A goose breeding path guiding system and method for a semi-open breeding scene

Abstract

The application discloses a kind of path guiding system and method for semi-open breeding scene of breeding goose, the system includes semi-open breeding farm and guiding control subsystem;Breeding farm includes rest activity area, egg house and the passage system connecting both, the passage system is configured as entrance dynamic line and exit dynamic line are separated, to avoid the intersection interference of breeding goose entry path;Guiding control subsystem includes individual identification, egg production cycle prediction, path planning and execution control etc.Modes.The present application is based on egg production cycle prediction model, in the prediction egg production time window before actively opening gate and activating sound and light guide, guide breeding goose along the planned path enters egg house, realize from " passive waiting " to " active guiding " change.The present application reduces nest egg rate, reduces artificial intervention and breeding goose stress, realizes the accurate correlation of breeding egg and breeding goose individual, provides data support for breeding breeding goose, suitable for large-scale, intelligent breeding goose breeding.

Description

Technical Field

[0001] This invention relates to the field of intelligent management of poultry farming, specifically to a path guidance system and method for breeding geese in a semi-open farming setting. Background Technology

[0002] Goose breeding is an important part of the poultry farming industry. Currently, egg production management of breeding geese mainly relies on manual methods, which suffers from problems such as difficulty in identification, untimely monitoring, difficulty in traceability, low efficiency, and a tendency to induce stress in geese. To improve the automation and intelligence level of goose breeding, some explorations have been conducted on existing related technologies.

[0003] Chinese patent CN109952974B discloses an intelligent egg-laying recording system and method for geese. This system uses a signal collector installed in the laying box to read the RFID electronic leg bands worn by breeding geese to identify their identity and the time they enter and leave the laying room. This information, combined with the egg-laying sequence, enables egg-laying recording. While this method automates egg-laying recording to some extent, its core logic relies on the breeding geese actively entering the fixed laying box, making it a passive management approach. In semi-open farming scenarios, without active intervention, breeding geese are highly likely to lay eggs outside the laying box (i.e., out-of-nest eggs), leading to recording failures.

[0004] Chinese patent CN118278580B discloses a method and system for measuring the egg production performance of floor-raised geese. This method collects multi-dimensional data on the geese's identity, weight, movement, and vision, and then integrates this data to determine whether egg production has occurred. The method focuses on improving the accuracy of egg production event monitoring through multi-dimensional data, but it still falls within the scope of monitoring and does not involve active intervention or path guidance for the egg-laying behavior of geese.

[0005] In addition, most existing technologies, such as CN119302243B, are mostly applicable to closed cage rearing scenarios. Their area division and event triggering depend on fixed cage structures and cannot be directly applied to semi-open farming scenarios with blurred area boundaries, many environmental disturbances, and high freedom of movement of geese.

[0006] In summary, existing technologies have gaps in the active path guidance of breeding geese in semi-open farming scenarios. There is a lack of technical solutions that enable active intervention based on egg production prediction to achieve automated and orderly guidance of breeding geese from the rest area to the egg-laying room. Summary of the Invention

[0007] This invention provides a path guidance system and method for breeding geese in a semi-open breeding scenario, which achieves the following in a semi-open breeding scenario: (1) accurate identification of individual breeding geese; (2) active path guidance for breeding geese based on egg production cycle prediction; (3) designing separate entrance and exit routes for the egg production room to avoid path intersections; (4) automated isolation management of breeding geese; and (5) establishing a precise correspondence between hatching eggs and individual breeding geese.

[0008] The technical solution adopted in this invention is as follows: A path guidance system for breeding geese in a semi-open farming setting, comprising: A semi-open breeding farm includes a resting and activity area for the daily activities of breeding geese, an egg-laying room physically isolated from the resting and activity area, and a passage system connecting the resting and activity area and the egg-laying room; the passage system is configured to separate the entrance flow and the exit flow to avoid cross-interference between the entry and exit paths of the breeding geese; The guidance control subsystem includes: Individual identification module, used to identify and locate the breeding geese entering the channel system; The egg-laying cycle prediction module is used to predict the egg-laying time window of individual breeding geese based on historical egg-laying data and real-time monitoring data through a prediction model. The path planning module is used to calculate and plan the optimal path from the current location of the breeding geese to the entrance of the laying house based on the current location of the breeding geese and the predicted egg-laying time window. The execution control module includes automatic gate units located on the entrance path and the exit path, and guide indicator units arranged along the channel system; The execution control module responds to the trigger signal of the egg-laying cycle prediction module by opening the automatic gate unit of the corresponding entrance path before the predicted egg-laying time window and activating the guidance indicator unit to guide the breeding geese into the egg-laying room along the planned path.

[0009] In one implementation, the passage system includes an entrance route from the resting activity area to the egg-laying room, and an exit route from the egg-laying room back to the resting activity area, wherein the entrance route and the exit route are physically separated in space. Alternatively, the channel system can be a single physical channel, with the entrance and exit routes separated by time multiplexing. The guidance and control subsystem also includes a route configuration module, which controls the channel system to be configured as an entrance route or an exit route at different time periods according to a preset time scheduling strategy.

[0010] As one implementation method, the traffic flow configuration module includes: The timing scheduling unit is used to store and execute the timing schedule for traffic flow switching; The clearance detection unit is used to detect whether there are breeding geese left in the channel before the flow line is switched, and to perform the switch after confirming that the channel is cleared. The signage switching unit is used to control the direction signs of the automatic gate and the directional signs of the guidance indicator unit to match the current traffic flow configuration.

[0011] As one implementation, the guidance control subsystem further includes: The isolation management module includes an egg-laying monitoring unit and an automatic isolation unit. The egg-laying monitoring unit is used to monitor the egg-laying behavior of breeding geese entering the egg-laying room. After the egg-laying monitoring unit determines that egg-laying has been completed, the automatic isolation unit controls the automatic gate unit that closes the exit passage to temporarily isolate the egg-laying geese in the egg-laying room. The data management module includes at least a data storage unit and a data analysis unit, used to record egg production data and analyze and establish the association between hatching eggs and individual breeding geese during the period when the automatic isolation unit isolates egg-laying individuals.

[0012] In one implementation, the isolation management module also includes a release control unit, which, after completing the recording of egg production data, controls the automatic gate unit to open the exit path and activates the guide indicator unit of the exit path to guide the breeding geese back to the resting activity area along the exit path.

[0013] In one implementation, the individual identification module includes one or more combinations of an RFID identification unit, a visual identification unit, a sensor unit, and a multimodal fusion unit; wherein: The RFID identification unit is used to read the RFID leg band information worn by the breeding geese; The visual recognition unit is used to capture images through a camera and identify the breeding geese; The sensor unit uses one or more combinations of pressure sensors and infrared sensors to assist in identifying breeding geese; The multimodal fusion unit is used to fuse RFID information and visual recognition results to determine the individual identity and current location of the breeding geese.

[0014] As one implementation, the visual recognition unit is selected from any of the following recognition methods: direct visual recognition based on the physical characteristics of the breeding goose; or, recognition based on the identification worn by the breeding goose, wherein the identification includes one or more of the following: a colored collar, a shaped collar, and a QR code foot tag.

[0015] In one implementation, the multimodal fusion unit is configured with a fusion priority strategy, which sets the priority of RFID signals to be higher than that of visual recognition results; when the RFID signal is consistent with the visual recognition results, a confirmation result is output; when the RFID signal is clear but the vision is blocked, the RFID position is used as the reference, and the visual trajectory is inferred using Kalman filtering; when there is no RFID signal but the goose is visually recognized, it is marked as an abnormal or unidentified individual and an alarm is triggered.

[0016] In one implementation, the egg-laying cycle prediction module includes a data acquisition unit, a feature extraction unit, a cycle prediction unit, and a prediction correction unit; the cycle prediction unit uses a machine learning model to predict the egg-laying time window of an individual breeding goose based on the egg-laying cycle features.

[0017] As one implementation, the prediction model includes any one or more combinations of the following: Machine learning models, including one or more of XGBoost, LSTM, and Transformer; The expert experience rule model makes predictions based on fixed egg-laying cycle rules of breed, season, and age. Hybrid prediction models combine machine learning model outputs with expert experience rule outputs for comprehensive prediction. Among them, the XGBoost model is used to process non-time series features and output the egg-laying probability, while the LSTM model is used to process time series data and predict specific egg-laying time windows.

[0018] In one implementation, the path planning module includes a map building unit, a path calculation unit, a conflict detection unit, and a path optimization unit; The map building unit is used to construct a spatial raster map of a semi-open scene; The path calculation unit is used to calculate the optimal path based on the current location of the breeding goose and the entrance location of the laying house using the A* algorithm; The conflict detection unit is used to detect whether there are conflicts among multiple planned paths; The path optimization unit is used to optimize the path timing according to the prediction time window, and to optimize the path by adjusting the guidance timing or path when the conflict detection unit detects a conflict in the planned path.

[0019] In one implementation, the guidance instruction unit includes any one or more combinations of the following: Fixed sound and light guidance device, including one or more of LED light strips, buzzers, and voice broadcasters; A mobile robot guidance device is used to autonomously navigate to the front of the breeding geese and guide them to move along a planned path; The ground projection guidance device projects dynamic guidance light spots or light strips onto the floor of the passageway using a projector.

[0020] This invention also provides a method for guiding breeding geese in semi-open farming scenarios, based on the system described in any of the above, which includes the following steps: Step S1: Individual identification, using multimodal fusion to identify the individual geese entering the channel system and their current location; Step S2: Egg production cycle prediction. Based on historical egg production data and real-time monitoring data, the prediction model is used to predict the egg production time window of individual breeding geese. Step S3: Path planning. Based on the current location of the breeding geese and the predicted egg-laying time window, calculate and plan the optimal path from the current location to the entrance of the egg-laying room. Step S4: Path guidance. At a preset time T before the predicted egg-laying time window, the automatic gate of the entrance path is opened, and the sound and light guidance along the optimal path is activated to guide the breeding geese into the egg-laying room. Step S5: Egg production monitoring, using sensors installed in the egg-laying room to monitor the egg-laying behavior of the breeding geese; Step S6: Automatic isolation. After detecting the laying of hatching eggs, the automatic gate of the exit flow line is closed, the laying individual is temporarily isolated in the laying room, and the laying data is recorded to establish the association between hatching eggs and individual breeding geese. Step S7: Individual release. After confirming that egg laying is complete and data recording is finished, open the automatic gate of the exit path and activate the sound and light guidance of the exit path to guide the breeding geese back to the resting activity area along the exit path. Step S8: Data management, storing all time data, updating individual records, generating reports, and feeding them back to the prediction model.

[0021] As one implementation method, the multimodal fusion recognition in step S1 includes: using the YOLOv8 algorithm to detect breeding goose targets in real time from the camera video stream, and using the DeepSORT algorithm to perform multi-target tracking on the detected breeding goose targets; at the same time, reading the electronic tag information of the breeding goose's leg band through an RFID reader; and performing time-stamp-based association matching between the breeding goose's location information obtained from visual tracking and the identity information read from RFID to determine the individual identity and movement trajectory of the breeding goose.

[0022] As one implementation method, the path planning in step S3 includes: using the A* algorithm to calculate the shortest path from the current location of the breeding goose to the entrance of the laying house on the constructed grid map; when multiple planned paths are detected to conflict, a priority queuing strategy is adopted to allocate higher path usage priority to individuals with higher egg-laying probability and more urgent prediction time, and to achieve staggered guidance by controlling the opening sequence of different gates.

[0023] Compared with the prior art, the technical solution of the present invention has the following beneficial effects: 1) Achieving prediction-based proactive guidance: Unlike the passive waiting mode of existing technologies, this invention uses an egg-laying cycle prediction module to proactively open the gate and issue guidance before the egg-laying behavior of breeding geese occurs, "inviting" the breeding geese into the egg-laying room, fundamentally reducing the occurrence rate of eggs laid outside the nest.

[0024] 2) Constructing a complete closed loop of "monitoring-guidance-intervention": This invention not only includes egg production monitoring, but its core innovation lies in planning and physically intervening in the movement path of breeding geese based on prediction results, realizing closed-loop management from "discovering problems" to "guiding behavior".

[0025] 3) Adapted to semi-open aquaculture scenarios: In view of the characteristics of blurred regional boundaries and many interferences in semi-open scenarios, this invention designs physically separated entrance / exit routes and combines multimodal fusion recognition and dynamic path planning technology to effectively solve the problems of long-distance path planning, dynamic obstacle avoidance and route separation.

[0026] 4) Improve management accuracy and efficiency: Through multimodal fusion recognition, the individual identification accuracy rate can reach over 99%; through automated guidance, isolation and release processes, more than 80% of manual intervention can be reduced; at the same time, the system establishes a precise association between hatching eggs and individual breeding geese, providing reliable data support for subsequent breeding and selection.

[0027] 5) Reduce stress response in geese: Physically separated entrance and exit routes avoid cross-path collisions and collisions between geese entering and leaving the flock. The orderly guidance and isolation process also significantly reduces disturbance and stress to the flock. Attached Figure Description

[0028] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of the present invention and do not constitute a limitation on the technical solutions of the present invention.

[0029] Figure 1 This is a schematic diagram of the overall layout of the main functions of the goose breeding path guidance system described in this invention.

[0030] Figure 2 This is an algorithm diagram of the egg production cycle prediction module described in this invention.

[0031] Figure 3 This is a flowchart of the goose breeding path guidance method described in this invention.

[0032] Figure 4 This is an algorithm flowchart of the path planning module described in this invention.

[0033] Figure 5This is a schematic diagram of the multimodal fusion recognition logic in an embodiment of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0035] The key algorithms and terms involved in this invention are explained below: YOLOv8 (You Only Look Once v8): An existing, state-of-the-art real-time object detection algorithm. In this invention, it is used to quickly identify "breeding geese" targets and their location coordinates from camera video streams.

[0036] DeepSORT (Simple Online and Realtime Tracking with a Deep AssociationMetric): An existing multi-target tracking algorithm. In this invention, it is used to lock the movement trajectory of the same breeding goose between consecutive video frames, preventing identity loss due to geese occlusion.

[0037] XGBoost (eXtreme Gradient Boosting): An existing gradient boosting decision tree algorithm. This invention is used to process non-time series features (such as weight, season, breed) and output the egg production probability.

[0038] LSTM (Long Short-Term Memory): A special type of recurrent neural network (RNN). This invention is used to process time-series data (such as historical egg-laying intervals and behavioral trajectories) to predict specific egg-laying time windows.

[0039] A* Algorithm: An existing heuristic pathfinding algorithm. This invention is used to calculate the shortest, conflict-free path from the current location of the breeding goose to the laying house in a constructed grid map.

[0040] Bayesian Inference: Used for probabilistic fusion of multimodal data, updating the probability distribution of breeding goose identity and status based on the confidence levels of different sensors.

[0041] Kalman Filter: Used to denoise and estimate the state of sensor data (such as GPS / UWB / visual coordinates) and smooth the movement trajectory of breeding geese.

[0042] Example 1 This embodiment provides a path guidance system and method for breeding geese in a semi-open farming scenario, applied to a large-scale breeding goose farm with a scale of 1000 geese. This farm adopts a semi-open farming model, featuring a large outdoor activity area and enclosed laying rooms, allowing the breeding geese high freedom of movement but also subjecting them to numerous environmental disturbances.

[0043] 1. System Hardware Configuration and Layout The goose path guidance system in this embodiment includes a semi-open breeding farm and a guidance and control subsystem.

[0044] The semi-open farm includes: a 5,000-square-meter outdoor rest and activity area for the breeding geese to engage in daily activities and rest; a closed laying house physically isolated from the rest and activity area, which contains 100 independent laying nests; an independent feeding area; and a passageway system connecting the above areas.

[0045] The passage system has a specific traffic flow design, which includes a dedicated entrance traffic flow from the rest and activity area to the egg-laying room, and a dedicated exit traffic flow from the egg-laying room back to the rest and activity area.

[0046] The entrance and exit routes are physically separated in space to avoid path intersections. This is the key structural basis for the present invention to achieve orderly guidance and reduce stress.

[0047] like Figure 1 As shown, the guidance and control subsystem includes an individual identification module, an egg-laying cycle prediction module, a path planning module, an execution control module, an isolation management module, and a data management module.

[0048] In terms of hardware configuration, 134.2kHz RFID readers (reading distance ≥50cm) and 1080P high-definition cameras are deployed at key nodes of the channel system. The egg-laying nests are equipped with pressure sensors with a range of 0-50kg and infrared sensors. Movement control employs bidirectional automatic gates (opening speed <3s), and guidance and indication units consisting of LED light strips and buzzers are deployed along the channel system.

[0049] 2. Example of Operation Process This example uses three representative breeding geese for illustration: breeding goose A (ID:001) is predicted to lay eggs soon; breeding goose B (ID:002) is in its egg-laying cycle but currently shows no signs of laying eggs; and breeding goose C (ID:003) is an abnormal individual that is not wearing a leg band.

[0050] The following example uses the typical egg-laying period from 05:00 to 10:00 on a typical egg-laying day, combined with... Figure 3 The flowchart shown illustrates the system's workflow in detail.

[0051] Step S1: Individual Identification When the breeding geese enter the passage system, an RFID reader reads the electronic tag information in their leg bands, while a high-definition camera captures images. The visual recognition unit of the individual identification module uses the YOLOv8 algorithm to detect the breeding geese in real time from the images and employs the DeepSORT algorithm to perform multi-target tracking on the detected targets, locking the movement trajectory of the same breeding goose between consecutive video frames to prevent identification loss due to mutual occlusion. The pressure sensor and infrared sensor in the sensor unit can assist in identifying the location and behavior of the breeding geese. The multimodal fusion unit performs time-stamp-based association matching between the identification information read by RFID and the trajectory information of visual tracking to determine the individual identity and current location of the breeding goose.

[0052] In this embodiment, the multimodal fusion unit is configured with a fusion priority strategy, setting the priority of RFID signals higher than that of visual recognition results. When the RFID signal matches the visual recognition result, a confirmation result is output; when the RFID signal is clear but the vision is obstructed, the RFID position is used as the reference, and Kalman filtering is used to denoise and smooth the position data to infer the visual trajectory; when there is no RFID signal but a goose is visually identified (such as breeding goose C), it is marked as an abnormal or unidentified individual, and an alarm is triggered, without path guidance.

[0053] The specific implementation of "multimodal fusion" can employ various existing technologies. For example, when the multimodal fusion unit receives an RFID signal, it generates an identification point with a precise timestamp and location coordinates. The visual tracking module outputs a continuous trajectory (track line). The fusion strategy is as follows: when the Euclidean distance between the identification point and the trajectory line in space and time is less than a preset threshold (e.g., 0.5 meters, 1 second), the identity of the trajectory line is marked as the ID corresponding to that RFID. If the visual trajectory is occluded, based on the position of the previous identification point, a Kalman filter (with its state vector set to [x, y, vx, vy]^T and observation vector set to [x, y]^T) is used to predict the position within a short period of time until the visual track is recaptured.

[0054] Step S2: Egg Laying Cycle Prediction The data management module stores historical egg production data for each breeding goose (including egg-laying intervals, egg-laying time preferences, weight changes, etc.). The data acquisition unit of the egg-laying cycle prediction module accesses this historical data, and the feature extraction unit extracts egg-laying cycle features from it. The cycle prediction unit uses a hybrid model of XGBoost and LSTM: the XGBoost model is used to process non-time-series features (such as breed, season, weight) and outputs the probability of the current egg-laying period; the LSTM model is used to process time-series data (such as past N egg-laying intervals, recent behavioral trajectories) and predicts specific egg-laying time windows. The prediction correction unit dynamically corrects the prediction results based on real-time monitoring data (such as current location, activity level).

[0055] Training and application of egg production cycle prediction models can employ various existing technologies. For example, an LSTM model takes as input a sequence of time intervals between the past five egg production cycles, passes through an LSTM layer with 128 hidden units, and outputs a vector. An XGBoost model takes as input features such as breed (one-hot encoding), current age, season, and activity level in the last 24 hours, and outputs an egg production probability. Concatenating the vector output from the LSTM with the probability value output from the XGBoost model and inputting it into a fully connected layer ultimately predicts the egg production time window in hourly increments.

[0056] At 05:00 AM, based on historical data analysis, the system predicted that the egg-laying time window for breeding goose A (ID:001) would be 06:30-07:30, with an egg-laying probability >95%; the egg-laying probability for breeding goose B was only 30%, and it was not included in this guidance plan.

[0057] Step S3: Path Planning The map building unit of the path planning module pre-constructs a raster map (0.5m × 0.5m resolution) of the semi-open scene. When the system time is 06:15 (i.e., 15 minutes before the prediction time window), the egg-laying cycle prediction module issues a guidance command. The path calculation unit obtains the current location of breeding goose A (e.g., northeast corner of the rest area) determined by the individual identification module, and uses the A* algorithm to calculate the shortest path from that location to the entrance of the egg-laying room on the raster map. The conflict detection unit detects that there are no obstacles on the path and no other high-priority guidance tasks, and determines that there is no conflict. The path optimization unit confirms that the current path is available and optimizes the path timing.

[0058] When the predicted egg-laying windows of multiple breeding geese overlap, causing target path conflicts, the path optimization unit adopts a priority queuing strategy to allocate higher path usage priority to individuals with higher egg-laying probability and more urgent prediction time, and achieves staggered guidance by controlling the opening sequence of different gates.

[0059] As a concrete example, the conflict detection unit unfolds the planned path for each breeding goose to be guided in a spatiotemporal coordinate system. If two paths occupy the same grid at the same time point, it is determined as a "catch-up conflict"; if two paths are in opposite directions and occupy adjacent grids within the same time period, it is determined as a "heading conflict". The priority P is calculated as P = w1 * P_egg + w2 / T_urgency, where P_egg is the egg-laying probability, T_urgency is the time from the start of the prediction window, and w1 and w2 are weighting coefficients.

[0060] Step S4: Path Guidance The execution control module responds to guidance commands, opening the first automatic gate leading to the egg-laying room entrance and activating the guidance indicator unit along the planned path, such as illuminating a green LED arrow light strip and emitting a gentle buzzing sound, forming an audio-visual guidance system. Under conditioned reflex training, goose A moves in the direction of the cursor.

[0061] The safety protection unit of the execution control module detects obstacles in real time through light curtain sensors and pressure mats to prevent injuries from being pinched by the turnstiles. If a temporary obstacle (such as a zookeeper's vehicle or a flock of geese blocking the path) is detected, the path planning module triggers local replanning to calculate an alternative route.

[0062] Step S5: Egg production monitoring After breeding goose A successfully enters the laying room, the automatic gate at the entrance closes. The egg-laying monitoring unit detects weight changes (a slow increase of about 120g) inside the nest using a pressure sensor, detects slight increases in local temperature using an infrared sensor, and combines visual recognition to confirm egg-laying behaviors such as "standing still after leaving the nest," thus comprehensively determining that egg-laying has occurred.

[0063] Step S6: Automatic Isolation After the hatching eggs are detected, the automatic isolation unit immediately locks the automatic gate of the exit path, temporarily isolating the breeding goose A in the egg-laying room.

[0064] Meanwhile, the data storage unit of the data management module records egg-laying data such as egg-laying time, egg weight, and egg photos. The data association unit accurately binds this data with the breeding goose ID:001, establishing a relationship between the breeding eggs and the individual breeding goose.

[0065] Step S7: Individual Release After confirming that egg laying is complete and data recording is complete, the release control unit issues a command to open the automatic gate of the exit path and activate the guide indicator unit of the exit path (such as blue light or different tones) to guide the breeding goose A back to the resting activity area along the dedicated exit path.

[0066] Because the entrance and exit routes are physically separated, the return path of breeding goose A completely avoids other breeding geese that may be heading to the laying house, effectively reducing group stress.

[0067] Step S8: Data Management After the entire egg-laying period ends, the data management module stores all time data in the database, updates the individual egg-laying record of breeding goose A, and generates reports containing indicators such as successful induction rate, out-of-nest egg rate, and egg-laying efficiency.

[0068] The data analysis unit optimizes and adjusts the parameters of the prediction model based on the deviation between the actual egg-laying time (06:50) and the prediction window (06:30-07:30) (such as fine-tuning the LSTM model parameters through backpropagation), and feeds this information back to the egg-laying cycle prediction module to improve the accuracy of future predictions.

[0069] At the same time, the system will record the abnormal information of breeding goose C and issue an alarm, prompting manual intervention.

[0070] 3. Examples of Exception Handling and Dynamic Adjustment During the process of guiding breeding goose A, breeding goose B accidentally approached the entrance path. The RFID reader identified its ID: 002. The system checked the prediction database and found that its egg-laying probability was only 30% and it was not in the guidance plan. Therefore, the system refused to open the gate at that location and changed the frequency of the sound and light guide (such as red light or rapid tone) to indicate "this road is closed", effectively preventing non-egg-laying individuals from accidentally entering and occupying resources.

[0071] For the abnormal individual goose C that was not wearing a leg band, the vision unit detected its presence in the rest area but there was no RFID signal. The multimodal fusion unit marked it as an "unidentified individual", did not plan a path for it, and popped up an alarm on the central control screen to prompt the zookeeper to intervene.

[0072] After adopting the system and method of this embodiment and testing for a complete egg-laying cycle (approximately 6 months), compared with the manual management mode of the same period last year, the farm's various indicators have been significantly improved: the individual identification accuracy rate reached over 99%, the active arrival rate of the egg-laying room reached over 95%, the hatching egg collection rate reached over 98%, the labor cost of the egg-laying room manager was reduced by over 80%, and due to orderly guidance and separation of movement lines, the stress-related behaviors of geese during the egg-laying period were reduced, and the overall egg-laying efficiency was improved by over 15%.

[0073] Example 2 This embodiment is basically the same as embodiment 1, except that the visual recognition unit of the individual recognition module adopts a different recognition scheme.

[0074] In this embodiment, the breeding geese do not wear RFID leg bands, but rather identification collars of different colors or shapes. These collars are made of weather-resistant materials and have bright colors (such as red, yellow, blue, and green) or unique geometric shapes (such as triangles, circles, and squares). Different combinations of colors or shapes correspond to different individual breeding goose identities. The cameras are color high-definition cameras deployed at key nodes of the channel system.

[0075] The workflow of the individual identification module is as follows: When the breeding geese enter the passage system, a color camera captures an image of the goose, including its collar. The visual recognition unit analyzes the captured image using a deep learning-based image classification or object detection algorithm (such as YOLOv8) to identify the collar's color and / or shape features. The system has a built-in mapping table between color-shape combinations and individual goose IDs; the identified collar features uniquely identify the goose.

[0076] Meanwhile, the current location of the breeding geese is determined by using stereo vision or monocular vision positioning algorithms from multi-view cameras.

[0077] This embodiment eliminates the need for breeding geese to wear RFID electronic leg bands, avoiding potential leg injuries or dislodgement issues. For certain breeds of breeding geese or stages of rearing where leg bands are unsuitable, this solution provides a less stressful alternative identification method. Using colored collars combined with visual recognition, the identification accuracy can reach over 95%, and the system deployment cost is relatively low, making it particularly suitable for cost-sensitive small and medium-sized farms.

[0078] Example 3 This embodiment is basically the same as embodiment 1, except that the cycle prediction unit of the egg production cycle prediction module uses a different prediction model.

[0079] In this embodiment, the periodic prediction unit uses the Transformer time series prediction model instead of the hybrid model of XGBoost and LSTM. The Transformer model is based on the self-attention mechanism, which can effectively capture long-distance dependencies in time series data and has good adaptability for predicting the behavior of breeding geese with complex periodicity (such as seasonal egg production and large fluctuations in individual egg production intervals).

[0080] Specifically, the data acquisition unit collects historical egg-laying time series data for each breeding goose, such as the time of each egg-laying session, the interval between egg-laying sessions, and the duration of egg-laying sessions over the past 6 months; the feature extraction unit encodes the above data into a format suitable for input to the Transformer model, including positional encoding and temporal feature encoding; the cycle prediction unit fine-tunes the pre-trained Transformer model or retrains it based on historical data from the farm to predict the next egg-laying time window for individual breeding geese.

[0081] Compared to the hybrid model in Example 1, this embodiment demonstrates higher prediction accuracy when dealing with egg-laying behavior that is highly periodic and has long-term dependence, and shows better adaptability, especially in scenarios where the egg-laying cycle changes suddenly.

[0082] Experimental data show that, using the Transformer model, under the same farm dataset and test conditions, the accuracy of predicting the egg-laying time window can be improved to over 90%, an improvement of 8%–12% compared to the hybrid model described in Example 1; the average prediction error can be controlled within 1.5 hours, a reduction of 20%–30% compared to the original model; for abnormal samples with fluctuations in the egg-laying cycle, the prediction stability is improved by over 25%, and the false negative rate is reduced by over 15%. Based on more accurate egg-laying time window prediction, the success rate of guiding breeding geese to lay eggs can be further improved to over 96%, and the rate of eggs laid outside the nest can be reduced by another 10%–20%, thereby further improving the efficiency of egg-laying management and the effect of hatching egg collection.

[0083] This embodiment can more fully explore the long-term temporal correlation information in the historical egg-laying behavior of breeding geese, improve the prediction accuracy and robustness of individual egg-laying time windows, and is particularly suitable for large-scale breeding goose farming scenarios with large fluctuations in egg-laying cycles, strong influence of environmental factors, or sufficient historical data.

[0084] Example 4 This embodiment is basically the same as Embodiment 1, except that the cycle prediction unit of the egg production cycle prediction module adopts a comprehensive prediction scheme that combines machine learning models with expert experience rules.

[0085] In this embodiment, the periodic prediction unit includes two parallel prediction channels: (1) Data-driven channel: The XGBoost+LSTM hybrid model described in Example 1 or the Transformer model described in Example 3 is used to generate preliminary prediction results P_data based on historical egg production data.

[0086] (2) Knowledge-driven channel: The rule prediction result P_rule is generated based on the expert experience rule base. The expert experience rule base contains fixed egg-laying cycle rules for different breeds, seasons, and ages, such as "the average egg-laying interval of Zhejiang White Geese in spring is 28-32 hours" and "the egg-laying cycle of breeding geese over 3 years old is extended by 15%". These rules are based on long-term breeding experience and are stored in the rule base in the form of conditional statements.

[0087] The cycle prediction unit also includes a fusion decision subunit for weighted fusion of P_data and P_rule. The fusion weights can be dynamically adjusted according to the actual application scenario: for individuals with sufficient data, the weight of P_data is increased; for newly introduced varieties or individuals with sparse data, the weight of P_rule is increased. The fusion formula is: P_final = α * P_data + (1-α) * P_rule, where α is a dynamic adjustment factor.

[0088] This embodiment combines the adaptability of data-driven approaches with the stability of knowledge-driven approaches, making it particularly suitable for the initial application stage of farms with mixed breeds and insufficient data accumulation. When encountering scenarios where historical data is not covered, such as extreme weather or feed changes, expert experience rules can provide reliable baseline predictions, preventing model failure. Practical application shows that this solution improves the robustness of the prediction model by more than 30%, significantly reducing misleading predictions caused by bias.

[0089] Example 5 This embodiment is basically the same as embodiment 1, except that the guidance instruction unit of the execution control module adopts a different guidance method.

[0090] In this embodiment, in addition to the fixed audio-visual guidance device, the guidance and indication unit also includes a movable robot guidance device. The robot guidance device is an autonomous mobile robot equipped with a high-precision positioning module (such as SLAM lidar), a path tracking controller, and various guidance actuators (such as a bionic sound source, an adjustable spotlight, a food reward dispenser, etc.).

[0091] Once the execution control module receives the guidance command, the path planning module simultaneously sends the optimal path information to both the fixed guidance device and the robotic guidance device. The robotic guidance device autonomously navigates to the current position of the breeding goose, then moves at an appropriate speed about 1-2 meters ahead of the goose along the planned path, emitting the mother goose's call through a bionic sound source and illuminating the direction of movement with a spotlight to guide the goose to follow. For individuals that are slow to react or sensitive to environmental changes, the robot can also provide a small amount of food as a reward to enhance the guidance effect. After the breeding goose enters the laying room, the robot automatically returns to the charging station to stand by.

[0092] This embodiment is particularly suitable for breeding scenarios with complex site layouts, severe obstructions, or high environmental noise. The mobile robot can provide personalized "one-on-one" guidance, overcoming the limitations of fixed guidance devices due to field of vision and distance. For timid breeding geese with strong stress responses, the robot's guidance method is gentler, increasing the success rate to over 98%, while further reducing the stress level of the breeding geese.

[0093] Example 6 This embodiment is basically the same as embodiment 1, except that the guidance indication unit of the execution control module uses ground projection path indication instead of traditional LED light strip.

[0094] In this embodiment, the guidance and instruction unit includes a projector array (such as a short-throw laser projector) deployed above the passageway system and a corresponding control system. When the execution control module receives a guidance command, the optimal path generated by the path planning module is converted into a series of continuous path point coordinates. Based on these coordinates, the control system controls the projector array to dynamically project bright guidance spots or continuous light bands onto the passageway floor.

[0095] Ground projection has the following characteristics: (1) High flexibility: the projection path can be changed in real time according to the planning results, without the need to lay physical light strips in advance; (2) Rich information: it can project various information such as arrows, numbers, and color changes to realize more complex guidance instructions (such as "forward", "turn left", "stop"); (3) Strong adaptability: for temporary changes in the movement or scenarios that require dynamic obstacle avoidance, the projection path can be updated in real time to guide the breeding geese to detour.

[0096] This embodiment offers significant advantages in scenarios requiring frequent adjustments to traffic flow or responses to unforeseen events. For example, when a section of the passageway needs temporary repair, the system can immediately adjust the projection path, guiding the geese to a backup passageway without requiring physical modifications. Compared to fixed LED light strips, the initial investment for the ground projection solution is slightly higher, but its operational flexibility is greatly improved, making it particularly suitable for research-oriented farms or scenarios requiring frequent traffic flow optimization experiments.

[0097] Example 7 This embodiment is basically the same as Embodiment 1, except that the circulation design of the passage system adopts a different scheme. The passage system is a single physical passage, and the entrance circulation path and exit circulation path are separated through time multiplexing. The guidance and control subsystem further includes a circulation configuration module, which controls the passage system to be configured as an entrance circulation path or an exit circulation path at different time periods according to a preset time scheduling strategy.

[0098] As a specific example, the circulation configuration module includes: The timing scheduling unit is used to store and execute the timing schedule for traffic flow switching; The clearance detection unit is used to detect whether there are breeding geese left in the channel before the flow line is switched, and to perform the switch after confirming that the channel is cleared. The signage switching unit is used to control the direction signs of the automatic gate and the directional signs of the guidance indicator unit to match the current traffic flow configuration.

[0099] In this embodiment, due to site limitations (such as insufficient space for farm renovation), physical separation of the entrance and exit traffic routes is not possible. Therefore, the system adopts a time-reuse scheme for the same passageway. Specifically, the passageway system contains only one physical passageway connecting the resting activity area and the laying room, but this passageway is configured as different functional traffic routes at different times.

[0100] The system achieves time reuse in the following ways: (1) Automatic gates with switchable signs are set at both ends of the channel, and the gates are equipped with LED displays that can change colors or patterns; (2) Ground projections or LED light strips with changeable direction indicators are laid along the channel; (3) The system clock and the egg production cycle prediction module work together. For example, during the peak egg production period from early morning to morning (04:00-11:00), the system configures the channel as an "entrance flow" mode: the gate at the entrance is open, the gate at the exit is closed, and all the guide signs point to the direction of the egg production room; during the non-peak egg production period (11:00-04:00 the next day), the system configures the channel as an "exit flow" mode: the gate at the exit is open, the gate at the entrance is closed, and all the guide signs point to the direction of the rest and activity area.

[0101] To avoid conflicts in entry and exit, the system adopts a strict time-series isolation strategy: (1) During the switching of the movement mode (such as around 11:00), a 15-30 minute clearing buffer period is set to ensure that there are no breeding geese left in the passage; (2) For individual abnormal situations (such as brooding geese) that need to enter the laying room during the non-laying period, the system marks them as special tasks, and temporarily switches the mode and guides them separately after confirming that the passage is empty.

[0102] This embodiment provides a compact solution for space-constrained scenarios. Although it sacrifices the parallel processing capacity for guiding multiple geese in and out simultaneously, it still meets the needs of medium-sized farms through precise timing control and predictive guidance. In practical applications, by adopting a time-reuse scheme, the original 3-meter-wide dual-channel layout can be compressed into a 1.5-meter-wide single channel, saving 50% of the channel floor space and providing a feasible path for automated transformation of farms with limited space. Because the entry and exit times are strictly staggered, this solution fundamentally eliminates the intersection of paths for geese entering and exiting, thus also achieving the beneficial effect of reducing stress.

[0103] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and changes to the form and details of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection of this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A path guidance system for breeding geese in a semi-open farming setting, characterized in that, include: A semi-open breeding farm includes a resting area for the daily activities of breeding geese, an egg-laying room physically isolated from the resting area, and a passageway system connecting the resting area and the egg-laying room; The passage system is configured with separate entrance and exit routes to avoid cross-interference between the geese's entry and exit paths; The guidance control subsystem includes: Individual identification module, used to identify and locate the breeding geese entering the channel system; The egg-laying cycle prediction module is used to predict the egg-laying time window of individual breeding geese based on historical egg-laying data and real-time monitoring data through a prediction model. The path planning module is used to calculate and plan the optimal path from the current location of the breeding geese to the entrance of the laying house based on the current location of the breeding geese and the predicted egg-laying time window. The execution control module includes automatic gate units located on the entrance path and the exit path, and guide indicator units arranged along the channel system; The execution control module responds to the trigger signal of the egg-laying cycle prediction module by opening the automatic gate unit of the corresponding entrance path before the predicted egg-laying time window and activating the guidance indicator unit to guide the breeding geese into the egg-laying room along the planned path.

2. The goose path guidance system according to claim 1, characterized in that, The passage system includes an entrance route from the rest and activity area to the egg-laying room, and an exit route from the egg-laying room back to the rest and activity area, wherein the entrance route and the exit route are physically separated in space. Alternatively, the channel system is a single physical channel, and the entrance flow and exit flow are separated by time multiplexing; the guidance and control subsystem also includes a flow configuration module, which is used to control the channel system to be configured as an entrance flow or an exit flow at different time periods according to a preset time scheduling strategy; Optionally, the traffic flow configuration module includes: The timing scheduling unit is used to store and execute the timing schedule for traffic flow switching; The clearance detection unit is used to detect whether there are breeding geese left in the channel before the flow line is switched, and to perform the switch after confirming that the channel is cleared. The signage switching unit is used to control the direction signs of the automatic gate and the pointing signs of the guidance indicator unit to match them with the current traffic flow configuration; Optionally, the guidance control subsystem further includes: The isolation management module includes an egg-laying monitoring unit and an automatic isolation unit. The egg-laying monitoring unit is used to monitor the egg-laying behavior of breeding geese entering the egg-laying room. After the egg-laying monitoring unit determines that egg-laying has been completed, the automatic isolation unit controls the automatic gate unit that closes the exit passage to temporarily isolate the egg-laying geese in the egg-laying room. The data management module includes at least a data storage unit and a data analysis unit, used to record egg production data and analyze and establish the association between hatching eggs and individual breeding geese during the period when the automatic isolation unit isolates egg-laying individuals.

3. The goose path guidance system according to claim 2, characterized in that, The isolation management module also includes a release control unit, which, after completing the recording of egg production data, controls the automatic gate unit to open the exit path and activates the guide indicator unit of the exit path to guide the breeding geese back to the resting activity area along the exit path.

4. The goose path guidance system according to claim 1, characterized in that, The individual identification module includes one or more combinations of an RFID identification unit, a visual identification unit, a sensor unit, and a multimodal fusion unit; wherein: The RFID identification unit is used to read the RFID leg band information worn by the breeding geese; The visual recognition unit is used to capture images through a camera and identify the breeding geese; The sensor unit uses one or more combinations of pressure sensors and infrared sensors to assist in identifying breeding geese; The multimodal fusion unit is used to fuse RFID information and visual recognition results to determine the individual identity and current location of the breeding geese; Optionally, the visual recognition unit is selected from any of the following recognition methods: direct visual recognition based on the physical characteristics of the breeding goose; or, recognition based on the identification worn by the breeding goose, wherein the identification includes one or more of the following: a colored collar, a shaped collar, and a QR code foot tag.

5. The goose path guidance system according to claim 4, characterized in that, The multimodal fusion unit is configured with a fusion priority strategy, which sets the priority of RFID signals to be higher than that of visual recognition results; when the RFID signal and the visual recognition result are consistent, a confirmation result is output. When the RFID signal is clear but the vision is obstructed, the RFID position is used as the reference, and the visual trajectory is inferred using Kalman filtering. When a goose is visually identified but there is no RFID signal, it is marked as an abnormal or unidentified individual, and an alarm is triggered.

6. The goose path guidance system according to claim 1, characterized in that, The egg-laying cycle prediction module includes a data acquisition unit, a feature extraction unit, a cycle prediction unit, and a prediction correction unit; the cycle prediction unit uses a machine learning model to predict the egg-laying time window of an individual breeding goose based on the egg-laying cycle features. Optionally, the prediction model includes any one or more combinations of the following: Machine learning models, including one or more of XGBoost, LSTM, and Transformer; The expert experience rule model makes predictions based on fixed egg-laying cycle rules of breed, season, and age. Hybrid prediction models combine machine learning model outputs with expert experience rule outputs for comprehensive prediction. Among them, the XGBoost model is used to process non-time series features and output the egg-laying probability, while the LSTM model is used to process time series data and predict specific egg-laying time windows.

7. The goose path guidance system according to claim 1, characterized in that, The path planning module includes a map building unit, a path calculation unit, a conflict detection unit, and a path optimization unit; The map building unit is used to construct a spatial raster map of a semi-open scene; The path calculation unit is used to calculate the optimal path based on the current location of the breeding goose and the entrance location of the laying house using the A* algorithm; The conflict detection unit is used to detect whether there are conflicts among multiple planned paths; The path optimization unit is used to optimize the path timing according to the prediction time window, and to optimize the path by adjusting the guidance timing or path when the conflict detection unit detects a conflict in the planned path. Optionally, the guidance indication unit includes any one or more combinations of the following: Fixed sound and light guidance device, including one or more of LED light strips, buzzers, and voice broadcasters; A mobile robot guidance device is used to autonomously navigate to the front of the breeding geese and guide them to move along a planned path; The ground projection guidance device projects dynamic guidance light spots or light strips onto the floor of the passageway using a projector.

8. A method for guiding breeding geese in a semi-open farming scenario, based on the system described in any one of claims 1 to 7, characterized in that, Includes the following steps: Step S1: Individual identification, using multimodal fusion to identify the individual geese entering the channel system and their current location; Step S2: Egg production cycle prediction. Based on historical egg production data and real-time monitoring data, the prediction model is used to predict the egg production time window of individual breeding geese. Step S3: Path planning. Based on the current location of the breeding geese and the predicted egg-laying time window, calculate and plan the optimal path from the current location to the entrance of the egg-laying room. Step S4: Path guidance. At a preset time T before the predicted egg-laying time window, the automatic gate of the entrance path is opened, and the sound and light guidance along the optimal path is activated to guide the breeding geese into the egg-laying room. Step S5: Egg production monitoring, using sensors installed in the egg-laying room to monitor the egg-laying behavior of the breeding geese; Step S6: Automatic isolation. After detecting the laying of hatching eggs, the automatic gate of the exit flow line is closed, the laying individual is temporarily isolated in the laying room, and the laying data is recorded to establish the association between hatching eggs and individual breeding geese. Step S7: Individual release. After confirming that egg laying is complete and data recording is finished, open the automatic gate of the exit path and activate the sound and light guidance of the exit path to guide the breeding geese back to the resting activity area along the exit path. Step S8: Data management, storing all time data, updating individual records, generating reports, and feeding them back to the prediction model.

9. The method for guiding breeding geese according to claim 8, characterized in that, The multimodal fusion recognition in step S1 includes: using the YOLOv8 algorithm to detect breeding goose targets in real time from the camera video stream, and using the DeepSORT algorithm to perform multi-target tracking on the detected breeding goose targets; at the same time, reading the electronic tag information of the breeding goose's leg band through an RFID reader; and performing time-stamp-based association matching between the breeding goose's location information obtained from visual tracking and the identity information read from RFID to determine the individual identity and movement trajectory of the breeding goose.

10. The method for guiding breeding geese according to claim 8, characterized in that, The path planning in step S3 includes: using the A* algorithm to calculate the shortest path from the current location of the breeding goose to the entrance of the laying house on the constructed grid map; when multiple planned paths are detected to conflict, a priority queuing strategy is adopted to allocate higher path usage priority to individuals with higher egg-laying probability and more urgent prediction time, and to achieve staggered guidance by controlling the opening sequence of different gates.

Images