Intelligent identification multifunctional bird repelling device and bird repelling method

Abstract

The invention discloses intelligent identification multifunctional bird repelling equipment and a bird repelling method. The bird repelling equipment comprises a monitoring camera, an intelligent controller, a laser transmitter group, a power supply box and a base. Bird repelling work in a monitoring area can be completed through cooperation of the patrol laser transmitter and the support laser transmitter, in the working process, the monitoring staying time of different monitoring areas is adjusted according to the bird density in the monitoring areas, meanwhile, the patrol working time of the patrol laser transmitter in the different monitoring areas is adjusted, and the working efficiency of the patrol laser transmitter is improved. And when more birds appear, the support laser transmitter can be started to enhance the bird repelling strength, so that the bird repelling working efficiency is high, and the problem of safe operation of a power grid caused by bird activities can be effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment protection, and in particular to an intelligent identification multifunctional bird-repelling device and a bird-repelling method. Background Art

[0002] As the main support for power transmission, the safety of transmission lines is crucial to the entire power system. In recent years, bird activities have brought great hidden dangers to the safe operation of transmission lines. There are many bird prevention technologies in relevant domestic and foreign materials. Most of them adopt single or new bird prevention equipment for comprehensive protection, and adopt single methods such as bird spikes and needle plates, which cannot meet the protection requirements and will cause varying degrees of harm to birds. For example, the protection range of bird spikes is small, and birds are very easy to adapt to equipment such as bird windmills. Even if they are installed in a "covered" manner on poles and towers, they only provide short-distance protection for poles and towers, and cannot achieve long-distance detection and protection. As a result, after birds have entered the protection area, the conventional single protection equipment cannot cause certain interference to the flight and stay of birds, thus losing the meaning of protection. In addition, some shortcomings have gradually been exposed after long-term use, and the meaning of protection has been lost.

[0003] After nearly a decade of development, bird repellent methods at home and abroad have been improved to varying degrees, but most of them are passive physical bird repellent devices such as bird thorns, bird needle plates, bird umbrellas, bird baffles, etc. This method is easy to operate and has low cost. However, the coverage area is small. At the same time, the equipment is easily damaged when exposed outdoors. This repellent method can definitely achieve significant results in the short term. However, if it continues for a long time, birds will inevitably adapt. The effect of bird repellent is getting worse year by year, and bird damage accidents continue to occur.

[0004] Active physical bird-proofing devices actively detect bird activity characteristics through infrared, radar and other means, and stimulate birds through voice, light, ultrasound, laser and other means to make them uncomfortable, thereby driving them away. Birds have keen hearing and vision, and electronic bird repellers use this physiological characteristic of birds to achieve the purpose of bird repelling and bird prevention. By playing ultrasound and sounds of bird's natural enemies, gunshots and other sounds to interfere with birds' hearing, birds are driven away from the protection area. It has achieved good results in the early application, and the red and blue flashing lights are used to interfere with birds' vision at night to achieve the effect of driving away.

[0005] Reference 1: Chinese patent document with publication number CN107242231A

[0006] Reference 1 records an intelligent laser bird-repelling device for a transmission tower, including an intelligent control system, which is respectively connected to a laser emitting device, an infrared sensing system and a power supply system, wherein the intelligent control system receives the sensing signal collected by the infrared sensing system, optimizes the bird-repelling algorithm according to the sensing signal, and controls the laser emitting device to emit a bird-repelling laser according to the bird-repelling algorithm, wherein the laser emitting device and the infrared sensing system are both connected to the power supply system, and the intelligent control system, the laser emitting device, the infrared sensing system and the power supply device are all fixed on the mounting base, and the mounting base is fixed on the transmission tower. The present invention is installed on a transmission tower, which can effectively prevent birds from nesting on the tower or moving around the tower, and reduces problems such as line short circuit, ground fault and insulator flashover caused by bird activities, which is conducive to ensuring the safe operation of the transmission line.

[0007] Reference 2: Chinese patent document with publication number CN 110692619 A

[0008] Reference 2 records an edge computing-based video surveillance-laser bird repellent integrated device and method, including a surveillance camera, an intelligent control unit, a laser emitting device, a power supply system and a base, wherein the intelligent control unit is respectively connected to the surveillance camera and the power supply system, and the surveillance camera, the intelligent control unit, the laser emitting device and the power supply system are all fixed on the base, and the base is fixed on a substation or a transmission tower; the intelligent control unit receives surveillance photos taken by the surveillance camera in real time, identifies bird activity behaviors, and performs deep learning based on bird activity behaviors to optimize the bird repellent algorithm, and controls the laser emitting device to emit bird repellent laser according to the bird repellent algorithm; the existing laser bird repellent device solves the problems of invalid movement, high energy consumption and misidentification, and can effectively reduce the problems of safe operation of the power grid caused by bird activities while safely monitoring the distribution network.

[0009] However, the methods for actively repelling birds are not limited to the above two. In order to enrich the technical route of active bird repelling, the applicant has designed an intelligent identification multifunctional bird repelling device and a bird repelling method that are different from the prior art. Summary of the invention

[0010] The purpose of the present invention is to enrich the technical route of active bird repelling and provide an intelligent identification multifunctional bird repelling device and a bird repelling method.

[0011] In order to solve the above technical problems, the technical solution adopted by the present invention is: an intelligent identification multifunctional bird-repelling device, including a monitoring camera, an intelligent controller, a laser transmitter group, a power supply box and a base;

[0012] The monitoring camera is a rotatable camera, and the monitoring range of the monitoring camera can cover the entire monitoring area;

[0013] The laser transmitter group includes a patrol laser transmitter and a support laser transmitter. The patrol laser transmitter is used to continuously transmit lasers in the monitoring area. The support laser transmitter receives instructions from the intelligent controller to transmit lasers into the designated monitoring area.

[0014] The intelligent controller receives the surveillance photos transmitted by all the sub-surveillance cameras in real time, calculates the bird density value of the corresponding surveillance area according to the surveillance photos, and chooses to start the patrol laser transmitter alone or start the patrol laser transmitter and the support laser transmitter at the same time according to the bird density value;

[0015] The power supply box is used to supply power to the surveillance camera group, the intelligent controller and the laser transmitter group;

[0016] The base includes a base body and a mounting rod vertically arranged on the base body, and the mounting rod is used to provide mounting support for the laser transmitter group, the power supply box and / or the monitoring camera group.

[0017] As a further optimization of the intelligent identification multifunctional bird-repelling device of the present invention: the patrol laser emitter is a semiconductor laser emitter, and the support laser emitter is a helium-neon laser emitter.

[0018] As a further optimization of the intelligent identification multifunctional bird-repelling device of the present invention: the patrol laser emitter and the support laser emitter are both arranged on the mounting pole, and the horizontal positions of the patrol laser emitter and the support laser emitter are consistent.

[0019] The present invention also provides a bird-repelling method:

[0020] The surveillance area of ​​the surveillance camera is divided into N sub-surveillance areas, and the surveillance camera performs cyclic monitoring on the N sub-surveillance areas;

[0021] The monitoring time of each sub-monitoring area in the initial round is set to t0. The intelligent controller receives the monitoring images transmitted by the monitoring camera and forms a monitoring image set; the monitoring image set is an array of (N, M, W, H, 3), where N is the number of sub-monitoring areas, M is the total number of images of each sub-monitoring area, and the image size is expressed as: W×H×3;

[0022] Perform image recognition on the initial round of surveillance image sets, calculate the bird density of each surveillance image, and obtain the surveillance image bird density dataset δ x ;

[0023] δx={δ 1-1 ,δ 1-2 ,δ 1-3 ......δ n-m}

[0024] Wherein, n represents the nth sub-monitoring area, and m represents the mth image of the nth sub-monitoring area;

[0025] According to the surveillance image bird density dataset δ x The bird density in each sub-monitoring area is calculated, and the bird density data set δ of the sub-monitoring area is obtained. Y ;

[0026] δ Y ={δ1,δ 2...... δ n}

[0027] Where n represents the nth sub-monitoring area, m represents the total number of images in the nth sub-monitoring area, and δ n The calculation formula is:

[0028]

[0029] Bird density dataset for the monitoring area δ Y , and assign a weight value Q to each sub-monitoring area according to the ranking, wherein the weight value Q of the last sub-monitoring area in the ranking is set to 1, the weight value Q of the second-to-last sub-monitoring area in the ranking is set to 2, and so on;

[0030] Calculate the weight coefficient K of each sub-monitoring area according to the weight value Q n , K n The calculation formula is as follows:

[0031]

[0032] Where n represents the nth sub-monitoring area;

[0033] The monitoring time of each sub-monitoring area in the next round is calculated according to the weight coefficient to obtain the monitoring time data set T y , where y represents the round of monitoring cycle;

[0034] Ty={t1,t2,t3......t n}

[0035] Where n represents the monitoring time of the nth sub-monitoring area in the yth monitoring cycle, t n The calculation formula is:

[0036] t n =K n ×(t o ×n)

[0037] The surveillance camera follows the monitoring time dataset T in the yth round of monitoring cycle. y Allocate monitoring time for each sub-monitoring area;

[0038] The patrol laser transmitter in the yth round of monitoring cycle follows the monitoring time data set T y Allocate working time for each sub-monitoring area and set support threshold δ k , when the bird density δ in the nth sub-monitoring area n Greater than the threshold δ k When the support laser transmitter is started, the support laser transmitter is made to scan the nth sub-monitoring area, and the duration of the scanning work is set to t k .

[0039] As a further optimization of the bird-repelling method of the present invention: the image of each sub-monitoring area is intercepted at an interval of 0.5-1S to obtain M monitoring images.

[0040] As a further optimization of the bird-repelling method of the present invention: the bird density calculation method of each monitoring image is as follows:

[0041] Perform noise reduction on the surveillance image to remove salt and pepper noise and Gaussian noise contained in the surveillance image;

[0042] Determine a judgment threshold, the pixel value of the bird part is higher than the threshold, and the background pixel value is lower than the threshold, then extract the bird target through threshold segmentation to obtain a binary image;

[0043] The binary image is obtained by threshold segmentation, and the bird density is calculated by counting the number of white pixels.

[0044] As a further optimization of the bird-repelling method of the present invention: before performing noise reduction processing, the color image is first converted into a grayscale image.

[0045] As a further optimization of the bird-repelling method of the present invention: the bird density calculation method of each monitoring image is as follows:

[0046] Build a YOLO model, input the surveillance image into the model, obtain the bounding box of the bird target, divide the surveillance image into several sub-areas, count the number of birds detected in each sub-area, and use the average or sum of the number of birds in each sub-area to represent the bird density.

[0047] As a further optimization of the bird-repelling method of the present invention: the scanning working time t of the supporting laser transmitter k It is 60-90S.

[0048] As a further optimization of a bird-repelling method of the present invention: when the bird density δ of a sub-monitoring area is n Two consecutive rounds are greater than the threshold δ k , increases the operating power of the supporting laser transmitter.

[0049] The present invention has the following beneficial effects: the present invention can complete the bird-repelling work in the monitoring area through the cooperation of the patrol laser transmitter and the support laser transmitter. During the working process, the monitoring residence time of different monitoring areas is adjusted according to the bird density in the monitoring area, and the patrol working time of the patrol laser transmitter in different monitoring areas is adjusted at the same time. In addition, when there are many birds, the support laser transmitter can be started to strengthen the bird-repelling force. The bird-repelling work is highly efficient and can effectively reduce the problems of safe operation of the power grid caused by bird activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a structural block diagram of the bird-repelling device of the present invention. DETAILED DESCRIPTION

[0051] In order to better understand the present invention, the content of the present invention is further explained below in conjunction with embodiments, but the content of the present invention is not limited to the following embodiments.

[0052] <Bird repellent equipment>

[0053] As shown in the figure: an intelligent identification multifunctional bird-repelling device, including a surveillance camera, an intelligent controller, a laser transmitter group, a power supply box and a base.

[0054] The surveillance camera is a rotatable camera, and the monitoring range of the surveillance camera can cover the entire monitoring area.

[0055] The laser transmitter group includes a patrol laser transmitter and a support laser transmitter. The patrol laser transmitter is used to continuously transmit lasers in the monitoring area, and the support laser transmitter receives instructions from the intelligent controller to transmit lasers into the designated monitoring area.

[0056] The intelligent controller receives the surveillance photos transmitted by all sub-surveillance cameras in real time, calculates the bird density value of the corresponding surveillance area based on the surveillance photos, and chooses to start the patrol laser transmitter alone or start the patrol laser transmitter and the support laser transmitter at the same time according to the bird density value.

[0057] The power supply box is used to supply power to the surveillance camera group, the intelligent controller and the laser transmitter group.

[0058] The base includes a base body and a mounting rod vertically arranged on the base body, and the mounting rod is used to provide mounting support for the laser transmitter group, the power supply box and / or the monitoring camera group.

[0059] The patrol laser emitter is a semiconductor laser emitter, and the support laser emitter is a helium-neon laser emitter.

[0060] Semiconductor laser emitters work based on the principle of electron transition in semiconductor materials. When an appropriate forward bias voltage is applied to the semiconductor material, the electrons in the semiconductor will transition from the valence band to the conduction band, forming a population inversion distribution state. In this state, the electrons in the conduction band will spontaneously transition back to the valence band and release energy. When certain resonance conditions are met, this energy will be emitted in the form of laser.

[0061] The interior of the HeNe laser transmitter is filled with a mixture of helium and neon, usually with a higher proportion of helium. When a high voltage is applied to both ends of the laser tube, the gas discharges. In this process, electrons are accelerated by the electric field, and these high-speed electrons collide with helium atoms, causing the helium atoms to transition from the ground state to the excited state.

[0062] The excited state of the helium atom has two metastable states, and the energies of these two metastable states are very close to the energies of the two excited states of the neon atom. Through the collision between helium atoms and neon atoms, the helium atom transfers energy to the neon atom, causing the neon atom to be excited to a specific high energy level, thereby achieving a population inversion between the specific energy levels of the neon atom.

[0063] When neon atoms transition from high energy levels back to low energy levels, photons are emitted. In the resonant cavity formed by the reflectors at both ends of the laser tube, photons are reflected back and forth, constantly stimulating neon atoms at high energy levels to emit photons, thereby amplifying the light. When the light gain is greater than the loss, a stable laser will be output from the partially transmitted reflector end.

[0064] The patrol laser emitter and the support laser emitter are both arranged on the mounting rod, and the horizontal positions of the patrol laser emitter and the support laser emitter are consistent.

[0065] <Bird-repelling methods>

[0066] An intelligent identification and bird-repelling method:

[0067] The bird-repelling method is specifically applied in a substation, and the bird-repelling equipment is arranged at the center of the substation.

[0068] The monitoring area of ​​the surveillance camera is divided into 4 sub-monitoring areas, and the surveillance camera performs cyclic monitoring on the 4 sub-monitoring areas.

[0069] During operation, the surveillance camera carries out cyclic monitoring operations in an orderly manner for the four sub-monitoring areas according to the preset intelligent control program. In each cycle, the camera first positions its image acquisition focus on the first sub-monitoring area, and uses its high-resolution imaging sensor to continuously collect data and capture images of the area at a specific frame rate. The acquired image information is immediately transmitted to the background image processing unit for preliminary feature extraction and analysis. After completing the acquisition of the predetermined monitoring time or data volume of the first sub-monitoring area, the camera smoothly transitions the monitoring direction to the second sub-monitoring area within a very short switching delay, repeating the above image acquisition, transmission and analysis process. During the switching process, the monitoring parameters can be adaptively adjusted according to factors such as the ambient light of different sub-monitoring areas and the expected target movement speed, so as to ensure the stability and consistency of the monitoring quality.

[0070] Afterwards, the camera monitors the third sub-monitoring area in the predetermined order. Finally, the camera monitors and scans the fourth sub-monitoring area. After completing the monitoring task of the area, a complete cycle of monitoring ends and the next cycle starts automatically. This cycle repeats itself, forming a continuous, stable and efficient cycle monitoring mechanism.

[0071] The monitoring time of each sub-monitoring area in the initial round is set to t0 (60S). The intelligent controller receives the monitoring images transmitted by the monitoring camera and forms a monitoring image set; the monitoring image set is an (N, M, W, H, 3) array, where N is the number of sub-monitoring areas, M is the total number of images of each sub-monitoring area, and the image size is expressed as: W×H×3.

[0072] Perform image recognition on the initial round of surveillance image sets, calculate the bird density of each surveillance image, and obtain the surveillance image bird density dataset δ x .

[0073] δx={δ 1-1 ,δ 1-2 ,δ 1-3 ......δ n-m}

[0074] Among them, n represents the nth sub-monitoring area, here, the value range of n is 1-4, and m represents the mth image of the nth sub-monitoring area.

[0075] The image of each sub-monitoring area is captured at an interval of 1S to obtain 60 monitoring images. That is to say, each sub-monitoring area has 60 monitoring images, and the four sub-monitoring areas have a total of 240 monitoring images.

[0076] According to the surveillance image bird density dataset δ xThe bird density in each sub-monitoring area is calculated, and the bird density data set δ of the sub-monitoring area is obtained. Y .

[0077] δ Y ={δ1,δ 2...... δ n}

[0078] Where n represents the nth sub-monitoring area, m represents the total number of images in the nth sub-monitoring area, and δ n The calculation formula is:

[0079]

[0080] The bird density (δ n-m There are two ways to calculate:

[0081] First method:

[0082] S1. Perform noise reduction processing on the surveillance image to remove salt and pepper noise and Gaussian noise contained in the surveillance image.

[0083] After acquiring surveillance images, due to interference from various external factors and limitations of the equipment itself, the images often inevitably contain different types of noise, among which salt and pepper noise and Gaussian noise are more common and will interfere with subsequent image analysis. In view of this situation, it is necessary to use professional image denoising algorithms to pre-process surveillance images to improve image quality. Before denoising, convert color images into grayscale images.

[0084] For salt and pepper noise, median filtering can be used to remove it. Median filtering replaces the gray value of each pixel in the image with the median of the gray values ​​of the pixels in its neighborhood, effectively filtering out those salt and pepper noise points that obviously deviate from the gray values ​​of the surrounding pixels, making the distribution of pixel gray values ​​in the image smoother and more reasonable.

[0085] As for Gaussian noise, Gaussian filtering is usually used to deal with it. Gaussian filtering is based on the weighted average principle of the Gaussian function. It assigns weights according to the distance between different pixels and the central pixel, and performs convolution operations on the image to reduce the impact of Gaussian noise and make the grayscale changes of the image more consistent with the characteristics of the actual object. After this series of operations, the salt and pepper noise and Gaussian noise in the monitoring image can be effectively removed, laying a good foundation for subsequent operations such as target extraction.

[0086] S2, determining a judgment threshold, the pixel value of the bird part is higher than the threshold, and the background pixel value is lower than the threshold, and then extracting the bird target by threshold segmentation to obtain a binary image;

[0087] On the basis of successful noise removal, in order to accurately extract the bird target from the monitoring image, a suitable judgment threshold needs to be determined. The determination of this threshold is crucial, as it plays a key role in distinguishing the bird part from the background part. Through the analysis of a large number of sample images and the study of the distribution law of bird and background pixel values, with the help of relevant image processing software or algorithm tools, an appropriate pixel value is found as the judgment threshold.

[0088] After the threshold is determined, the pixel value of the bird part in the image is higher than this threshold, while the background pixel value is lower than this threshold. Then, the threshold segmentation technology is used to process the image. Threshold segmentation is to divide the pixels in the image into different categories based on the set threshold. Specifically, the part with pixel values ​​higher than the threshold is judged as the area where the bird target is located, and the part with pixel values ​​lower than the threshold is classified as the background area. After such a segmentation operation, the originally complex monitoring image is converted into a binary image consisting of only black and white pixel values, in which the white part represents the bird target and the black part represents the background, thereby clearly extracting the bird target.

[0089] S3. A binary image is obtained based on the threshold segmentation, and the bird density is calculated by counting the number of white pixels.

[0090] According to the binary image obtained by threshold segmentation, since the bird targets appear as white pixels in the binary image, the number of birds is indirectly reflected by counting the number of white pixels.

[0091] The specific calculation process can be achieved with the help of professional image analysis algorithms or image processing libraries in programming languages. The entire binary image is scanned row by row and column by column, and once white pixels are detected, they are counted. After counting the number of white pixels, combined with relevant parameters such as the area of ​​the actual monitoring area corresponding to the monitoring image, a reasonable mathematical model is established, such as dividing the number of white pixels by the number of pixels corresponding to the unit area, and then the number of birds per unit area, that is, the bird density, is obtained. In this way, the density of bird distribution in the monitoring area can be intuitively understood, providing strong data support for subsequent related work such as bird ecology research and environmental assessment.

[0092] Second method:

[0093] Build a YOLO model, input the surveillance image into the model, obtain the bounding box of the bird target, divide the surveillance image into several sub-areas, count the number of birds detected in each sub-area, and use the average or sum of the number of birds in each sub-area to represent the bird density.

[0094] The construction process of the YOLO model involves multiple key steps, from the selection of the network architecture to the configuration of the parameters of each layer, all of which need to be carefully arranged according to the specific task requirements. First, you need to choose the appropriate YOLO version. Different versions have differences in network depth, feature extraction capabilities, and detection accuracy. After comprehensively considering factors such as the resolution of the monitoring image, the size and complexity of the bird target, select the corresponding version, such as YOLOv5 or YOLOv8.

[0095] The YOLO model is usually composed of multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer is responsible for extracting image features and capturing feature information of different scales through convolution kernels of different sizes. As the number of network layers increases, the abstraction level of features continues to increase. The pooling layer plays a role in downsampling, which reduces the amount of data while retaining key feature information to prevent overfitting. The fully connected layer is used for the final classification and regression tasks, mapping the extracted features to specific categories and the bounding box coordinates of the target.

[0096] After building the basic network structure, the model needs to be trained. Collect a large number of surveillance image samples containing bird targets and accurately labeled as training data sets, and prepare validation sets and test sets to evaluate the performance of the model. During the training process, by defining a suitable loss function to measure the difference between the predicted results and the actual annotations, the model parameters are continuously adjusted using optimization algorithms (such as stochastic gradient descent, etc.), so that the model can accurately learn the feature representation of bird targets and the corresponding bounding box positioning method. After multiple rounds of iterative training, until the model reaches satisfactory performance indicators on the validation set and test set, a YOLO model suitable for bird target detection in surveillance images is successfully built.

[0097] When the built YOLO model is ready, the next step is to input the actual surveillance image into the model for target detection. As the input data of the model, the surveillance image must be in a format that meets the requirements of the model and usually requires necessary preprocessing, such as unifying the image size and normalizing the pixel values, to ensure that the model can process it normally and stably.

[0098] After the preprocessed surveillance image is sent to the trained YOLO model, each layer of the network inside the model begins to extract and analyze the image features in turn. From the basic features of the image such as texture and color extracted by the shallow network to the more abstract and semantic features obtained by the deep network, these feature information will be comprehensively utilized in the subsequent processing process.

[0099] Then, through the final classification and regression branches of the model, the bird targets in the image are classified and the bounding box positioning is predicted. The final output will clearly give the bounding box coordinate information corresponding to each detected bird target. This bounding box accurately defines the position of the bird target in the monitoring image in the form of a rectangle, including the coordinates of its upper left corner and lower right corner, etc., so that the specific location of each bird target in the image can be clearly known.

[0100] In order to analyze the distribution density of birds in surveillance images in a more detailed and quantitative manner, it is necessary to divide the surveillance images into several appropriate sub-areas based on factors such as the size of the surveillance images, the usual activity range of bird targets, and the analysis accuracy requirements.

[0101] There are many ways to divide the image, such as using a uniform grid division method to divide the image into rectangular sub-regions of equal size according to a fixed number of rows and columns. This method is simple and intuitive and convenient for subsequent statistical calculations. Alternatively, adaptive division can be performed based on different scene features in the image, such as dividing sub-regions of different shapes and sizes based on different landforms, vegetation distribution and other background elements in the image, so that the divided sub-regions are more in line with the actual habitat and activity patterns of birds, and are more conducive to accurately reflecting the distribution density of birds.

[0102] After the surveillance image is divided into multiple sub-areas, the number of birds in each sub-area is counted based on the bird target bounding box information obtained by the YOLO model.

[0103] Specifically, traverse each sub-region and check the bounding box coordinates of all bird targets at the same time to determine whether the center coordinates of each bounding box fall within the sub-region currently being counted. If the center coordinates of the bounding box are within the sub-region, the corresponding bird target is counted into this sub-region. By performing such traversal and judgment operations on all sub-regions in turn, the number of birds contained in each sub-region can be accurately counted. This quantity information intuitively reflects the distribution of birds in different sub-regions and is the key basis for further calculating bird density.

[0104] Bird density aims to reflect the number of birds in a unit area. Different indicators can be used to represent it based on the divided sub-areas and the number of birds counted.

[0105] A common method is to calculate the average number of birds in each sub-area. The number of birds counted in all sub-areas is added up, and then divided by the total number of sub-areas. The average value can be used as a measure of bird density. This average value can roughly reflect the average density of bird distribution in the area covered by the entire monitoring image, and show the overall distribution of birds.

[0106] In addition, the sum of the number of birds in the sub-area can also be used to represent the bird density, but it usually needs to be considered in combination with relevant parameters such as the total area of ​​the sub-area.

[0107] Bird density dataset for the monitoring area δ Y The density in the sub-monitoring area is sorted (arranged from large to small), and a weight value Q is assigned to each sub-monitoring area according to the sorting, wherein the weight value Q of the last sub-monitoring area in the sorting is set to 1, the weight value Q of the second to last sub-monitoring area in the sorting is set to 2, and so on.

[0108] For example, the second sub-monitoring area is ranked first, and its corresponding weight value is Q=4; the third sub-monitoring area is ranked second, and its corresponding weight value is Q=3; the first sub-monitoring area is ranked third, and its corresponding weight value is Q=2; the fourth sub-monitoring area is ranked fourth, and its corresponding weight value is Q=1.

[0109] Calculate the weight coefficient K of each sub-monitoring area according to the weight value Q n , K n The calculation formula is as follows:

[0110]

[0111] Wherein, n represents the nth sub-monitoring area.

[0112] K1=0.2, K2=0.4, K3=0.3, K4=0.1.

[0113] The monitoring time of each sub-monitoring area in the next round is calculated according to the weight coefficient to obtain the monitoring time data set T y , where y represents the round of the monitoring cycle.

[0114] Ty={t1,t2,t3......t n}

[0115] Where n represents the monitoring time of the nth sub-monitoring area in the yth monitoring cycle, t n The calculation formula is:

[0116] t n =K n ×(t o ×n)

[0117] For example, in the second monitoring cycle:

[0118] t1=0.2*60*4=48S, t2=0.4*60*4=96S, t3=0.3*60*4=72S, t4=0.1*60*4=24S.

[0119] The surveillance camera follows the monitoring time dataset T in the yth round of monitoring cycle. y Allocate monitoring time to each sub-monitoring area.

[0120] The patrol laser transmitter in the yth round of monitoring cycle follows the monitoring time data set T y Allocate working time for each sub-monitoring area and set support threshold δ k , for example, when the bird density δ in the second sub-monitoring area n Greater than the threshold δ k , start the supporting laser transmitter and make it scan the second sub-monitoring area. The duration of the scanning is set to t k . Supports scanning working time of laser transmitter t k After the above scanning time is over, the supporting laser returns to its position. If the supporting laser transmitter receives a lower command during the scanning work, the command will be queued for execution as a standby command.

[0121] When the bird density δ in a sub-monitoring area n Two consecutive rounds are greater than the threshold δ k , increases the operating power of the supporting laser transmitter.

[0122] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art may make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. An intelligent identification multifunctional bird repellent device, characterized in that: Includes surveillance camera, intelligent controller, laser transmitter group, power supply box and base; The monitoring camera is a rotatable camera, and the monitoring range of the monitoring camera can cover the entire monitoring area; The laser transmitter group includes a patrol laser transmitter and a support laser transmitter. The patrol laser transmitter is used to continuously transmit lasers in the monitoring area. The support laser transmitter receives instructions from the intelligent controller to transmit lasers into the designated monitoring area. The intelligent controller receives the surveillance photos transmitted by all the sub-surveillance cameras in real time, calculates the bird density value of the corresponding surveillance area according to the surveillance photos, and chooses to start the patrol laser transmitter alone or start the patrol laser transmitter and the support laser transmitter at the same time according to the bird density value; The power supply box is used to supply power to the surveillance camera group, the intelligent controller and the laser transmitter group; The base includes a base body and a mounting rod vertically arranged on the base body, and the mounting rod is used to provide mounting support for the laser transmitter group, the power supply box and / or the monitoring camera group.

2. The intelligent identification multifunctional bird repellent device as claimed in claim 1, characterized in that: The patrol laser emitter is a semiconductor laser emitter, and the support laser emitter is a helium-neon laser emitter.

3. The intelligent identification multifunctional bird repellent device as claimed in claim 1, characterized in that: The patrol laser emitter and the support laser emitter are both arranged on the mounting rod, and the horizontal positions of the patrol laser emitter and the support laser emitter are consistent.

4. A method for repelling birds using the device of claim 1, characterized in that: The surveillance area of ​​the surveillance camera is divided into N sub-surveillance areas, and the surveillance camera performs cyclic monitoring on the N sub-surveillance areas; The monitoring time of each sub-monitoring area in the initial round is set to t0. The intelligent controller receives the monitoring images transmitted by the monitoring camera and forms a monitoring image set; the monitoring image set is an array of (N, M, W, H, 3), where N is the number of sub-monitoring areas, M is the total number of images of each sub-monitoring area, and the image size is expressed as: W×H×3; Perform image recognition on the initial round of surveillance image sets, calculate the bird density of each surveillance image, and obtain the surveillance image bird density dataset δ x ; δx={δ 1-1 ,d 1-2 ,d 1-3 ......d n-m } Wherein, n represents the nth sub-monitoring area, and m represents the mth image of the nth sub-monitoring area; According to the surveillance image bird density dataset δ x The bird density in each sub-monitoring area is calculated, and the bird density data set δ of the sub-monitoring area is obtained. Y ; d Y ={δ1,δ 2...... d n } Where n represents the nth sub-monitoring area, m represents the total number of images in the nth sub-monitoring area, and δ n The calculation formula is: Bird density dataset for the monitoring area δ Y The density in is sorted in descending order, and a weight value Q is assigned to each sub-monitoring area according to the sorting, wherein the weight value Q of the last sub-monitoring area in the sorting is set to 1, the weight value Q of the second-to-last sub-monitoring area in the sorting is set to 2, and so on; Calculate the weight coefficient K of each sub-monitoring area according to the weight value Q n , K n The calculation formula is as follows: Where n represents the nth sub-monitoring area; The monitoring time of each sub-monitoring area in the next round is calculated according to the weight coefficient to obtain the monitoring time data set T y , where y represents the round of monitoring cycle; Ty={t1,t2,t3......t n } Where n represents the monitoring time of the nth sub-monitoring area in the yth monitoring cycle, t n The calculation formula is: t n =K n ×(t o ×n) The surveillance camera follows the monitoring time dataset T in the yth round of monitoring cycle. y Allocate monitoring time for each sub-monitoring area; The patrol laser transmitter in the yth round of monitoring cycle follows the monitoring time data set T y Allocate working time for each sub-monitoring area and set support threshold δ k , when the bird density δ in the nth sub-monitoring area n Greater than the threshold δ k When the support laser transmitter is started, the support laser transmitter is made to scan the nth sub-monitoring area, and the duration of the scanning work is set to t k .

5. The intelligent identification multifunctional bird-repelling method as claimed in claim 4, characterized in that: The image of each sub-monitoring area is captured at an interval of 0.5-1S to obtain M monitoring images.

6. The intelligent identification multifunctional bird-repelling method as claimed in claim 4, characterized in that: The bird density of each monitoring image is calculated as follows: Perform noise reduction on the surveillance image to remove salt and pepper noise and Gaussian noise contained in the surveillance image; Determine a judgment threshold, the pixel value of the bird part is higher than the threshold, and the background pixel value is lower than the threshold, then extract the bird target through threshold segmentation to obtain a binary image; The binary image is obtained by threshold segmentation, and the bird density is calculated by counting the number of white pixels.

7. The intelligent identification multifunctional bird-repelling method as claimed in claim 6, characterized in that: Before denoising, the color image is converted to grayscale.

8. The intelligent identification multifunctional bird-repelling method as claimed in claim 4, characterized in that: The bird density of each monitoring image is calculated as follows: Build a YOLO model, input the surveillance image into the model, obtain the bounding box of the bird target, divide the surveillance image into several sub-areas, count the number of birds detected in each sub-area, and use the average or sum of the number of birds in each sub-area to represent the bird density.

9. The intelligent identification multifunctional bird-repelling method as claimed in claim 4, characterized in that: The scanning working time t of the supporting laser transmitter k It is 60-90S.

10. The intelligent identification multifunctional bird-repelling method according to claim 4, characterized in that: When the bird density δ in a sub-monitoring area n Two consecutive rounds are greater than the threshold δ k , increases the operating power of the supporting laser transmitter.

Images