Novel system and method for observing and recording behaviors in honeycomb

Abstract

The invention relates to a novel system and method for observing and recording behaviors in a honeycomb, and belongs to the technical field of bee behavior research. The beehive is installed at the top of the supporting cabinet, limiting grooves for installing honeycomb frames are formed in the beehive, monitoring windows are formed in the two sides of the beehive, transparent glass plates are installed on the monitoring windows, the monitoring boxes are installed on the two sides of the beehive, one side of each monitoring box is open and is in butt joint with the corresponding monitoring window, and cameras are installed in the monitoring boxes. The controller and the signal transmission module are both installed in the supporting cabinet, the camera is connected with the controller, and the controller is connected with the monitoring rear end through the signal transmission module. A camera is used for observing and recording behaviors of marked bees in a honeycomb and generating a video, and computer software is used for carrying out statistical analysis on the bee behavior video, so that data such as a crawling track of the bees on a honeycomb, honeycomb probing duration frequency, inter-individual behavior duration frequency and the like can be obtained; therefore, the behavior in the honeycomb can be comprehensively, deeply and accurately investigated and studied.

Description

Technical Field

[0001] The present invention belongs to the technical field of bee behavior research, and particularly relates to a novel observation and recording system and method for bee behavior inside the hive. Background Art

[0002] Bees are completely social model insects. They not only provide humans with green, nutritious, and health-care bee products such as honey and royal jelly, but more importantly, their pollination role contributes to the increased production and income of crops, fruits, vegetables, etc., and the role they play in maintaining ecological diversity. In order to comprehensively understand this important insect, scientists have been studying bees from different perspectives. Currently, there are still many blanks in bee biology that have not been filled. One very important reason is that we cannot truly, comprehensively, and efficiently monitor the natural living state of bees. Observing bee behavior inside the hive still only stays at observing outside the hive or opening the hive. Observing outside the hive cannot well understand the situation inside the beehive. Vibrations, external light, or temperature changes generated during opening the hive will interfere with the normal order of the bee colony. Especially during winter, opening the hive for observation has a greater impact on the bee colony, affecting the normal behavior of bees inside the hive and the comprehensiveness and accuracy of the observation results. Summary of the Invention

[0003] In order to overcome the problems in the background art that vibrations, external light, or temperature changes generated during opening the hive will interfere with the bee colony, opening the hive for observation in winter has a greater impact on the bee colony, affecting the normal behavior of bees inside the hive and the comprehensiveness and accuracy of the observation results. The present invention collects the behavior of marked bees inside the hive through video, and analyzes the data of the collected bee behavior through the YOLO11 object detection model, obtaining data including the crawling trajectory of the marked bees on the honeycomb, the visiting duration and frequency of specific cells, and the duration and frequency data of the mutual behavior between marked bee individuals, so as to more comprehensively, deeply, and accurately investigate and study the behavior of bees inside the hive.

[0004] The present invention provides a novel observation and recording system and method for bee behavior inside the hive; by opening monitoring windows on both sides of the beehive, observing the marked bee colony through a camera, collecting video information in real time through the camera, and recording the behavior of each marked bee, it is possible to observe the behavior of bees inside the bee colony without opening the beehive, and without disturbing the bee colony, observe the behavior of bees inside the bee colony; ensuring the natural nest temperature of the bee colony, the dark environment, and minimizing the interference with the normal order of the bee colony, thereby improving the comprehensiveness and accuracy of observing the behavior of bees inside the hive.

[0005] To achieve the above object, the present invention is implemented by the following technical solutions:

[0006] A novel beehive behavior observation and recording system mainly includes a supporting cabinet 1, a beehive 2, a monitoring box 3, a camera 4, a controller 5, a signal transmission module 6, and a monitoring back end; the beehive 2 is installed on the top of the supporting cabinet 1, and a limiting groove 21 for installing a honeycomb frame is provided in the beehive 2. Monitoring windows 22 are opened on both sides of the beehive 2, and transparent glass plates are installed on the monitoring windows 22. The monitoring box 3 is installed on both sides of the beehive 2, and one side of the monitoring box 3 is opened and docked on the monitoring window 22. A camera 4 is installed in the monitoring box 3, and the controller 5 and the signal transmission module 6 are both installed in the supporting cabinet 1, the camera 4 is connected to the controller 5, and the controller 5 is connected to the monitoring back end through the signal transmission module 6.

[0007] Furthermore, the side panels on both sides of the beehive 2 can be detachably installed through tower buckles 23.

[0008] Furthermore, infrared fill light strips 24 are installed on the side panels of both sides of the beehive 2 , and the infrared fill light strips 24 are connected to the controller 5 .

[0009] Furthermore, the camera 4 is installed in the monitoring box 3 through a position adjustment device, and the position adjustment device includes a longitudinal guide rail 41, a longitudinal slider 42, a longitudinal electric push rod 43, a transverse slider 44, and a transverse electric push rod 45. The longitudinal guide rail 41 is installed in the monitoring box 3, the longitudinal slider 42 can be slidably installed on the longitudinal guide rail 41, the longitudinal electric push rod 43 is installed in the monitoring box 3 and is transmission-connected with the longitudinal slider 42, the longitudinal slider 42 is provided with a transverse dovetail groove 46, the transverse slider 44 can be slidably installed in the transverse dovetail groove 46, the transverse electric push rod 45 is installed on the longitudinal slider 42 and is transmission-connected with the transverse slider 44, the camera 4 is installed on the transverse slider 44, and the longitudinal electric push rod 43 and the transverse electric push rod 45 are both connected to the controller 5.

[0010] Furthermore, the support cabinet 1 is provided with an operating door 11 for easy opening and maintenance. The top of the monitoring box 3 is provided with an openable cover plate 31.

[0011] Furthermore, a box cover 25 is provided on the top of the beehive 2 for easy operation and management, and a limiting edge 26 is provided on the box cover 25 to be sleeved on the top of the beehive 2.

[0012] In another aspect, a novel method for observing and recording the behavior of bees in a nest is provided, wherein the method is applicable to the system, and the recording method comprises:

[0013] Design a bee back tag and then stick the tag on the bee's back;

[0014] The marked bees are placed in the observation beehive, and the front and back sides of a honeycomb are photographed with a high-definition camera to obtain a video of the bees' behavior in the nest;

[0015] Extract pictures from the collected video at a certain frame rate. For the extracted pictures, use a data annotation tool for annotation;

[0016] Use the YOLO11 object detection model for training. First, annotate and train the model for the circular label area; then detect the circular label area and intercept the label image, annotate and train the model for the triangle inside the intercepted image; then detect the triangle identifier, determine the rotation angle according to the center points of the triangle and the circular label, rotate the circular label area according to the rotation angle, and after rotating it to be upright, annotate and train the character recognition model for individual characters.

[0017] Adopt a method of combining three models. Use the circular label detection model to detect the circular label on the back of the bee. After extracting the circular label area, use the triangle detection model to detect the triangle identifier, obtain the rotation angle and rotate the circular label area so that the triangle identifier is located directly above the image. Then use the character recognition model to recognize the rotated circular label area, detect the characters and determine which bee it is.

[0018] Detect the empty honeycomb frame and locate individual cells, determine the coordinate positions of each cell, compare the cell coordinates on the captured empty honeycomb frame with the cell coordinates in the video to calculate the offset, and correct the coordinates of the cells.

[0019] Determine whether the bee is exploring the cell and the exploration duration by calculating the coordinate points and time difference, and record and store the detected data;

[0020] By querying the stored data, reproduce the crawling trajectory of the bee on the honeycomb frame within a period of time, and locate the data such as which cells the bee has explored and the exploration duration frequency within this period of time.

[0021] Furthermore, the label on the back of the bee is a circle with a diameter of 3.00 mm, the background is black, the upper part of the label is an isosceles triangle with a direction identifier, the color is white, and the lower part is a combination of two of numbers, uppercase and lowercase letters, and the font color is white.

[0022] Furthermore, the annotation area is the circular label area on the back of the bee, and an area with obvious label features is selected for annotation

[0023] The beneficial effects of the present invention:

[0024] The present invention observes the marked bees through a camera by opening monitoring windows on both sides of the beehive, and collects video information in real time, records the behaviors of the marked bees, and can observe the behaviors of bees in the bee colony without opening the beehive, which can observe the behaviors of bees in the bee colony without disturbing the bee colony; it ensures the natural nest temperature of the bee colony, the dark environment and the minimization of interference with the normal order of the bee colony, thereby improving the comprehensiveness and accuracy of the observation of the behaviors of bees in the hive.

[0025] The present invention collects the behaviors of the marked bees in the hive through video, and analyzes the data of the collected bee behaviors through the YOLO11 object detection model, obtaining data including the crawling trajectories of the marked bees on the honeycomb frames, the visiting duration and frequency of specific cells, and the duration and frequency data of the mutual behaviors among the marked bees. Thus, it can investigate and study the behaviors of bees in the hive more comprehensively, deeply and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a three-dimensional schematic diagram of the present invention.

[0027] Figure 2 is a three-dimensional schematic diagram of the partially opened state of the present invention.

[0028] Figure 3 is a main-axis isometric three-dimensional schematic diagram of the disassembled state of the beehive of the present invention.

[0029] Figure 4 is a bottom-axis isometric three-dimensional schematic diagram of the disassembled state of the beehive of the present invention.

[0030] Figure 5 is a three-dimensional schematic diagram of the camera installation structure of the present invention.

[0031] Figure 6 is a flowchart of the method of the present invention.

[0032] Figure 7 is a schematic diagram of the label of the present invention.

[0033] Figure 8 is a flowchart of the combined detection using multiple models of the present invention.

[0034] Figure 9 is a schematic diagram of the rotation angle detection.

[0035] Figure 10 is the detection effect diagram of the triangular direction identification.

[0036] Figure 11 is the final effect diagram after the combined detection and recognition by calling three models.

[0037] Figure 12 is a reproduction diagram of the bee trajectory. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] In order to make the objectives, technical solutions, and beneficial effects of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings to facilitate understanding by those skilled in the art.

[0039] Embodiment 1:

[0040] As Figures 1 - 5 shown, the present invention discloses a new type of in-hive bee behavior observation and recording system. The new type of in-hive bee behavior observation and recording system mainly includes a support cabinet 1, a beehive 2, a monitoring box 3, a camera 4, a controller 5, a signal transmission module 6, and a monitoring backend. The beehive 2 is installed on the top of the support cabinet 1. The beehive 2 is provided with a limit groove 21 for installing honeycomb frames. Monitoring windows 22 are opened on both sides of the beehive 2. Transparent glass plates are installed on the monitoring windows 22. The monitoring box 3 is installed on both sides of the beehive 2. One side of the monitoring box 3 is open and butted against the monitoring window 22. The camera 4 is installed in the monitoring box 3. The controller 5 and the signal transmission module 6 are both installed in the support cabinet 1. The camera 4 is connected to the controller 5. The controller 5 is connected to the monitoring backend through the signal transmission module 6. By opening monitoring windows on both sides of the beehive, the marked bees are observed through the camera, and video information is collected in real time to record the behaviors of the marked bees. There is no need to open the beehive for observation, and the behaviors of bees in the bee colony can be observed without disturbing the normal order of the bee colony. It ensures the natural nest temperature of the bee colony, the dark environment, and minimizes the interference with the normal order of the bee colony, thereby improving the comprehensiveness and accuracy of the observation of the in-hive bee behaviors.

[0041] The two side plates of the beehive 2 are detachably installed through tower buckles 23, which is convenient for cleaning and daily management of the inside of the beehive 2.

[0042] Infrared supplementary light strips 24 are installed on the two side plates of the beehive 2. The infrared supplementary light strips 24 are connected to the controller 5. Supplementary lighting is carried out through the infrared supplementary light strips 24, which ensures clear vision and avoids the influence of visible light on the bee colony.

[0043] The camera 4 is installed in the monitoring box 3 through a position adjustment device, and the position adjustment device includes a longitudinal guide rail 41, a longitudinal slider 42, a longitudinal electric push rod 43, a transverse slider 44, and a transverse electric push rod 45. The longitudinal guide rail 41 is installed in the monitoring box 3, the longitudinal slider 42 can be slidably installed on the longitudinal guide rail 41, the longitudinal electric push rod 43 is installed in the monitoring box 3 and is transmission-connected with the longitudinal slider 42, the longitudinal slider 42 is provided with a transverse dovetail groove 46, the transverse slider 44 can be slidably installed in the transverse dovetail groove 46, the transverse electric push rod 45 is installed on the longitudinal slider 42 and is transmission-connected with the transverse slider 44, the camera 4 is installed on the transverse slider 44, and the longitudinal electric push rod 43 and the transverse electric push rod 45 are both connected to the controller 5; the longitudinal electric push rod 43 and the transverse electric push rod 45 can be used to adjust the position of the camera 4 in the front, back, left, and right directions, adjust the observation field of view, and improve the convenience and clarity of observation and monitoring.

[0044] The support cabinet 1 is provided with an operating door 11 for easy opening and maintenance. The top of the monitoring box 3 is provided with an openable cover plate 31.

[0045] The top of the beehive 2 is provided with a box cover 25 for easy operation and management, and the box cover 25 is provided with a limiting edge 26 which is sleeved on the top of the beehive 2 to ensure that the beehive cover is stably and reliably connected.

[0046] Working process:

[0047] The present invention opens monitoring windows on both sides of the beehive, observes the marked bees through cameras and collects video information in real time, records the behavior of the marked bees, and does not need to open the beehive for observation. The behavior of the bees in the bee colony can be observed without disturbing the normal order of the bee colony; the natural nest temperature and dark environment of the bee colony and the minimization of interference with the normal order of the bee colony are ensured, thereby improving the comprehensiveness and accuracy of the observation of the behavior of the bees in the nest.

[0048] The present invention can implement video recording of the behavior of marked bees in the nest. After data analysis of the behavior video by computer software, data including the crawling track of the marked bees on the honeycomb, the duration and frequency of visits to specific nest cells, and the duration and frequency of mutual behavior between marked bees can be obtained. This helps us to investigate and study the behavior of bees in the nest more comprehensively, deeply and accurately.

[0049] Embodiment 2:

[0050] Based on the system constructed in Example 1, a new method for observing and recording the behavior of bees in a nest is constructed, which specifically includes:

[0051] S1. Design the bee back label: The label is a circle with a diameter of 3.00 mm, the background is black, the upper part of the label is an isosceles triangle with a direction identifier, the color is white, and the lower part is a combination of two of numbers, uppercase and lowercase letters, and the font color is white. The designed back label can be directly printed on A4 paper and then carefully pasted on the back plate of the bee.

[0052] The number of individual bees in a bee colony is large and their body size is small, so it is very difficult to distinguish them without marking. The key to the study of bee nest behavior lies in whether individual bees can be uniquely identified. The present invention designs a back label combined with numbers and uppercase and lowercase letters to mark individual bees, which solves problems such as the lack of individual uniqueness in traditional paint marking, the small sample size of pure digital marking, the need for special equipment to make and print bar code labels for bar code marking, and the inapplicability of RFID marking to the hive environment, etc. The newly designed back label for individual bees is a circle with a diameter of 3.00 mm, the background is black, the upper part of the label is an isosceles triangle with a direction identifier, the color is white, and the lower part is a combination of two of numbers, uppercase and lowercase letters. The width of each single character is 1.20 mm, the height is 1.75 mm, the character spacing is 0.20 mm, the font color is white, and the width of the font line is 0.20 mm. The designed back label can be directly printed on A4 paper, with the characteristics of low cost and large marking quantity. At the same time, compared with markings such as circular codes and two-dimensional codes, the characters can be recognized by the human eye, which is convenient for comparing the accuracy of the detection results of software recognition and human eye recognition. The shape of the back label and the labeled bees are as follows Figure 7 .

[0053] S2. Put the labeled bees into the observation hive and take pictures of both sides of a honeycomb by a high-definition camera to collect the in-hive behavior video of the bees.

[0054] Paste the label on the back of the bee, put the labeled bee into the observation hive, and take pictures of both sides of a honeycomb by a high-definition camera to collect the in-hive behavior video of the bees. Extract pictures from the collected video at a certain frame rate. For the extracted pictures, use a data annotation tool for annotation. The annotation area is the circular label area on the back of the bee, and select the area with obvious label features for annotation.

[0055] At the same time, in order to facilitate the recognition of the characters in the bee back label, the circular label area needs to be rotated. In order to obtain the rotation direction angle, the center point of the circular label area and the center point of the triangular identifier area need to be determined. In order to obtain the center point of the triangular identifier area, the center point of the triangular identifier area is marked and trained, and then detected and recognized.

[0056] S3. Extract pictures from the collected video at a certain frame rate. For the extracted pictures, use a data annotation tool to annotate them. The annotation area is the circular label area on the back of the bee. Select the area with obvious label features for annotation.

[0057] For the rotated image, in order to recognize each character, it is necessary to perform annotation training on individual characters and then perform detection and recognition. At the same time, in order to ensure the recognition effect, for incomplete or dirty characters, even if the characters are not fully displayed or have certain stains, custom annotation can be performed according to the unique features and structures of the characters, so that the model can better understand and adapt to characters in different situations, significantly improving the recognition accuracy and robustness.

[0058] S4. Use the YOLO11 object detection model. First, train the circular label detection model, extract the circular label area and annotate the triangular area, then train the triangular detection model, obtain the rotation angle and rotate the circular label area, rotate the triangular identifier to the directly above of the back label, and finally annotate the rotated characters and perform the character detection model training.

[0059] In step S4, the YOLO11 object detection model is used to perform layer-by-layer detection and recognition on the circular label, triangular identifier and back label characters in the bee back label. The specific implementation process is as follows:

[0060] 1. Data preparation and annotation

[0061] Data collection: First, extract pictures from the video of the bee's in-nest behavior collected in step S2 at a certain frame rate, annotate the circular labels in the pictures, then cut out the circular label area, and annotate the triangle and characters.

[0062] Annotation categories:

[0063] Circular label detection: Annotate the circular labels in the image, defined as a category "Circle".

[0064] Triangular area detection: Detect the circular label and cut out the circular label area, and further annotate the white isosceles triangular area, defined as a category "Triangle".

[0065] Character detection: After detecting the triangle in the cut-out circular label area, rotate the circular label to the positive direction according to the position of the triangle, and annotate individual numbers, uppercase and lowercase letter characters in the rotated label area, defined as multiple categories (such as "A", "B", "1", "2", etc.).

[0066] Annotation Tool: Use LabelImg or similar tools to annotate images, generating corresponding Bounding Boxes and class labels.

[0067] 2. Model Architecture and Configuration

[0068] YOLO11 Model Architecture: YOLO is a deep learning-based object detection algorithm. Its core principle is to transform the object detection task into a regression problem, and simultaneously predict the bounding boxes and class probabilities of multiple objects in an image through a single convolutional neural network. Its basic architecture includes a backbone network, a neck network, and a head network. YOLO can directly predict the positions and classes of all objects in an image by running the network only once on an image, with the advantage of real-time detection, and is suitable for large-scale bee behavior monitoring.

[0069] Configuration File:

[0070] Input Size: Normalize the input image size (such as 640×640) according to the label size (diameter 3.00mm) and video resolution.

[0071] Number of Classes: Set according to the annotated classes. For example, there is 1 class for circular labels, 1 class for triangular regions, and several classes for character categories.

[0072] Anchor Box Setting: Based on the target size distribution in the dataset, adjust the size and ratio of the anchor boxes to improve detection accuracy.

[0073] 3. Phased Training Process

[0074] 3.1 Training of Circular Label Detection Model

[0075] First, use the YOLO11 model to train the circular label detection model. The specific steps are as follows:

[0076] Loss Function: Adopt the multi-task loss function of YOLO, including:

[0077] Location Loss (L_coord): Measure the position error between the predicted bounding box and the ground truth bounding box.

[0078] Confidence Loss (L_conf): Measure the confidence that the bounding box contains an object.

[0079] Class Loss (L_cls): Measure the difference between the predicted class and the ground truth class.

[0080] Expressed by the mathematical formula as

[0081]

[0082] Among them, λ coord 、λnoobj and λ cls are loss weights, S is the number of grid divisions, and B is the number of bounding boxes per grid.

[0083] Training process: Use the pre-divided circular label dataset for multiple rounds of iterative training, and optimize the model parameters through backpropagation until the loss function converges.

[0084] 3.2 Training of the triangular region detection model

[0085] After the circular label detection model training is completed, perform annotation training on the triangular identifiers within the circular label region. The specific steps are as follows:

[0086] Data extraction: Use the circular label detection model to detect the bee nest behavior images, cut out the detected circular label regions and annotate the triangles, which are used as the dataset for the new stage of training.

[0087] Loss function and optimization: Also adopt the multi-task loss function of YOLO

[0088] Training process: Use the cropped triangular region dataset for training, and optimize the model parameters through multiple rounds of iteration to improve the detection accuracy of the triangular region.

[0089] 3.3 Training of the rotated character detection model

[0090] After the triangular region detection model training is completed, rotate the circular label region according to the position of the triangle and perform annotation training on the characters. The specific steps are as follows:

[0091] Data extraction and rotation: For the cropped circular label region, use the triangular detection model for detection. After detecting the triangular identifier, calculate the center coordinates of the triangular identifier. Based on the cropped circular label region, connect the center of the triangular region with the center of the circular label region, and the angle between the connection line and the X-axis can be obtained. Then use the rotation matrix to rotate the image to standardize the character direction.

[0092] Mathematically expressed as:

[0093]

[0094] where θ is the angle to be rotated.

[0095] Data annotation: Annotate the rotated characters and define the category of each character (such as "A", "B", "a", "b", "1", "2", etc.).

[0096] Loss function and optimization: Adopt a similar multi-task loss function, but the number of categories is adjusted according to the character categories.

[0097] Training process: Use the rotated character dataset for training, optimize the model parameters through multiple rounds of iteration, and improve the detection and recognition accuracy of characters.

[0098] By training the YOLO11 model in stages, gradually achieve high-precision detection and recognition of circular labels, triangular identifiers, and characters on the bee back labels. The training of each stage depends on the detection results of the previous stage to ensure the hierarchical and modular nature of the overall system. With the support of mathematical formulas, ensure the accuracy of the model in geometric transformation and object detection. Finally, achieve the detection of the unique identification of a large number of individual bees and behavior tracking, providing reliable data support for the study of bee in-nest behavior.

[0099] S5. Obtain and recognize the back label area of the bee through the circular label detection model, calculate the rotation angle, rotate the triangular identifier to the directly above of the back label, and use the character recognition model to recognize the characters in the rotated label area to detect which bee it is.

[0100] Adopt the method of combined detection with multiple models to detect the video of bee in-nest behavior collected. Use the circular label detection model to detect the back label area, and then use the triangular detection model to detect the internal triangular identifier in this area after detecting the back label area. After detecting the identifier, calculate the rotation angle, then rotate the triangle to the directly above of the label area, and finally use the character recognition model to recognize the characters in the rotated label area to detect which bee it is. The entire recognition process is as follows Figure 8 :

[0101] As Figure 9 shown, the circular label detection model and the triangular detection model detect the circular label area and the triangular identifier, and then the center of the back label area and the center point of the triangular identifier can be obtained. The center point of the back label area is C = (C x , C y ), and the center point of the triangular identifier is S = (S x , S y ). The calculation of the rotation angle θ is based on the difference between the center of the triangle and the center of the back label. The horizontal difference Δ x = S x - C x , and the vertical difference Δ y = S y - C y . Then use the arctangent function arctan to calculate the angle between the line segment pointing from the character center to the triangle center and the x-axis Finally, it needs to be rotated to point to the positive y-axis direction, so use 90° - θ to obtain the final rotation angle.

[0102] After obtaining the rotation angle θ, rotate the label area to ensure that the triangular symbol is directly above the label after rotation. The detection effect of the triangular direction identifier is as follows Figure 10 , where the green box in the figure represents the circular label area, the blue box represents the detected triangular area, and the right figure shows the effect after rotation.

[0103] After rotating to the positive direction, the character recognition model can be called for character recognition. During the recognition process, the order of the detected characters is random. To ensure that the final obtained character combination is correct, according to the coordinates of the two target character boxes and arranging them in ascending order of the X coordinate of the character boxes, the correct character combination can be obtained. The following Figure 11 is the final effect after combining the detection and recognition by calling three models.

[0104] S6. Detect and locate each honeycomb cell, determine the position of each honeycomb cell, and adjust the coordinates according to the actual situation to compensate for the influence caused by the offset of the honeycomb.

[0105] In order to completely reproduce the trajectory of the bees on the honeycomb frame and locate which honeycomb cells have been visited, it is necessary to detect and locate each honeycomb cell. First, place an empty honeycomb frame without bees in the observation box and take a picture of this empty honeycomb frame. Then, detect each honeycomb cell. First, convert the input grayscale image into a binary image. In the binary image, the honeycomb cells appear black, while the surrounding background appears white. By detecting the boundaries of the binary image, the connected regions of the binary image can be extracted, and information such as the shape, size, and position of each honeycomb cell in the image can be obtained.

[0106] After obtaining the contour of each honeycomb cell, the center coordinates and diameter of each honeycomb cell can be obtained, so as to determine the position of each honeycomb cell. At the same time, record the coordinates of the four honeycomb cells in the upper right, lower right, upper left, and lower left corners of the honeycomb frame. Since the honeycomb frame will be taken out and put back during the experiment, it will cause a slight offset in the position of the honeycomb frame. Therefore, each time the video is recognized, first detect the four honeycomb cells in the upper right, lower right, upper left, and lower left corners of the first frame of the video, and select the center point of one of the honeycomb cells and compare it with the corresponding center point coordinates on the empty honeycomb frame to calculate the offset in the x and y axes. Finally, after obtaining the center coordinates of the back label during the detection of the labeled bees, add this offset. In this way, the image can be dynamically corrected to ensure the accuracy of the detection result positioning.

[0107] Since the label is attached to the back of the bee, but the bee explores the hive with its head. When the head enters the cell, it drives the back label to disappear from the picture together. However, there is a certain distance between the position where the back label disappears and the bee's head. To accurately represent the position of the bee's head, it is also necessary to measure the distance between the center position of the bee's head and the center position of the back label. After the model detects the label and obtains the center coordinates of the label, offset a certain distance from the center point of the label towards the bee's head to obtain the accurate position of the bee's head. During the process of attaching the back label to the bee, the triangular identifier points to the bee's head. Therefore, the direction from the center of the label to the center of the triangle is the direction of the bee's head. Finally, store the time when the circular back label is detected, the characters on the back label, and the position coordinates of the bee's head on the honeycomb.

[0108] S7. Determine whether the bee is exploring a cell, which cell it is exploring, and the exploration duration by calculating the coordinate points and time differences, and store the detected data.

[0109] Such as Figure 12 , after detecting and recognizing the characters on the back label of the video of the bee's behavior in the hive and correcting the cell positioning, store the data. By querying the data corresponding to the bees with back labels within a certain period of time stored, connecting these time points, the crawling trajectory of the labeled bees on the comb within a certain period of time can be reproduced. When a bee wants to explore a certain cell, it will put its head into this cell, and the label attached to the bee's back will also disappear from the picture. By calculating the time difference between two consecutive coordinate points, after obtaining the time difference, then compare the distance between these two points. If this distance difference does not exceed the diameter of a cell, it is determined that the bee has explored the cell during this period, and the exploration duration is the time difference between the disappearance point and the next appearance point. After calculating and processing the consecutive coordinate points within a certain period of time, it is possible to know which cells this bee has explored during this period and the exploration duration of each cell.

[0110] After determining the center and diameter of each cell on an empty comb, select a period of time and query which back labels appear within the diameter range of this cell during this period. The calculation method is as follows: The coordinates of the center of the cell are (h, k), the radius is r, and the coordinates of the position of the bee's back label are (x, y). If (x - h) 2 +(y - k) 2 ≤r 2It can be determined that the back label appears within the honeycomb diameter range. For each back label, find the position where the back label appears at the next time point. Since the time difference between when the bee enters the cell and reappears is approximately equal to 2 seconds, if the time difference exceeds 2 seconds and the position of the back label is within the diameter range of this cell, it is determined that the bee corresponding to this back label has visited this cell. After calculating all the points within this time period, it can be obtained which bees with back labels have visited this cell during this time period.

[0111] S8. By querying the stored data, reproduce the crawling trajectory of the bee with a back label on the comb within a time period, and obtain which cells the bee has explored during this time period and the exploration duration for each cell.

[0112] Through its observation and recording system, the present invention collects the behavior of the marked bees in the hive through video, and analyzes the data of the collected bee behavior through the YOLO11 object detection model, obtaining data including the crawling trajectory of the marked bees on the comb, the visiting duration frequency of specific cells, and the duration frequency data of the mutual behavior between marked bee individuals, so as to investigate and study the behavior of bees in the hive more comprehensively, deeply, and accurately.

[0113] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A new type of bee nest behavior observation and recording system, characterized by: The novel beehive behavior observation and recording system comprises a supporting cabinet (1), a beehive (2), a monitoring box (3), a camera (4), a controller (5), a signal transmission module (6), and a monitoring back end; the beehive (2) is installed on the top of the supporting cabinet (1), a limiting groove (21) for installing a honeycomb frame is provided in the beehive (2), monitoring windows (22) are provided on both sides of the beehive (2), and transparent glass plates are installed on the monitoring windows (22); the monitoring box (3) is installed on both sides of the beehive (2), one side of the monitoring box (3) is opened and docked on the monitoring window (22), a camera (4) is installed in the monitoring box (3), the controller (5) and the signal transmission module (6) are both installed in the supporting cabinet (1), the camera (4) is connected to the controller (5), and the controller (5) is connected to the monitoring back end through the signal transmission module (6).

2. A novel beehive behavior observation and recording system as claimed in claim 1, characterized in that: The side panels on both sides of the beehive (2) are detachably mounted via tower buckles (23).

3. A novel beehive behavior observation and recording system as claimed in claim 1 or 2, characterized in that: Infrared fill light strips (24) are installed on the side panels on both sides of the beehive (2), and the infrared fill light strips (24) are connected to the controller (5).

4. A novel beehive behavior observation and recording system as claimed in claim 3, characterized in that: The camera (4) is installed in the monitoring box (3) through a position adjustment device. The position adjustment device comprises a longitudinal guide rail (41), a longitudinal slider (42), a longitudinal electric push rod (43), a transverse slider (44), and a transverse electric push rod (45). The longitudinal guide rail (41) is installed in the monitoring box (3). The longitudinal slider (42) can be slidably installed on the longitudinal guide rail (41). The longitudinal electric push rod (43) is installed in the monitoring box (3) and is transmission-connected with the longitudinal slider (42). The longitudinal slider (42) is provided with a transverse dovetail groove (46). The transverse slider (44) can be slidably installed in the transverse dovetail groove (46). The transverse electric push rod (45) is installed on the longitudinal slider (42) and is transmission-connected with the transverse slider (44). The camera (4) is installed on the transverse slider (44). The longitudinal electric push rod (43) and the transverse electric push rod (45) are both connected to the controller (5).

5. A novel beehive behavior observation and recording system as claimed in any one of claims 1, 2 and 4, characterized in that: The support cabinet (1) is provided with an operating door (11) which is convenient for opening and maintenance.

6. A novel beehive behavior observation and recording system as claimed in claim 5, characterized in that: The top of the monitoring box (3) is provided with an openable cover plate (31).

7. A novel beehive behavior observation and recording system as claimed in any one of claims 1, 2, 4 and 6, characterized in that: The top of the beehive (2) is provided with a box cover (25) for easy operation and management, and the box cover (25) is provided with a limiting edge (26) which is sleeved on the top of the beehive (2).

8. A novel method for observing and recording the behavior of bees in a nest, wherein the method is applicable to the system according to any one of claims 1 to 7, characterized in that: The recording method includes: Design a bee back tag and then stick the tag on the bee's back; The marked bees are placed in the observation beehive, and the front and back sides of a honeycomb are photographed with a high-definition camera to obtain a video of the bees' behavior in the nest; Extract images from the collected video at a certain frame rate, and use data annotation tools to annotate the extracted images; Use the YOLO11 target detection model for training. First, annotate the circular label area and train the model. Then detect the circular label and capture the label area image. Annotate the triangle inside the captured image and train the model. Then detect the triangle identifier, determine the rotation angle based on the center point of the triangle and the center point of the circular label, rotate the circular label area according to the rotation angle, and annotate the single character after rotating to the positive direction and train the character recognition model. The three models are combined. The circular label detection model detects the circular label on the back of the bee, and the triangle detection model is used to detect the triangle identifier after extracting the circular label area. The rotation angle is obtained and the circular label area is rotated so that the triangle identifier is located directly above the image. The character recognition model is then used to identify the rotated circular label area. Once the characters are detected, it is determined which bee it is. The empty combs are detected and individual cells are located to determine the coordinates of each cell. The coordinates of the cells on the empty comb images are compared with the coordinates of the cells in the video to calculate the offset and correct the coordinates of the cells. By calculating the coordinate points and time difference, it is determined whether the bees are exploring the nest and the exploration time, and the detected data is recorded and stored; By querying the stored data, the crawling trajectory of the bees within a period of time can be reproduced on the honeycomb, and the nest cells that the bees have explored within this period of time can be located, as well as data such as the exploration duration and frequency.

9. A novel method for observing and recording the behavior of bees in a nest according to claim 8, characterized in that: The back label is a circle with a diameter of 3.00 mm, a black background, an isosceles triangle with a direction mark on the upper part of the label, the color is white, and the lower part is a combination of numbers and uppercase and lowercase letters in pairs, the font color is white.

10. A novel method for observing and recording the behavior of bees in a nest according to claim 8, characterized in that: The marked area is the circular label area on the back of the bee, and the area with obvious label features is selected for marking.

Images