to the adaptability of the laser bird-repelling. The method divides historical birds and new birds into designated areas, performs light resistance analysis based on the bird-repelling data of historical birds, and optimizes the adaptability of laser parameters based on actual bird-repelling conditions, thereby avoiding the poor bird-repelling effect caused by laser tolerance.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a bird-repelling method and system based on radar laser technology. Background Art

[0002] Bird activities are unavoidable natural activities, and corresponding preventive measures must be taken to drive them away to avoid threatening losses, such as the safe operation of power transmission lines, airports, power equipment, etc.; intrusion in orchards and other areas, etc., and bird repellent operations must be carried out in a timely manner.

[0003] Traditional bird-repelling methods in the existing technology are costly and ineffective. Laser bird-repelling methods are less expensive, but birds are adaptable to the same type of laser. When the same laser is used to repel the same bird multiple times, the bird's tolerance to the laser leads to a decrease in the bird-repelling effect. There is a problem that the adaptability of laser bird repellent leads to poor bird-repelling effect. Summary of the Invention

[0004] The present application provides a bird-repelling method and system based on radar laser technology, which is used to solve the technical problem in the prior art that birds have adaptability to the same type of laser, and when the same laser is used to repel the same bird multiple times, the tolerance to the laser leads to a decrease in the bird-repelling effect, and the adaptability of the laser bird-repelling leads to poor bird-repelling effect.

[0005] In view of the above problems, the present application provides a bird-repelling method and system based on radar laser technology.

[0006] In a first aspect, the present application provides a bird-repelling method based on radar laser technology, the method comprising:

[0007] Collecting an image of a designated area to be bird-repellent to obtain a regional image, wherein the regional image includes images of a plurality of birds currently in the designated area;

[0008] performing image segmentation on the bird images of the plurality of birds within the region image to obtain a plurality of bird images;

[0009] Using multiple historical bird images in a bird-repelling database, performing similarity analysis on the multiple bird images, determining whether multiple birds in the multiple bird images are historical birds, and obtaining multiple new birds and multiple historical birds, the bird-repelling database including multiple sets of bird marking information, bird images, and bird-repelling laser parameters;

[0010] Marking the multiple new birds, and adding the bird images of the multiple new birds to the bird repellent database, searching the bird repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repelling of the multiple historical birds within a historical time period;

[0011] Analyzing the light resistance of the historical birds to the multiple laser parameters according to the multiple historical laser parameter sets, obtaining multiple light resistance parameter sets, and constructing a bird repellent light resistance evaluation function;

[0012] According to the bird-repelling light resistance evaluation function and the multiple light resistance parameter sets, the laser parameters currently used for bird repelling are adjusted and optimized to obtain the optimal laser parameters, laser bird repelling is performed on the designated area, and the optimal laser parameters are added to the bird repelling database.

[0013] In a second aspect, the present application provides a bird-repelling system based on radar laser technology, the system comprising:

[0014] An image acquisition module is used to acquire images of a designated area to be bird-repellent, and obtain a regional image, wherein the regional image includes images of multiple birds currently in the designated area;

[0015] an image segmentation module, the image segmentation module being used to perform image segmentation on the bird images of the plurality of birds in the regional image to obtain a plurality of bird images;

[0016] a similarity analysis module, the similarity analysis module being configured to use a plurality of historical bird images in a bird repellent database to perform similarity analysis on the plurality of bird images, determine whether a plurality of birds in the plurality of bird images are historical birds, and obtain a plurality of new birds and a plurality of historical birds, the bird repellent database including a plurality of sets of bird marking information, bird images, and bird repellent laser parameters;

[0017] a laser parameter acquisition module, the laser parameter acquisition module being used to mark the multiple new birds, add the bird images of the multiple new birds to the bird repellent database, and search the bird repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repelling of the multiple historical birds within a historical period;

[0018] A function construction module, the function construction module being used to analyze the light resistance of the plurality of historical birds to a plurality of laser parameters based on the plurality of historical laser parameter sets, obtain a plurality of light resistance parameter sets, and construct a bird-repelling light resistance evaluation function;

[0019] A parameter tuning control module is used to adjust and optimize the laser parameters currently used for bird repelling according to the bird repelling light resistance evaluation function and the multiple light resistance parameter sets, obtain optimal laser parameters, perform laser bird repelling on the designated area, and add the optimal laser parameters to the bird repelling database.

[0020] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0021] An embodiment of the present application provides a bird-repelling method based on radar laser technology, which collects images of a designated area to be bird-repellent to obtain an area image, performs image segmentation on bird images of multiple birds in the area image, and obtains multiple bird images; uses multiple historical bird images in a bird-repellent database, performs similarity analysis on the multiple bird images, performs historical bird judgment, obtains multiple new birds and multiple historical birds, marks the multiple new birds, and adds the bird images of the multiple new birds to the bird-repellent database; searches the bird-repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repelling of the multiple historical birds within a historical time. The light resistance of the multiple historical birds to multiple laser parameters is analyzed to obtain multiple light resistance parameter sets, and a bird repellent light resistance evaluation function is constructed. In combination with the multiple light resistance parameter sets, the laser parameters currently used for bird repelling are adjusted and optimized to obtain the optimal laser parameters for laser bird repelling in a specified area, and the optimal laser parameters are added to the bird repellent database. This solves the technical problem in the prior art that birds have adaptability to the same type of laser, and when the same laser is used to repel the same bird multiple times, the tolerance to the laser leads to a decrease in the bird repellent effect, and the adaptability of laser bird repellent leads to poor bird repellent effect. The designated area is divided into historical birds and new birds, and the light resistance analysis is performed based on the bird repellent data of historical birds. The adaptability of laser parameters is optimized based on the actual bird repellent situation to avoid the poor bird repellent effect caused by laser tolerance. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This application provides a flow chart of a bird-repelling method based on radar laser technology;

[0023] Figure 2 This application provides a schematic diagram of a similarity analysis process for multiple bird images in a bird repelling method based on radar laser technology;

[0024] Figure 3 This application provides a schematic diagram of the process of obtaining the optimal laser parameters in a bird-repelling method based on radar laser technology;

[0025] Figure 4 This application provides a schematic structural diagram of a bird repellent system based on radar laser technology.

[0026] Description of the accompanying symbols: image acquisition module 11, image segmentation module 12, similarity analysis module 13, laser parameter acquisition module 14, function construction module 15, parameter tuning control module 16. DETAILED DESCRIPTION

[0027] The present application provides a bird-repelling method and system based on radar laser technology, which collects and segments regional images to obtain multiple bird images, combines historical bird identification with a bird-repelling database, marks new birds and adds them to the database, retrieves historical laser parameters of historical birds and performs light resistance analysis, and adjusts and optimizes current laser parameters to perform bird-repelling operations. The application is used to solve the technical problem in the prior art that birds have adaptability to the same type of laser, and when the same laser is used to repel the same bird multiple times, the tolerance to the laser leads to a decrease in the bird-repelling effect, and the adaptability of the laser bird-repelling leads to poor bird-repelling effect. Example 1

[0028] like Figure 1 As shown, the present application provides a bird-repelling method based on radar laser technology, the method comprising:

[0029] Step S100: collecting an image of a designated area to be bird-repellent, and obtaining a regional image, wherein the regional image includes images of multiple birds currently in the designated area;

[0030] Specifically, bird activities are unavoidable natural activities, and corresponding preventive measures must be taken to drive them away to avoid threatening losses. For example, the operation safety of power transmission lines, airports, power equipment, etc.; in areas such as orchards, bird infestation, etc., requires timely bird repelling operations. This application provides a bird repelling method based on radar laser technology, which identifies bird images in a specified area and divides historical birds into historical birds and new birds. It combines the bird repelling data of historical birds to perform light resistance analysis to optimize laser parameters, and conducts targeted and effective processing based on the actual bird repelling situation, reducing the problem of birds developing tolerance to the same laser and affecting the bird repelling effect, thereby improving the bird repelling effect.

[0031] Specifically, the designated area is the target area for bird repellent operations, such as a power grid or other area requiring bird repellent. Using image acquisition equipment, multi-angle image acquisition is performed on the designated area to ensure that images of multiple birds currently within the designated area are included to ensure complete coverage of image information. Sequential integration of captured images is performed based on the shift in acquisition angles to obtain an image of the area. This image of the area serves as the source data for analysis of the bird repellent operations.

[0032] Step S200: performing image segmentation on the bird images of the plurality of birds in the regional image to obtain a plurality of bird images;

[0033] Furthermore, the bird images of the multiple birds in the regional image are segmented to obtain multiple bird images. Step S200 of the present application further includes:

[0034] Step S210: mining historical area images collected from multiple bird-repelling areas to obtain multiple historical area images, and segmenting and marking bird images in the multiple historical area images to obtain multiple sample bird image segmentation results;

[0035] Step S220: constructing an encoder and a decoder based on semantic segmentation;

[0036] Step S230: using the multiple historical area images and the multiple sample bird image segmentation results, performing supervised training on the encoder and decoder until a convergence condition is met, thereby obtaining a bird image segmentation channel;

[0037] Step S240: inputting the region image into the bird image segmentation channel to perform image segmentation to obtain the multiple bird images.

[0038] Specifically, historical images of the bird-repelling area are mined. For example, historical images of the bird-repelling area within a predetermined time interval are retrieved, where the historical images have environmental and temporal differences, to obtain multiple images of the historical area. The bird images within the multiple historical area images are identified as targets, and the identification targets and backgrounds are segmented. The areas to which the segmented bird images belong are marked as the segmentation results of the multiple sample bird images. Furthermore, an encoder and decoder are constructed based on semantic segmentation, i.e., the main network architecture includes multiple convolutional layers and multiple pooling layers. The encoder and decoder have the same hierarchical structure and are relatively distributed.

[0039] Furthermore, the multiple historical region images are mapped to the multiple sample bird image segmentation results, and used as training samples. The mapped historical region images are used as input images for the encoder, and the sample bird image segmentation results are used as output images for the decoder, for supervised training to generate the bird image segmentation channel. The bird segmentation channel is further monitored based on the training samples, and deviations between the output image and the multiple sample bird image segmentation results are corrected to determine whether the convergence condition, such as output accuracy, is met. If not, training samples that do not meet the convergence condition are extracted for repeated training and verification until the preset convergence condition is met, thereby obtaining the constructed bird image segmentation channel.

[0040] The region images are then fed into the bird image segmentation pipeline. The encoder uses convolutional and pooling layers to reduce the size of the feature maps of the region images, transforming them into lower-dimensional representations. The decoder receives this representation and performs upsampling via transposed convolutions to restore the spatial dimensions. Each transposed convolution is then expanded to the feature map size to complete image segmentation and obtain the multiple bird images, enabling accurate and efficient segmentation and localization of the bird image regions.

[0041] Step S300: using a plurality of historical bird images in a bird repellent database, performing similarity analysis on the plurality of bird images, determining whether a plurality of birds in the plurality of bird images are historical birds, and obtaining a plurality of new birds and a plurality of historical birds, wherein the bird repellent database includes a plurality of sets of bird marking information, bird images, and bird repellent laser parameters;

[0042] Furthermore, if Figure 2 As shown, a plurality of historical bird images in the bird-repelling database are used to perform similarity analysis on the plurality of bird images. Step S300 of the present application further includes:

[0043] Step S310: constructing a bird similarity analysis channel based on the Siamese network, wherein the bird similarity analysis channel includes a first convolution path, a second convolution path, and a similarity calculation branch;

[0044] Step S320: combining the multiple historical bird images with the multiple bird images in sequence, inputting the images into the bird similarity analysis channel, performing image convolution feature extraction and similarity calculation, and obtaining multiple similarity information sets;

[0045] Step S330: determine whether the similarity information in the multiple similarity information sets is greater than a similarity threshold; if so, determine it as a historical bird; if not, determine it as a new bird, and obtain the multiple new birds and multiple historical birds.

[0046] Furthermore, based on the twin network, a bird similarity analysis channel is constructed. Step S310 of this application also includes:

[0047] Step S311: constructing the first convolution path and the second convolution path with the same network structure based on the convolutional neural network, wherein the first convolution path and the second convolution path include multiple convolution layers and multiple pooling layers;

[0048] Step S312: Based on the fully connected layer, construct the similarity calculation branch including the similarity calculation rule, connect the first convolution path and the second convolution path, the similarity calculation rule includes obtaining all image features extracted by convolution in the first convolution path and the second convolution path, and calculating the proportion of the same image convolution features to obtain similarity.

[0049] Specifically, the bird repellent database is a storage database built based on historical bird repellent data. It includes bird tagging information, such as bird numbers and images, and laser parameters used to repel the bird, including color, wavelength, and power. Multiple sets of bird tagging information, bird images, and laser parameters are integrated and combined to form the bird repellent database. Based on the bird repellent database, multiple historical bird images are retrieved and similarity analysis is performed between the historical bird images and the multiple bird images to determine whether the birds in the multiple bird images are historical birds (i.e., birds that have been repelled in the past). If they are historical birds, this indicates a certain degree of overall light resistance, requiring detailed analysis and parameter optimization.

[0050] Specifically, based on the convolutional neural network, the first convolution path and the second convolution path with the same network structure are constructed, including multi-layer convolution layers and multi-layer pooling layers with the same number of levels and distribution, for extracting convolution features of the input image. The similarity rule is obtained, that is, the convolution image features of the first convolution path and the second convolution path received by the flow are mapped and proofread, the same convolution features are extracted, the number of the same convolution features and the total number of convolution features are counted and the ratio is calculated as the image similarity. The similarity calculation branch is constructed based on the fully connected layer, and the similarity rule is embedded in the similarity calculation branch. The first convolution path and the second convolution path are arranged in parallel, and are placed in the similarity calculation branch to generate the bird similarity analysis channel. Modeling is performed based on the twin network, and the fitness is analyzed by directly comparing the convolution calculations, which has a relatively high processing efficiency and occupies less computing resources.

[0051] Furthermore, the multiple historical bird images are sequentially combined with the multiple bird images to determine multiple groups of bird images. The multiple groups of bird images are traversed and respectively input into the first convolution path and the second convolution path within the bird similarity analysis channel to extract image convolution features and transfer them to the similarity calculation branch. The similarity of each group of images is then evaluated in accordance with the similarity calculation rule to obtain the multiple similarity information sets.

[0052] A similarity threshold is further set, i.e., a custom-defined critical similarity based on the similarity determination criteria. The similarity information in the multiple similarity information sets is traversed and compared with the similarity threshold. If the similarity information is greater than the similarity threshold, the bird image is determined to be consistent with the historical bird image and is used as the historical bird. If the similarity information is less than or equal to the similarity threshold, i.e., the similarity is low, the bird is determined to be a new bird. The similarity comparison results are then aggregated to obtain the multiple new birds and the multiple historical birds.

[0053] Among them, for the multiple new birds, it is the first time to drive away the birds, and the issue of light resistance does not need to be considered; for the multiple historical birds, subsequent analysis of the overall light resistance is required to find the optimal laser parameters to obtain the laser parameters with the lowest laser adaptability, the greatest bird-repelling stimulation, and the best bird-repelling effect for the current multiple birds as a whole.

[0054] Step S400: marking the multiple new birds, and adding the bird images of the multiple new birds to the bird repellent database; searching the bird repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repelling of the multiple historical birds within a historical period;

[0055] Specifically, based on the marking pattern in the bird repellent database, the multiple new birds are marked, for example, by sequential numbering, and combined with the bird images of the multiple new birds, they are added to the bird repellent database to update the bird repellent database. Simultaneously, the marking information of the multiple historical birds is matched and determined, used as a search target, and the bird repellent database is searched. The historical laser parameters of the laser bird repellent are retrieved within the historical time period, i.e., the reference time interval for the light resistance impact analysis. The retrieved historical laser parameters are mapped and associated with the multiple historical birds to generate the multiple historical laser parameter sets. The multiple historical laser parameter sets correspond one-to-one to the multiple historical birds and serve as the parameter data source for light resistance analysis and laser parameter tuning.

[0056] Step S500: Analyzing the light resistance of the multiple historical birds to multiple laser parameters based on the multiple historical laser parameter sets, obtaining multiple light resistance parameter sets, and constructing a bird-repelling light resistance evaluation function;

[0057] Furthermore, the light resistance of the plurality of historical birds to a plurality of laser parameters is analyzed according to the plurality of historical laser parameter sets to obtain a plurality of light resistance parameter sets. Step S500 of the present application further includes:

[0058] Step S510: obtaining, based on the multiple historical laser parameter sets, the multiple historical bird-repelling times of the multiple historical laser parameters, to obtain multiple historical bird-repelling times sets;

[0059] Step S520: searching the bird-repelling record data in the historical period to obtain a set of sample bird-repelling times and a set of sample light resistance levels;

[0060] Step S530: constructing a light resistance analysis channel based on a decision tree according to the sample bird-repelling times set and the sample light resistance level set;

[0061] Step S540: inputting the multiple historical bird-repelling times in the multiple historical bird-repelling times sets into the photostability analysis channel to obtain the multiple photostability parameter sets.

[0062] Furthermore, the present application further includes step S550, comprising:

[0063] Step S551: Construct a bird-repelling light-resistance evaluation function, as follows:

[0064] ;

[0065] in, is the photostability score of the j-th laser parameter for the multiple historical birds, is the number of historical birds, is the light resistance level of the i-th historical bird to the j-th laser parameter, is the weight assigned according to the photostability level of the i-th historical bird to the j-th laser parameter.

[0066] Specifically, the historical bird-repelling frequency of laser parameters is used as a basis for light resistance evaluation based on the multiple historical laser parameter sets. The bird-repelling frequency statistics for the multiple historical laser parameters received by the multiple birds are respectively calculated, and the statistical results of the bird-repelling frequency for the same bird under different historical laser parameters are determined as a historical bird-repelling frequency set. The multiple historical laser parameters of the multiple historical birds are counted to obtain the multiple historical bird-repelling frequency sets. The more times a bird is repelled by a laser parameter, the higher the light resistance level for that laser parameter.

[0067] Furthermore, historical bird-repelling record data is retrieved to identify and extract the sample bird-repelling frequency and sample light resistance level sets. These two sets of sample data can be directly extracted from the bird-repelling record data and are in one-to-one correspondence. Furthermore, based on the sample bird-repelling frequency set and the sample light resistance level set, a light resistance analysis channel is constructed using a decision tree to accurately and efficiently analyze and decide on light resistance based on the bird-repelling frequency.

[0068] Specifically, based on the set of sample bird-repelling frequency, an item is randomly extracted as the first decision node of the first decision layer, and the set of sample bird-repelling frequency is binary-classified based on the first decision node. Again, based on the set of sample bird-repelling frequency, an item is randomly extracted as the second decision node of the second decision layer, and the binary classification result is further divided. The above steps are repeated until the convergence condition is met, for example, the maximum construction level is reached, and the construction of the Nth decision layer and the determination of the Nth decision node are completed. The first decision layer, the second decision layer, and the Nth decision layer are hierarchically associated to generate a classification tree based on the number of bird-repelling frequency. The set of sample light resistance levels is further used as decision data to match and identify the classification tree, complete the construction of the decision tree, and generate the light resistance analysis channel based on the decision tree.

[0069] Furthermore, the multiple historical bird-repelling times in the multiple historical bird-repelling times set are input into the light-resistance analysis channel, and the decision tree built into the channel is traversed to perform hierarchical division and attribution of each bird-repelling time, and the identification data of the attribution result is used as the corresponding light-resistance parameter, and the multiple historical bird-repelling times are associated with the corresponding light-resistance parameters, and the light-resistance parameter set is generated through integration.

[0070] Furthermore, the bird-repelling light resistance evaluation function is constructed: ,in, is the photostability score of the j-th laser parameter for the multiple historical birds, is the number of historical birds, is the light resistance level of the i-th historical bird to the j-th laser parameter, The weight is assigned according to the size of the light resistance level of the i-th historical bird to the j-th laser parameter. The above parameters can all be determined based on the early processing of the embodiment of the present application. The weight configuration is proportional to the light resistance level of each historical bird to the laser parameter. The sum of the assigned weights is 1. The bird-repellent light resistance evaluation function is used to calculate the light resistance score.

[0071] Step S600: According to the bird-repelling light resistance evaluation function and the multiple light resistance parameter sets, the laser parameters currently used for bird-repelling are adjusted and optimized to obtain the optimal laser parameters, laser bird-repelling is performed on the designated area, and the optimal laser parameters are added to the bird-repelling database.

[0072] Furthermore, if Figure 3 As shown, according to the bird-repelling light resistance evaluation function and the multiple light resistance parameter sets, the laser parameters currently used for bird-repelling are adjusted and optimized to obtain the optimal laser parameters. Step S600 of this application also includes:

[0073] Step S610: randomly generating a laser parameter as an initial laser parameter, wherein the initial laser parameter includes an initial laser color, an initial laser wavelength, and an initial laser power;

[0074] Step S620: Calculating an initial light resistance score based on the multiple light resistance parameter sets and the bird-repelling light resistance evaluation function;

[0075] Step S630: adjusting the initial laser parameters according to a preset optimization step size to obtain first laser parameters, and calculating a first light resistance score based on the multiple light resistance parameter sets and the bird repellent light resistance evaluation function;

[0076] Step S640: Calculating and adjusting the preset optimization step length according to the first light resistance score and the initial light resistance score to obtain a first optimization step length;

[0077] Step S650: adjusting the first laser parameters according to the first optimization step size and performing iterative optimization;

[0078] Step S660: continue iterative optimization until the optimization times threshold is reached, output the laser parameter with the largest light resistance score during the optimization process, and obtain the optimal laser parameter.

[0079] Furthermore, the preset optimization step length is calculated and adjusted according to the first light fastness score and the initial light fastness score to obtain a first optimization step length. Step S640 of the present application further includes:

[0080] Step S641: calculating the ratio of the first light resistance score to the initial light resistance score;

[0081] Step S642: using the inverse of the ratio to adjust and calculate the preset optimization step size to obtain the first optimization step size.

[0082] Specifically, the initial laser color, initial laser wavelength, and initial laser power are randomly generated as the initial laser parameters, and parameter optimization analysis is performed based on the initial laser parameters. Based on the multiple light resistance parameter sets, the number of bird repelling events and light resistance levels of the initial laser parameters for the multiple historical bird species are retrieved and determined, and the results are input into the bird repelling light resistance evaluation function to calculate the initial light resistance score.

[0083] Furthermore, a preset optimization step size is set, i.e., a custom-defined initialization step size interval for performing a single adjustment on each of the laser parameters, for example, a laser power adjustment step size of 1W. Based on the preset optimization step size, the initial laser parameters are adjusted to serve as the first laser parameters. The plurality of light resistance parameter sets are further traversed to determine the number of bird repelling events and light resistance levels associated with the first laser parameter for the plurality of historical bird species. These are input into the bird repelling light resistance evaluation function to calculate and obtain the first light resistance score.

[0084] Furthermore, the preset optimization step size is adjusted based on the first light resistance score and the initial light resistance score, i.e., a positive feedback adjustment is performed. Specifically, a ratio of the first light resistance score to the initial light resistance score is calculated, and the inverse of the ratio is used as the adjustment criterion. The inverse of the ratio is multiplied by the preset optimization step size, and the result is used as the first optimization step size. The optimization step size is adjusted synchronously with the recursive optimization iterations to ensure that the optimization step size is compatible with the real-time optimization process.

[0085] Based on the first optimization step size, the first laser parameter is then adjusted to serve as the second laser parameter, and a second light resistance score is calculated. The above steps are repeated for iterative optimization until the optimization threshold (i.e., the maximum number of optimization iterations) is met. Laser parameter optimization is then stopped, and the laser parameter with the highest light resistance score during the optimization process is selected as the optimal laser parameter, i.e., the laser parameter with the lowest overall laser light resistance rating for multiple historical birds, to maximize the bird repellent effect on the multiple historical birds. Furthermore, based on the optimal laser parameter, laser bird repellent is performed on the designated area, and the optimal laser bird repellent parameter is added to the bird repellent database to update the database, ensure the database's timeliness, and improve the accuracy of subsequent analysis in the laser parameter optimization process.

[0086] The bird-repelling method based on radar laser technology provided in the embodiment of the present application has the following technical effects:

[0087] 1. Collect regional images and perform image segmentation to identify only target bird images. Combined with the Siamese network, a similarity analysis channel is constructed to analyze the similarity between historical bird images and collected bird images. This allows for efficient and accurate classification of historical and new bird species, facilitating subsequent targeted processing of historical birds.

[0088] 2. Combined with historical laser parameters, the light resistance of historical birds to various laser parameters is analyzed. Combined with the constructed bird-repellent light resistance evaluation function, the laser parameters are optimized to determine the laser parameters with the lowest overall laser adaptability, the greatest bird-repellent irritation, and the best bird-repellent effect. Example 2

[0089] Based on the same inventive concept as the bird-repelling method based on radar laser technology in the above embodiment, Figure 4 As shown, the present application provides a bird-repelling system based on radar laser technology, the system comprising:

[0090] An image acquisition module 11 is used to acquire images of a designated area to be bird-repellent, and obtain a regional image, wherein the regional image includes images of multiple birds currently in the designated area;

[0091] An image segmentation module 12 is configured to segment the bird images of the plurality of birds in the regional image to obtain a plurality of bird images;

[0092] A similarity analysis module 13 is configured to use a plurality of historical bird images in a bird-repelling database to perform similarity analysis on the plurality of bird images, determine whether a plurality of birds in the plurality of bird images are historical birds, and obtain a plurality of new birds and a plurality of historical birds, wherein the bird-repelling database includes a plurality of sets of bird tag information, bird images, and bird-repelling laser parameters;

[0093] The laser parameter acquisition module 14 is used to mark the multiple new birds, add the bird images of the multiple new birds to the bird repellent database, and search the bird repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repellent of the multiple historical birds in the historical time;

[0094] A function construction module 15 is configured to analyze the light resistance of the plurality of historical birds to a plurality of laser parameters based on the plurality of historical laser parameter sets, obtain a plurality of light resistance parameter sets, and construct a bird-repelling light resistance evaluation function;

[0095] The parameter tuning control module 16 is used to adjust and optimize the laser parameters currently used for bird repelling according to the bird repelling light resistance evaluation function and the multiple light resistance parameter sets, obtain the optimal laser parameters, perform laser bird repelling on the designated area, and add the optimal laser parameters to the bird repelling database.

[0096] Furthermore, the image segmentation module 12 further includes:

[0097] A sample acquisition module is used to mine historical area images collected from multiple bird-repelling areas, obtain multiple historical area images, and segment and mark bird images in the multiple historical area images to obtain multiple sample bird image segmentation results;

[0098] A building module, wherein the building module is used to build an encoder and a decoder based on semantic segmentation;

[0099] a channel acquisition module, configured to perform supervised training on the encoder and decoder using the plurality of historical area images and the plurality of sample bird image segmentation results until convergence conditions are met, thereby obtaining bird image segmentation channels;

[0100] The bird image acquisition module is used to input the regional image into the bird image segmentation channel, perform image segmentation, and obtain the multiple bird images.

[0101] Furthermore, the similarity analysis module 13 further includes:

[0102] A similarity analysis channel construction module, wherein the similarity analysis channel construction module is used to construct a bird similarity analysis channel based on the Siamese network, wherein the bird similarity analysis channel includes a first convolution path, a second convolution path, and a similarity calculation branch;

[0103] a similarity information acquisition module, configured to sequentially combine the plurality of historical bird images with the plurality of bird images, input the images into the bird similarity analysis channel, perform image convolution feature extraction and similarity calculation, and obtain a plurality of similarity information sets;

[0104] A threshold judgment module is used to judge whether the similarity information in the multiple similarity information sets is greater than a similarity threshold. If so, it is judged to be a historical bird; if not, it is judged to be a new bird, and the multiple new birds and multiple historical birds are obtained.

[0105] Furthermore, the similarity analysis channel construction module also includes:

[0106] A convolutional path construction module, wherein the convolutional path construction module is used to construct the first convolutional path and the second convolutional path with the same network structure based on a convolutional neural network, wherein the first convolutional path and the second convolutional path include multiple convolutional layers and multiple pooling layers;

[0107] A branch construction module is used to construct the similarity calculation branch including the similarity calculation rule based on the fully connected layer, connect the first convolution path and the second convolution path, and the similarity calculation rule includes obtaining all image features extracted by convolution in the first convolution path and the second convolution path, and calculating the proportion of the same image convolution features to obtain similarity.

[0108] Furthermore, the function building module 15 also includes:

[0109] A bird-repelling frequency acquisition module is used to obtain the number of times the multiple historical birds are repelled by the multiple historical laser parameters according to the multiple historical laser parameter sets, and obtain multiple historical bird-repelling frequency sets;

[0110] A sample data acquisition module, the sample data acquisition module is used to search the bird-repelling record data in the historical time period to obtain a sample bird-repelling frequency set and a sample light resistance level set;

[0111] A light resistance analysis channel construction module, wherein the light resistance analysis channel construction module is used to construct a light resistance analysis channel based on a decision tree according to the sample bird-repelling frequency set and the sample light resistance level set;

[0112] A light resistance parameter acquisition module is used to input multiple historical bird-repelling times in the multiple historical bird-repelling times sets into the light resistance analysis channel to obtain the multiple light resistance parameter sets.

[0113] Furthermore, the function building module 15 also includes:

[0114] An evaluation function construction module is used to construct a bird-repellent light-resistance evaluation function, as shown in the following formula:

[0115] ;

[0116] in, is the photostability score of the j-th laser parameter for the multiple historical birds, is the number of historical birds, is the light resistance level of the i-th historical bird to the j-th laser parameter, is the weight assigned according to the photostability level of the i-th historical bird to the j-th laser parameter.

[0117] Furthermore, the parameter tuning control module 16 further includes:

[0118] An initial laser parameter determination module, the initial laser parameter determination module is used to randomly generate a laser parameter as an initial laser parameter, wherein the initial laser parameter includes an initial laser color, an initial laser wavelength, and an initial laser power;

[0119] a light resistance score calculation module, configured to calculate an initial light resistance score based on the plurality of light resistance parameter sets and the bird-repelling light resistance evaluation function;

[0120] a first light resistance score calculation module, configured to adjust the initial laser parameters according to a preset optimization step size to obtain first laser parameters, and calculate a first light resistance score based on the multiple light resistance parameter sets and the bird repellent light resistance evaluation function;

[0121] a first optimization step length acquisition module, configured to calculate and adjust the preset optimization step length according to the first light resistance score and the initial light resistance score to obtain a first optimization step length;

[0122] an iterative optimization module, configured to adjust the first laser parameter according to the first optimization step size and perform iterative optimization;

[0123] The optimal laser parameter acquisition module is used to continue iterative optimization until a threshold number of optimization times is reached, output the laser parameter with the largest light resistance score during the optimization process, and obtain the optimal laser parameter.

[0124] Furthermore, the first optimization step length acquisition module further includes:

[0125] a score ratio calculation module, configured to calculate a ratio of the first light resistance score to an initial light resistance score;

[0126] A step length adjustment module is used to use the inverse of the ratio to adjust the preset optimization step length to obtain the first optimization step length.

[0127] Through the above detailed description of a bird-repelling method based on radar laser technology in this specification, those skilled in the art can clearly understand a bird-repelling method and system based on radar laser technology in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0128] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A bird-repelling method based on radar laser technology, characterized in that: The method comprises: Collecting an image of a designated area to be bird-repellent to obtain a regional image, wherein the regional image includes images of a plurality of birds currently in the designated area; performing image segmentation on the bird images of the plurality of birds within the region image to obtain a plurality of bird images; Using multiple historical bird images in a bird-repelling database, performing similarity analysis on the multiple bird images, determining whether multiple birds in the multiple bird images are historical birds, and obtaining multiple new birds and multiple historical birds, the bird-repelling database including multiple sets of bird marking information, bird images, and bird-repelling laser parameters; Marking the multiple new birds, and adding the bird images of the multiple new birds to the bird repellent database, searching the bird repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repelling of the multiple historical birds within a historical time period; Analyzing the light resistance of the historical birds to the multiple laser parameters according to the multiple historical laser parameter sets, obtaining multiple light resistance parameter sets, and constructing a bird repellent light resistance evaluation function; According to the bird-repelling light resistance evaluation function and the multiple light resistance parameter sets, the laser parameters currently used for bird repelling are adjusted and optimized to obtain the optimal laser parameters, laser bird repelling is performed on the designated area, and the optimal laser parameters are added to the bird repelling database.

2. The method according to claim 1, characterized in that Performing image segmentation on the bird images of the plurality of birds within the region image to obtain the plurality of bird images comprises: Mining historical area images collected from multiple bird-repelling areas to obtain multiple historical area images, and segmenting and marking bird images in the multiple historical area images to obtain multiple sample bird image segmentation results; Based on semantic segmentation, build encoder and decoder; Using the multiple historical area images and the multiple sample bird image segmentation results, supervised training is performed on the encoder and decoder until a convergence condition is met, thereby obtaining a bird image segmentation channel; The region image is input into the bird image segmentation channel to perform image segmentation to obtain the multiple bird images.

3. The method according to claim 1, characterized in that Using multiple historical bird images in a bird-repelling database, similarity analysis is performed on the multiple bird images, including: Based on the Siamese network, a bird similarity analysis channel is constructed, wherein the bird similarity analysis channel includes a first convolution path, a second convolution path and a similarity calculation branch; Combining the multiple historical bird images with the multiple bird images in sequence, inputting the images into the bird similarity analysis channel, performing image convolution feature extraction and similarity calculation, and obtaining multiple similarity information sets; It is determined whether the similarity information in the plurality of similarity information sets is greater than a similarity threshold; if so, it is determined to be a historical bird; if not, it is determined to be a new bird, and the plurality of new birds and the plurality of historical birds are obtained.

4. The method according to claim 3, characterized in that Based on the twin network, a bird similarity analysis channel is constructed, including: Based on a convolutional neural network, constructing the first convolution path and the second convolution path with the same network structure, wherein the first convolution path and the second convolution path include multiple convolution layers and multiple pooling layers; Based on the fully connected layer, a similarity calculation branch including a similarity calculation rule is constructed to connect the first convolution path and the second convolution path. The similarity calculation rule includes obtaining all image features extracted by convolution in the first convolution path and the second convolution path, and calculating the proportion of the same image convolution features to obtain similarity.

5. The method according to claim 1, characterized in that Analyzing the photostability of the plurality of historical birds to a plurality of laser parameters according to the plurality of historical laser parameter sets to obtain a plurality of photostability parameter sets, including: According to the multiple historical laser parameter sets, obtaining the multiple historical bird-repelling times of the multiple historical laser parameters, and obtaining multiple historical bird-repelling times sets; Search the bird-repelling record data in the historical period to obtain the sample bird-repelling times set and the sample light resistance level set; According to the sample bird-repelling times set and the sample light-tolerance level set, a light-tolerance analysis channel is constructed based on a decision tree; Multiple historical bird-repelling times in the multiple historical bird-repelling times sets are input into the photostability analysis channel to obtain the multiple photostability parameter sets.

6. The method according to claim 5, characterized in that Construct the bird-repellent light-resistance evaluation function as follows: ； in, is the photostability score of the j-th laser parameter for the multiple historical birds, is the number of historical birds, is the light resistance level of the i-th historical bird to the j-th laser parameter, is the weight assigned according to the photostability level of the i-th historical bird to the j-th laser parameter.

7. The method according to claim 1, characterized in that According to the bird-repelling light resistance evaluation function and the plurality of light resistance parameter sets, the laser parameters currently used for bird-repelling are adjusted and optimized to obtain the optimal laser parameters, including: Randomly generating a laser parameter as an initial laser parameter, wherein the initial laser parameter includes an initial laser color, an initial laser wavelength, and an initial laser power; Calculating an initial light resistance score according to the multiple light resistance parameter sets and the bird repellent light resistance evaluation function; Adjusting the initial laser parameters according to a preset optimization step size to obtain first laser parameters, and calculating a first light resistance score based on the multiple light resistance parameter sets and the bird repellent light resistance evaluation function; calculating and adjusting the preset optimization step length according to the first light resistance score and the initial light resistance score to obtain a first optimization step length; Adjusting the first laser parameter according to the first optimization step size and performing iterative optimization; The iterative optimization is continued until a threshold of optimization times is reached, and the laser parameters with the largest photostability score during the optimization process are output to obtain the optimal laser parameters.

8. The method according to claim 7, characterized in that Calculating and adjusting the preset optimization step length according to the first light resistance score and the initial light resistance score to obtain a first optimization step length, comprising: calculating a ratio of the first light fastness score to the initial light fastness score; The preset optimization step size is adjusted and calculated using the inverse of the ratio to obtain the first optimization step size.

9. A bird-repelling system based on radar laser technology, characterized in that: The system comprises: An image acquisition module is used to acquire images of a designated area to be bird-repellent, and obtain a regional image, wherein the regional image includes images of multiple birds currently in the designated area; an image segmentation module, the image segmentation module being used to perform image segmentation on the bird images of the plurality of birds in the regional image to obtain a plurality of bird images; a similarity analysis module, the similarity analysis module being configured to use a plurality of historical bird images in a bird repellent database to perform similarity analysis on the plurality of bird images, determine whether a plurality of birds in the plurality of bird images are historical birds, and obtain a plurality of new birds and a plurality of historical birds, the bird repellent database including a plurality of sets of bird marking information, bird images, and bird repellent laser parameters; a laser parameter acquisition module, the laser parameter acquisition module being used to mark the multiple new birds, add the bird images of the multiple new birds to the bird repellent database, and search the bird repellent database based on the marking information of the multiple historical birds to obtain multiple historical laser parameter sets for laser bird repelling of the multiple historical birds within a historical period; A function construction module, the function construction module being used to analyze the light resistance of the plurality of historical birds to a plurality of laser parameters based on the plurality of historical laser parameter sets, obtain a plurality of light resistance parameter sets, and construct a bird-repelling light resistance evaluation function; A parameter tuning control module is used to adjust and optimize the laser parameters currently used for bird repelling according to the bird repelling light resistance evaluation function and the multiple light resistance parameter sets, obtain optimal laser parameters, perform laser bird repelling on the designated area, and add the optimal laser parameters to the bird repelling database.

Images