A laser bird repelling device and method of use

Abstract

The application provides an automatic bird repelling device and a use method, and belongs to the technical field of laser bird repelling. The automatic bird repelling device comprises a support platform which is fixedly arranged relative to the ground; a wide-angle camera which is arranged on the support platform and is used for continuously imaging a set monitoring risk area to realize landing detection of a target; a two-axis holder which is arranged on the support platform and is used for horizontal and vertical rotating movement relative to the support platform; a zoom camera and a laser bird repeller which are both arranged on a movable end of the two-axis holder, the zoom camera is used for secondary identification of the target in the monitoring risk area where the landed harmful bird is located; the laser bird repeller is used for emitting a laser beam to the area where the harmful bird is located after the secondary identification of the target; and a control unit which is integrated on the support platform, is used for receiving the landing detection result of the target of the wide-angle camera and the secondary identification result of the target of the zoom camera, and is further used for driving the two-axis holder or the laser bird repeller to act.

Description

Technical Field

[0001] This invention relates to the field of laser bird deterrence technology, and in particular to a laser bird deterrence device and its usage method. Background Technology

[0002] High-value aquaculture and fruit-growing areas are frequently attacked by birds, causing significant economic losses to farmers and growers, especially during the harvest season when large numbers of birds peck at the produce, making it difficult to control. Furthermore, birds are highly adaptable, and conventional methods only offer short-term partial effects, rarely maintaining long-term deterrence. Common bird-repelling methods include: 1. Manually using a long pole or hand gestures; this method is ineffective against birds at a distance. 2. Throwing airburst bombs at the flock, using the loud noise to frighten the birds; this method is cumbersome, and the birds quickly return after the airburst, limiting its long-term effectiveness. 3. Using ultrasound to stimulate birds, causing physiological discomfort and thus repelling them; however, this method has a limited effective range, requires multiple ultrasound devices for large-scale application, and is difficult to install and maintain; furthermore, it is only effective against specific bird species, lacking versatility. 4. Using a laser to shine a strong light into the birds' eyes, causing discomfort and thus repelling them; this method requires the laser to be aimed directly at the bird's eye, making long-distance operation very difficult and potentially damaging to the birds' eyes.

[0003] Therefore, it is essential to provide a laser bird deterrent device and its usage method, which identifies whether a target entering a protected area is a harmful bird, predicts the bird's movement trajectory and stopping position, and uses a laser virtual rod to drive away the bird without causing harm to it, thereby improving the actual effect and efficiency of bird deterrence. Summary of the Invention

[0004] In view of this, the present invention proposes a laser bird deterrent device and its usage method that performs a target detection in a preset area, identifies a target that has a risk of landing, and then performs a second target detection and further uses laser to simulate sticks to drive away the bird.

[0005] On one hand, the present invention provides a laser bird deterrent device, comprising:

[0006] The support platform is fixed relative to the ground.

[0007] A wide-angle camera, mounted on the support platform, is used to continuously image the designated monitoring risk area, establish a unified visual benchmark, and achieve target landing detection.

[0008] A two-axis gimbal is mounted on the support platform and rotates relative to the support platform in both horizontal and vertical directions.

[0009] A zoom camera, mounted on the movable end of the two-axis gimbal, is used to focus on the monitoring risk area where the landed harmful bird is located, thereby achieving secondary identification of the target.

[0010] A laser deterrent device is installed at the movable end of the two-axis gimbal and is used to emit a laser beam to the area where the zoom camera has identified the target as a harmful bird. The laser deterrent device is arranged parallel to the optical path of the zoom camera and the pointing direction is synchronized.

[0011] The control unit, integrated on the support platform, communicates with the wide-angle camera, the two-axis gimbal, the zoom camera, and the laser deflector. It receives the target landing detection results from the wide-angle camera and the secondary recognition results of the target from the zoom camera. The control unit also drives the two-axis gimbal or the laser deflector.

[0012] Based on the above technical solutions, preferably, the set monitoring risk area is an area at a certain height above the ground or water surface within the field of view of the wide-angle camera.

[0013] Preferably, the wide-angle camera achieves target landing detection, including the following:

[0014] Wide-angle cameras preprocess the acquired images;

[0015] Inter-frame differential processing involves performing differential processing on two consecutive preprocessed images, calculating the gray values ​​of corresponding pixels in the two images and subtracting them, taking the absolute value of the gray value subtraction result, and using the absolute value of the gray value subtraction result to generate a differential grayscale image.

[0016] Change pixel extraction involves detecting changes in a differential grayscale image and extracting the changed regions between two frames.

[0017] Noise reduction is performed by using morphological algorithms to manipulate the changed regions, statistically analyzing their size, and setting a pixel area threshold for the changed regions. This threshold is based on the number of pixels occupied by the most distant typical harmful bird in the field of view. Changed regions outside the pixel area threshold range are filtered out.

[0018] Target extraction involves determining whether the remaining changed area after noise reduction is located within the set monitoring risk area. If the changed area is located within the set monitoring risk area, target extraction is performed on the changed area. The centroid of the changed area is calculated using the weighted centroid method, and the centroid represents the target corresponding to the changed area. The attribute information of the target is stored in the target queue. If the changed area is not located within the set monitoring risk area, target extraction is not performed.

[0019] Target search, identification and recording of the correspondence of each target in continuous images, and construction of a trajectory queue based on each target;

[0020] Trajectory tracking is performed separately for each target's trajectory queue to obtain trajectory data for each target: if the target's trajectory is outside the set monitoring risk area, the target is deleted from the target queue; if the target's trajectory is still within the set monitoring risk area and is moving towards the ground, the trajectory data of the target moving towards the ground is output to the landing trajectory queue.

[0021] Landing detection involves checking the landing trajectory queue for targets. It calculates changes in adjacent position information to determine if a target has landed. If the target has not landed, further trajectory tracking is performed; if the target has landed, a bird-repelling mechanism is triggered.

[0022] After the information is sent and the bird deterrence mechanism is triggered, the wide-angle camera sends the position of the landed target in a unified visual reference to the control unit. The control unit then drives the zoom camera and laser deterrent to take further action and execute the bird deterrence action.

[0023] Further preferably, the wide-angle camera preprocesses the acquired image, specifically by statistically analyzing the overall brightness and variance of the image, calculating the image sharpness using the gradient difference method, and if the image sharpness is less than a set sharpness threshold, adjusting the gain of the wide-angle camera, and then performing histogram equalization on the image after gain adjustment to make the gray-level probability distribution of the image after histogram equalization uniform, thus obtaining the preprocessed image; if the image sharpness is not less than the set sharpness threshold, then no preprocessing is performed on the image.

[0024] More preferably, the target search identifies and records the correspondence of each target in continuous images, constructing a trajectory queue based on each target. Specifically, it employs a three-dimensional target search based on Manhattan distance. Based on the statistical analysis of the two-dimensional Manhattan distance between the current target and another target in an image at a certain moment, it analyzes the relationship between the current target and targets in the previous moment's image, finding the continuous correspondence of the current target in the time dimension; let the current target be k, and the other target be j. Let the distance between target k and target j at time i be the two-dimensional Manhattan distance. Let be the two-dimensional Manhattan distance between the current target k at time i and the target j at time i-1; min(·) is the minimum two-dimensional Manhattan distance between the current target k and another target in the image at time i; search the target queue at time i-1, and when the current target k at time i and the target j at time i-1 satisfy the formula At that time, the information of the current target k and target j is saved to the trajectory queue, and the information of the current target k and target j in the continuous image is recorded.

[0025] In a further preferred embodiment, the landing detection involves performing landing detection on targets in the landing trajectory queue, calculating changes in adjacent position information of the trajectory to determine whether the target has landed. Specifically, within a given time range, if the change in the position of the target's center of gravity in the height direction within the set monitoring risk area does not exceed twice the distance of the longest path between any two points on the contour boundary corresponding to the initial dwell time, then the target is determined to have landed.

[0026] In a further preferred embodiment, after the control unit receives the position of the landed target in a unified visual reference, it obtains the deviation between the current position of the target and the current pointing direction of the zoom camera based on the calibration relationship between the zoom camera and the wide-angle camera. Based on the deviation, it adjusts the servo motor of the two-axis gimbal so that the zoom camera and the laser bird deterrent point to the current position of the target, and the zoom camera confirms the harmful bird for the landed target.

[0027] In a further preferred embodiment, the zoom camera confirms the presence of harmful birds on the ground by acquiring continuous images of the target, performing differential change detection on the continuous images, and normalizing the results of the differential change. When the range of change of the target exceeds a set threshold, the zoom camera determines that there is an active target in the field of view, which is a harmful bird. The set threshold for the range of change of the target is set according to the number of pixels occupied by the most distant typical harmful bird in the field of view. The zoom camera sends a drive-away command to the laser bird deterrent. After the laser bird deterrent is activated, it enlarges the laser spot and directs it toward the current pointing direction of the zoom camera. It constructs a virtual rectangle around the target's center of gravity and fits several trajectories passing through the center of gravity within the rectangle. The endpoints of each trajectory are located on two different sides of the virtual rectangle. The output spot of the laser bird deterrent moves back and forth along the endpoints of the different trajectories.

[0028] On the other hand, the present invention provides a method of using a laser bird deterrent device, comprising the following steps:

[0029] S1: Configure the aforementioned laser bird deterrent device; arrange wide-angle cameras and two-axis gimbals at different positions on the support platform. The servo motors built into the two-axis gimbals enable the pitch and rotation movements of the movable end of the two-axis gimbals. The movable end of the two-axis gimbals is equipped with a zoom camera and a laser deflector. The laser deflector and the zoom camera's optical path are set parallel and their pointing directions are synchronized; establish a unified visual reference based on the wide-angle camera, that is, calibrate the correspondence between the camera coordinate system of the zoom camera and the camera coordinate system of the wide-angle camera, so that the image obtained by the wide-angle camera or the image obtained by the zoom camera can be transformed into a corresponding position in the world coordinate system; configure a control unit on the support platform;

[0030] S2: Within the field of view of the wide-angle camera, select an area at a certain height above the ground or water surface as the monitoring risk area. The wide-angle camera and zoom camera ignore objects outside the monitoring risk area.

[0031] S3: When a target enters the monitored risk area, the wide-angle camera performs landing detection of the target, including the following:

[0032] S31: The wide-angle camera preprocesses the acquired images;

[0033] S32: Inter-frame differential processing, performing differential processing on two consecutive preprocessed images, calculating the gray values ​​of corresponding pixels in the two images and subtracting them, taking the absolute value of the gray value subtraction result, and using the absolute value of the gray value subtraction result to generate a differential grayscale image.

[0034] S33: Change pixel extraction, performs change detection on the differential grayscale image, and extracts the change region in the two frames of the image;

[0035] S34: Noise reduction processing, which operates on the changing region through morphological algorithms, counts the size of the changing region, sets a pixel area threshold for the changing region, and filters out the changing region that is not within the pixel area threshold range;

[0036] S35: Target extraction, determining whether the remaining change area after noise reduction is located within the set monitoring risk area: If the change area is located within the set monitoring risk area, target extraction is performed on the change area, the centroid of the change area is calculated using the weighted centroid method, the centroid represents the target corresponding to the change area, and the target's attribute information is stored in the target queue; if the change area is not located within the set monitoring risk area, target extraction is not performed.

[0037] S36: Target search, identify and record the correspondence of each target in continuous images, and construct a trajectory queue based on each target;

[0038] S37: Trajectory tracking. Track each target's trajectory queue separately to obtain trajectory data for each target. If the target's trajectory is outside the set monitoring risk area, delete the target from the target queue. If the target's trajectory is still within the set monitoring risk area and is moving towards the ground, output the trajectory data of the target moving towards the ground to the landing trajectory queue.

[0039] S38: Landing Detection. This function performs landing detection on targets in the landing trajectory queue, calculating changes in adjacent position information to determine if the target has landed. If the target has not landed, further trajectory tracking is performed; if the target has landed, a bird-repelling mechanism is triggered.

[0040] S39: Information transmission. After the bird deterrence mechanism is triggered, the zoom camera sends the position of the landed target in a unified visual reference to the control unit.

[0041] S4: After receiving the position of the landing target in the unified visual reference, the control unit drives the change of the current position of the zoom camera, so that the zoom camera and the laser bird deterrent point to the target position.

[0042] S5; The zoom camera performs a secondary confirmation of the landed target. When the range of change of the target after landing exceeds the set threshold, the zoom camera confirms that the target is a harmful bird and sends the information to the control unit. At this time, the control unit activates the laser bird deterrent device. The laser bird deterrent device has a built-in laser generator and beam expander. The laser generator generates a laser beam, and the beam expander thickens the laser beam. The thickened laser beam is shot towards the current pointing direction of the zoom camera, constructing several trajectories passing through the center of the target. The thickened laser beam sweeps along the trajectory and appears around the harmful bird, making the bird visually perceive the thickened laser beam as a physical stick. In order to avoid being hit by the stick, it will quickly flee, thereby achieving timely removal of the predatory harmful bird.

[0043] S6: The zoom camera continuously monitors changes in the field of view and performs differential change detection on the continuously acquired images. If the change in the field of view exceeds the threshold, it indicates that there are still harmful birds in the field of view. At this time, the laser deterrent is guided to point to the area in the field of view where the change is drastic until the change in the field of view of the zoom camera is less than the threshold, indicating that there are no active targets in the field of view. After all the harmful birds are driven away, the laser is turned off, and the driving away operation is completed.

[0044] Preferably, the pixel area threshold for setting the changing region in step S34 is obtained by combining the bird populations and seasonal changes in the aquaculture and fruit planting areas, based on the range of pixel area values ​​of the pest bird species appearing in the current season in the wide-angle camera.

[0045] The present invention provides a laser bird deterrent device and its usage method, which, compared with the prior art, have the following advantages:

[0046] Beneficial effects:

[0047] (1) Based on a deep analysis of the hunting behavior characteristics of pest birds in breeding or planting areas, this invention proposes an efficient method for detecting pest birds and identifying their landing behavior. It drives away only pest birds with potential hunting behavior, greatly improving the system's response speed and reducing the system's operating frequency. In order to meet the needs of a wide monitoring area, the camera for close-up imaging is obtained by combining a wide-angle camera and a zoom camera. The depth of field can be dynamically adjusted to meet the needs of large-scale, close-up fine imaging.

[0048] (2) Based on the biological characteristics of bird vision, the laser beam is thickened by expanding the laser spot. After enhancement, the laser beam will form a virtual stick effect in the bird's field of vision. The bird will have the illusion of being driven away by the stick and will quickly escape. The virtual stick does not need to directly stimulate the bird's eyes. The efficiency is far greater than the driving mechanism of laser directly stimulating the bird's eyes. Since there is no need to aim at the bird's eyes, the operation is faster and will not damage the bird's visual system.

[0049] (3) This scheme uses a wide-angle camera to identify and predict the trajectory of targets entering the monitoring risk area. Only when the trajectory of the target is pointing downwards toward the ground or water surface is it identified as a potential target. Then, a zoom camera is used for secondary target identification, which makes the identification of harmful birds more accurate. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a structural block diagram of a laser bird deterrent device and its usage method according to the present invention;

[0052] Figure 2 This is a flowchart illustrating the landing detection of a target using a wide-angle camera in a laser bird deterrent device and its usage method according to the present invention. Detailed Implementation

[0053] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0054] Existing laser bird-repelling methods utilize the intense light of a laser beam to shine into the bird's eyes, causing discomfort and thus driving them away. This method requires the laser to be aimed directly at the bird's eye, making long-distance operation very difficult, time-consuming, and potentially damaging to the bird's eyes. Therefore, if... Figure 1 As shown, in one aspect, the present invention provides a laser bird deterrent device, comprising:

[0055] The support platform is fixed relative to the ground and is used to support a wide-angle camera, a two-axis gimbal, a zoom camera, a laser drive, and a control unit.

[0056] A wide-angle camera, mounted on the supporting platform, is used to continuously image the designated monitoring risk area, establish a unified visual benchmark, and achieve target landing detection. The designated monitoring risk area refers to a region selected within the wide-angle camera's field of view at a certain height above the ground or water surface, such as several meters above the ground or water surface.

[0057] A two-axis gimbal is mounted on the support platform and rotates horizontally and vertically relative to the platform. The movable end of the two-axis gimbal can pitch relative to the horizontal plane and rotate vertically.

[0058] A zoom camera, mounted on the movable end of the two-axis gimbal, is used to focus on the monitoring risk area where the fallen harmful bird is located, enabling secondary identification of the target. Wide-angle cameras have a large market, but the target size in the image is small, and the detected target may be falling objects, leaves, etc. These targets usually do not change shape and position after landing. Therefore, using a zoom camera for secondary identification is beneficial to eliminate irrelevant targets, accurately identify harmful birds, and improve the accuracy of subsequent laser bird deterrence.

[0059] A laser deterrent device is installed at the movable end of the two-axis gimbal and is used to emit a laser beam to the area where the zoom camera has identified the target as a harmful bird. The laser deterrent device is arranged parallel to the optical path of the zoom camera and the pointing direction is synchronized.

[0060] The control unit, integrated on the support platform, communicates with the wide-angle camera, the two-axis gimbal, the zoom camera, and the laser deflector. It receives the target landing detection results from the wide-angle camera and the secondary recognition results of the target from the zoom camera. The control unit also drives the two-axis gimbal or the laser deflector.

[0061] The wide-angle camera enables target landing detection, specifically including the following:

[0062] 1) The wide-angle camera preprocesses the acquired image; specifically, it calculates the overall brightness and variance of the image, uses the gradient difference method to calculate the image sharpness, and if the image sharpness is less than the set sharpness threshold, it adjusts the gain of the wide-angle camera. The image after gain adjustment is then subjected to histogram equalization to make the gray level probability distribution of the image after histogram equalization uniform, thus obtaining the preprocessed image; if the image sharpness is not less than the set sharpness threshold, it is not preprocessed and directly enters the subsequent inter-frame difference processing operation.

[0063] 2) Inter-frame differential processing: Perform differential processing on two consecutive preprocessed images, calculate the gray values ​​of corresponding pixels in the two images, subtract them, calculate the absolute value of the gray value subtraction result, and use the absolute value of the gray value subtraction result to generate a differential grayscale image.

[0064] 3) Change pixel extraction: Perform change detection on the differential grayscale image and extract the change region between the two frames;

[0065] 4) Noise reduction processing: Morphological algorithms are used to operate on the changing regions, the size of the changing regions is counted, and a pixel area threshold for the changing regions is set. The pixel area threshold for the changing regions is set by the number of pixels occupied by the typical harmful bird at the farthest point in the field of view; changing regions that are not within the range of the pixel area threshold are filtered out.

[0066] 5) Target extraction: Determine whether the remaining change area after noise reduction is located within the set monitoring risk area. If the change area is located within the set monitoring risk area, target extraction is performed on the change area. The centroid of the change area is calculated using the weighted centroid method, and the centroid represents the target corresponding to the change area. The attribute information of the target is stored in the target queue. If the change area is not located within the set monitoring risk area, target extraction is not performed. The target queue stores one or more targets that have entered the set monitoring risk area. They can be saved separately according to the smallest row and column order of the edge position of each target in the wide-angle image.

[0067] If there are multiple targets, they are arranged into a target queue according to the order of their rows and columns in the image. For images acquired at different times, the number and position of targets in the images are not exactly the same, thus allowing for the creation of target queues corresponding to different times.

[0068] The weighted centroid method mentioned here binarizes the extracted target region, preserving the grayscale values ​​of pixels within the target boundary while setting the grayscale values ​​of pixels outside the boundary to 0, thus obtaining the target's outline. Then, the centroid of the target is calculated using the grayscale values ​​of each pixel within the target outline and the grayscale centroid method.

[0069] f(u, v) is the gray value at the image coordinates (u, v).

[0070] 6) Target search: Identify and record the correspondence of each target in continuous images, and construct a trajectory queue based on each target.

[0071] Specifically, the method employs a three-dimensional target search based on Manhattan distance. By statistically analyzing the two-dimensional Manhattan distance between the current target and another target in an image at a given moment, it identifies the relationship between the current target and targets in the previous moment's image, thus finding the continuous correspondence of the current target in the time dimension. Let the current target be k, and the other target be j. Let the distance between target k and target j at time i be the two-dimensional Manhattan distance. Let be the two-dimensional Manhattan distance between the current target k at time i and the target j at time i-1; min(·) is the minimum two-dimensional Manhattan distance between the current target k and another target in the image at time i; search the target queue at time i-1, and when the current target k at time i and the target j at time i-1 satisfy the formula At that time, the information of the current target k and target j is saved to the trajectory queue, and the information of the current target k and target j in the continuous image is recorded.

[0072] 7) Trajectory tracking: Track each target's trajectory queue separately to obtain trajectory data for each target. If the target's trajectory is outside the set monitoring risk area, the target is deleted from the target queue. If the target's trajectory is still within the set monitoring risk area and is moving towards the ground, the trajectory data of the target moving towards the ground is output to the landing trajectory queue.

[0073] In practice, the trajectory data can be converted to three-dimensional coordinates in the world coordinate system. The trajectory curve of the target is fitted according to the three-dimensional coordinates. If the trajectory curve is oriented towards the ground of the set monitoring risk area, the current target is retained in the target queue. If it is determined that the trajectory curve is oriented outside the set monitoring risk area, it will continue to track for several sampling cycles. After confirming that the target has moved outside the set monitoring risk area, the target is removed from the target queue.

[0074] 8) Landing detection: Detect the landing of targets in the landing trajectory queue, calculate the changes in adjacent position information of the trajectory to determine whether the target has landed. If the target has not landed, further trajectory tracking is performed; if the target has landed, the bird deterrence mechanism is triggered.

[0075] Specifically, within a given time frame, if the change in the center of gravity of a target in the height direction within a designated monitoring risk area does not exceed twice the distance of the longest path between any two points on the contour boundary corresponding to the initial dwell time, then the target is considered to have landed.

[0076] 9) Information transmission: After the bird deterrence mechanism is triggered, the wide-angle camera sends the position of the landed target in the unified visual reference to the control unit. The control unit drives the zoom camera and laser deterrent to take further action and execute the bird deterrence action.

[0077] Once the control unit receives the position of the landed target in a unified visual reference, it obtains the deviation between the target's current position and the zoom camera's current pointing direction based on the calibration relationship between the zoom camera and the wide-angle camera. Based on the deviation, it adjusts the servo motor of the two-axis gimbal so that the zoom camera and the laser bird deterrent point to the target's current position, and the zoom camera confirms the harmful bird based on the landed target.

[0078] The zoom camera identifies harmful birds by acquiring continuous images of the landed target. These images are then subjected to differential change detection, and the results are normalized. If the range of change of the landed target exceeds a set threshold, the zoom camera determines that a moving target, a harmful bird, exists in the field of view. This threshold is set based on the number of pixels occupied by a typical harmful bird at the farthest point in the field of view. The zoom camera then sends a deflection command to a laser bird deterrent. Upon activation, the laser bird deterrent enlarges its laser beam and directs it in the zoom camera's current direction. It constructs a virtual rectangle around the target's center of gravity and fits several trajectories passing through the center of gravity within this rectangle. The endpoints of these trajectories lie on two different sides of the virtual rectangle. The laser bird deterrent's output beam moves back and forth along these endpoints. The enlarged laser beam acts like a virtual stick in the bird's vision, startling and driving the bird away.

[0079] In addition, the present invention provides a method for using a laser bird deterrent device, comprising the following steps:

[0080] S1: Configure the aforementioned laser bird deterrent device; arrange wide-angle cameras and two-axis gimbals at different positions on the support platform. The servo motors built into the two-axis gimbals enable the pitch and rotation movements of the movable end of the two-axis gimbals. The movable end of the two-axis gimbals is equipped with a zoom camera and a laser deflector. The laser deflector and the zoom camera's optical path are set parallel and their pointing directions are synchronized; establish a unified visual reference based on the wide-angle camera, that is, calibrate the correspondence between the camera coordinate system of the zoom camera and the camera coordinate system of the wide-angle camera, so that the image obtained by the wide-angle camera or the image obtained by the zoom camera can be transformed into a corresponding position in the world coordinate system; configure a control unit on the support platform;

[0081] S2: Within the field of view of the wide-angle camera, select an area at a certain height above the ground or water surface as the monitoring risk area. The wide-angle camera and zoom camera ignore objects outside the monitoring risk area.

[0082] S3: When a target enters the monitored risk area, the wide-angle camera performs landing detection of the target, including the following:

[0083] S31: The wide-angle camera preprocesses the acquired images;

[0084] S32: Inter-frame differential processing, performing differential processing on two consecutive preprocessed images, calculating the gray values ​​of corresponding pixels in the two images and subtracting them, taking the absolute value of the gray value subtraction result, and using the absolute value of the gray value subtraction result to generate a differential grayscale image.

[0085] S33: Change pixel extraction, performs change detection on the differential grayscale image, and extracts the change region in the two frames of the image;

[0086] S34: Noise reduction processing. Morphological algorithms are used to operate on the changing areas, count the size of the changing areas, and set a pixel area threshold for the changing areas. The pixel area threshold for the changing areas is set by the number of pixels occupied by the typical harmful bird at the farthest point in the field of view. Changing areas that are not within the range of the pixel area threshold are filtered out. The pixel area threshold for the changing areas mentioned here is obtained by combining the bird populations and seasonal changes in aquaculture and fruit planting areas, based on the range of pixel area values ​​of the harmful bird species appearing in the current season in the wide-angle camera. Even in the same area, the bird species appearing in different seasons are not exactly the same.

[0087] S35: Target extraction, determining whether the remaining change area after noise reduction is located within the set monitoring risk area: If the change area is located within the set monitoring risk area, target extraction is performed on the change area, the centroid of the change area is calculated using the weighted centroid method, the centroid represents the target corresponding to the change area, and the target's attribute information is stored in the target queue; if the change area is not located within the set monitoring risk area, target extraction is not performed.

[0088] S36: Target search, identify and record the correspondence of each target in continuous images, and construct a trajectory queue based on each target;

[0089] S37: Trajectory tracking. Track each target's trajectory queue separately to obtain trajectory data for each target. If the target's trajectory is outside the set monitoring risk area, delete the target from the target queue. If the target's trajectory is still within the set monitoring risk area and is moving towards the ground, output the trajectory data of the target moving towards the ground to the landing trajectory queue.

[0090] S38: Landing Detection. This function performs landing detection on targets in the landing trajectory queue, calculating changes in adjacent position information to determine if the target has landed. If the target has not landed, further trajectory tracking is performed; if the target has landed, a bird-repelling mechanism is triggered.

[0091] S39: Information transmission. After the bird deterrence mechanism is triggered, the zoom camera sends the position of the landed target in a unified visual reference to the control unit.

[0092] S4: After receiving the position of the landing target in the unified visual reference, the control unit drives the change of the current position of the zoom camera, so that the zoom camera and the laser bird deterrent point to the target position.

[0093] S5; The zoom camera performs a secondary confirmation of the landed target. When the range of change of the target after landing exceeds the set threshold, the zoom camera confirms that the target is a harmful bird and sends the information to the control unit. At this time, the control unit activates the laser bird deterrent device. The laser bird deterrent device has a built-in laser generator and beam expander. The laser generator generates a laser beam, and the beam expander thickens the laser beam. The thickened laser beam is shot towards the current pointing direction of the zoom camera, constructing several trajectories passing through the center of the target. The thickened laser beam sweeps along the trajectory and appears around the harmful bird, making the bird visually perceive the thickened laser beam as a physical stick. In order to avoid being hit by the stick, it will quickly flee, thereby achieving timely removal of the predatory harmful bird.

[0094] S6: The zoom camera continuously monitors changes in the field of view and performs differential change detection on the continuously acquired images. If the change in the field of view exceeds the threshold, it indicates that there are still harmful birds in the field of view. At this time, the laser deterrent is guided to point to the area in the field of view where the change is drastic until the change in the field of view of the zoom camera is less than the threshold, indicating that there are no active targets in the field of view. After all the harmful birds are driven away, the laser is turned off, and the driving away operation is completed.

[0095] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A laser bird deterrent device, characterized in that, include: The support platform is fixed relative to the ground. A wide-angle camera, mounted on the support platform, is used to continuously image the designated monitoring risk area, establish a unified visual benchmark, and achieve target landing detection. The target landing detection involves acquiring changing regions in the image. If the changing region is located within a designated monitoring risk area, the target is extracted from the changing region. The centroid of the changing region is calculated using a weighted centroid method, and the centroid represents the target corresponding to the changing region. The trajectory of each target is acquired, and target search is performed. The correspondence of each target in continuous images is identified and recorded, and a trajectory queue based on each target is constructed. Target tracking is performed on the trajectory queue of each target. If the target's trajectory is within the designated monitoring risk area and moves towards the ground, the target is landed. The change in the adjacent position information of the target's trajectory is calculated to determine whether the target has landed. Specifically, within a given time range, if the change in the position of the target's centroid in the height direction within the designated monitoring risk area does not exceed twice the distance of the longest path between any two points on the contour boundary corresponding to the initial dwell time, the target is determined to have landed, triggering the bird deterrence mechanism. The target search identifies and records the correspondence of each target in continuous images, constructing a trajectory queue based on each target. Specifically, it employs a three-dimensional target search based on Manhattan distance. Based on the statistical analysis of the two-dimensional Manhattan distance between the current target and another target in an image at a given moment, it analyzes the relationship between the current target and targets in the previous moment's image, finding the continuous correspondence of the current target in the time dimension. Let the current target be... k Another goal is j , For the first i Current target at any given moment k With the goal j The two-dimensional Manhattan distance; For the first i Current target at any given moment k With the i -1 time target j The two-dimensional Manhattan distance; Is it to obtain the first i Current target in time image k The minimum two-dimensional Manhattan distance between the target and another target; for the th i The target queue is searched at time -1. i Current goal at any given moment k With the i The goal at time -1 j Satisfy the formula At that time, the current target k and target j The information is saved to the trajectory queue, recording the current target. k and target j Information in a series of images; A two-axis gimbal is mounted on the support platform and rotates relative to the support platform in both horizontal and vertical directions. A zoom camera, mounted on the movable end of the two-axis gimbal, is used to focus on the monitoring risk area where the landed harmful bird is located, thereby achieving secondary identification of the target. A laser deterrent device is installed at the movable end of the two-axis gimbal and is used to emit a laser beam to the area where the zoom camera has identified the target as a harmful bird. The laser deterrent device is arranged parallel to the optical path of the zoom camera and the pointing direction is synchronized. The control unit, integrated on the support platform, communicates with the wide-angle camera, the two-axis gimbal, the zoom camera, and the laser deflector. It receives the target landing detection results from the wide-angle camera and the secondary recognition results of the target from the zoom camera. The control unit also drives the two-axis gimbal or the laser deflector.

2. The laser bird deterrent device according to claim 1, characterized in that, The designated monitoring risk area is an area selected at a certain height above the ground or water surface within the field of view of the wide-angle camera.

3. The laser bird deterrent device according to claim 2, characterized in that, The wide-angle camera's target landing detection also includes the following: Wide-angle cameras preprocess the acquired images; Inter-frame differential processing involves performing differential processing on two consecutive preprocessed images, calculating the gray values ​​of corresponding pixels in the two images and subtracting them, taking the absolute value of the gray value subtraction result, and using the absolute value of the gray value subtraction result to generate a differential grayscale image. Change pixel extraction involves detecting changes in a differential grayscale image and extracting the changed regions between two frames. Noise reduction is performed by manipulating the changed regions using morphological algorithms, statistically analyzing the size of the changed regions, and setting a pixel area threshold for the changed regions. The pixel area threshold for the changed regions is set based on the number of pixels occupied by the most distant typical harmful bird in the field of view. Changed regions outside the pixel area threshold range are filtered out. Then, target search is performed. After the target search is completed and the information is sent, the bird deterrence mechanism is triggered. The wide-angle camera then sends the position of the landed target in a unified visual reference to the control unit. The control unit then drives the zoom camera and laser deterrent to take further action and execute the bird deterrence action.

4. The laser bird deterrent device according to claim 3, characterized in that, The wide-angle camera preprocesses the acquired images by statistically analyzing the overall brightness and variance of the image, calculating the image sharpness using the gradient difference method, and adjusting the gain of the wide-angle camera if the image sharpness is less than a set sharpness threshold. The image after gain adjustment is then subjected to histogram equalization to make the gray-level probability distribution of the image after histogram equalization uniform, thus obtaining the preprocessed image. If the image sharpness is not less than the set sharpness threshold, no preprocessing is performed on the image.

5. A laser bird deterrent device according to claim 3, characterized in that, After receiving the position of the landed target in a unified visual reference, the control unit obtains the deviation between the target's current position and the zoom camera's current pointing direction based on the calibration relationship between the zoom camera and the wide-angle camera. Based on the deviation, it adjusts the servo motor of the two-axis gimbal so that the zoom camera and the laser bird deterrent point to the target's current position, and the zoom camera confirms the harmful bird based on the landed target.

6. A laser bird deterrent device according to claim 3, characterized in that, The zoom camera identifies harmful birds by acquiring continuous images of the target after it lands. These images are then subjected to differential change detection, and the results are normalized. If the range of change of the target exceeds a set threshold, the zoom camera determines that a moving target exists in the field of view and is a harmful bird. This threshold is set based on the number of pixels occupied by a typical harmful bird at the farthest point in the field of view. The zoom camera then sends a deflection command to the laser bird deterrent. Once activated, the laser bird deterrent enlarges its laser spot and directs it in the zoom camera's current direction. It constructs a virtual rectangle around the target's center of gravity and fits several trajectories passing through the center of gravity within this rectangle. The endpoints of each trajectory lie on two different sides of the virtual rectangle. The output spot of the laser bird deterrent moves back and forth along the endpoints of these different trajectories.

7. A method of using a laser bird deterrent device, characterized in that, The steps include the following: S1: Configure the laser bird deterrent device as described in any one of claims 1-6; arrange a wide-angle camera and a two-axis gimbal at different positions on the support platform; the servo motor built into the two-axis gimbal realizes the pitch and rotation of the movable end of the two-axis gimbal; the movable end of the two-axis gimbal is equipped with a zoom camera and a laser deflector; the laser deflector and the optical path of the zoom camera are set parallel and the pointing direction is synchronized; establish a unified visual reference based on the wide-angle camera, that is, calibrate the correspondence between the camera coordinate system of the zoom camera and the camera coordinate system of the wide-angle camera, so that the image obtained by the wide-angle camera or the image of the zoom camera can be transformed into a corresponding position in the world coordinate system; configure a control unit on the support platform; S2: Within the field of view of the wide-angle camera, select an area at a certain height above the ground or water surface as the monitoring risk area. The wide-angle camera and zoom camera ignore objects outside the monitoring risk area. S3: When a target enters the monitored risk area, the wide-angle camera performs landing detection of the target, including the following: S31: The wide-angle camera preprocesses the acquired images; S32: Inter-frame differential processing, performing differential processing on two consecutive preprocessed images, calculating the gray values ​​of corresponding pixels in the two images and subtracting them, taking the absolute value of the gray value subtraction result, and using the absolute value of the gray value subtraction result to generate a differential grayscale image. S33: Change pixel extraction, performs change detection on the differential grayscale image, and extracts the change region in the two frames of the image; S34: Noise reduction processing, which operates on the changing region through morphological algorithms, counts the size of the changing region, sets a pixel area threshold for the changing region, and filters out the changing region that is not within the pixel area threshold range; S35: Target extraction, determine whether the remaining change area after noise reduction is located in the set monitoring risk area: if the change area is located in the set monitoring risk area, then target extraction is performed on the change area, the centroid of the change area is calculated using the weighted centroid method, the centroid is used to represent the target corresponding to the change area, and the attribute information of the target is stored in the target queue. If the area of ​​change is not within the designated monitoring risk area, target extraction will not be performed. S36: Target search, identify and record the correspondence of each target in continuous images, and construct a trajectory queue based on each target; S37: Trajectory tracking. Track each target's trajectory queue separately to obtain trajectory data for each target. If the target's trajectory is outside the set monitoring risk area, then delete the target from the target queue. If the target trajectory is still within the set monitoring risk area and is moving towards the ground, the target trajectory data moving towards the ground will be output to the landing trajectory queue. S38: Landing Detection. This function performs landing detection on targets in the landing trajectory queue, calculating changes in adjacent position information to determine if the target has landed. If the target has not landed, further trajectory tracking is performed; if the target has landed, a bird-repelling mechanism is triggered. S39: Information transmission. After the bird deterrence mechanism is triggered, the zoom camera sends the position of the landed target in a unified visual reference to the control unit. S4: After receiving the position of the landing target in the unified visual reference, the control unit drives the change of the current position of the zoom camera, so that the zoom camera and the laser bird deterrent point to the target position. S5; The zoom camera performs a secondary confirmation of the landed target. When the range of change of the target after landing exceeds the set threshold, the zoom camera confirms that the target is a harmful bird and sends the information to the control unit. At this time, the control unit activates the laser bird deterrent device. The laser bird deterrent device has a built-in laser generator and beam expander. The laser generator generates a laser beam, and the beam expander thickens the laser beam. The thickened laser beam is shot towards the current pointing direction of the zoom camera, constructing several trajectories passing through the center of the target. The thickened laser beam sweeps along the trajectory and appears around the harmful bird, making the bird visually perceive the thickened laser beam as a physical stick. In order to avoid being hit by the stick, it will quickly flee, thereby achieving timely removal of the predatory harmful bird. S6: The zoom camera continuously monitors changes in the field of view and performs differential change detection on the continuously acquired images. If the change in the field of view exceeds the threshold, it indicates that there are still harmful birds in the field of view. At this time, the laser deterrent is guided to point to the area in the field of view where the change is drastic until the change in the field of view of the zoom camera is less than the threshold, indicating that there are no active targets in the field of view. After all the harmful birds are driven away, the laser is turned off, and the driving away operation is completed.

8. The method of using a laser bird deterrent device according to claim 7, characterized in that, The pixel area threshold for setting the changing area mentioned in step S34 is obtained by combining the bird populations and seasonal changes in the aquaculture and fruit planting areas, and based on the range of pixel area values ​​of the pest bird species appearing in the current season in the wide-angle camera.

Images