Intelligent dirt detection and feather cleaning method and system for caged waterfowls

Abstract

The invention discloses an intelligent dirt detection and feather cleaning method and system for caged waterfowls, and relates to the field of waterfowl breeding, and the method comprises the steps: moving a skip car with a core processing module, a high-definition camera and a nozzle, and obtaining a two-dimensional code label and waterfowl data in a breeding cage through the high-definition camera; removing the breeding cage net in the image through the core processing module based on a defense algorithm, and recovering the shielded part; an improved MDA-DepLabV3 + model is used for detecting the pollution degree of the two-dimensional code labels and the waterfowl feathers, and a waterfowl area and a stain area are segmented to obtain a stain area proportion; a cleaning instruction is obtained by comparing the area proportion of the stains with a set threshold value through a core processing module, and a telescopic rod is started according to the cleaning instruction to move a spray head to the top of a breeding cage; and executing the corresponding cleaning instruction by using the nozzle. The method solves the problems that an existing method easily causes waterfowl stress, lacks an intelligent recognition algorithm for detecting the waterfowl feather pollution degree and lacks intelligent regulation and control.

Description

Technical Field

[0001] The invention relates to the field of waterfowl breeding, and in particular to an intelligent dirt detection and feather cleaning method and system for caged waterfowl. Background Art

[0002] At present, there are two main modes of waterfowl farming: caged and non-caged. Compared with the non-caged mode, the caged waterfowl model can help breeders better control the feed intake and feeding time of waterfowl, thereby ensuring that waterfowl get proper nutrition and effectively managing feed costs. In addition, the caged model is easy to clean the farm, which helps to improve the breeding environment, avoid water pollution, strengthen disease prevention and control, and improve animal welfare. In recent years, the protection of basic farmland has been strengthened, and the traditional breeding industry has been affected. The traditional breeding model of many farmers, especially the ground flat breeding model with large land area and low land use efficiency, has gradually turned to the high-density and higher land use efficiency stacked cage model.

[0003] The existing large-scale flat-farming model mainly relies on manual or simple spraying for cleaning waterfowl wings. Regular spraying of water in traditional waterfowl farming consumes a lot of manpower and may also pollute natural water resources such as rivers. It also requires regular manual operations and is difficult to adapt to welfare cage models for waterfowl of different specifications. With the development of waterfowl farming, the scale of farms is getting higher and higher, but the degree of automation is low. While improving farming efficiency, it becomes particularly important to improve the farming environment and enhance animal welfare.

[0004] Waterfowl love water by nature. Although the traditional cage model has many advantages in feeding and management, it brings challenges to waterfowl wing cleaning. In the cage model, due to space limitations and the particularity of waterfowl feathers, the existing cleaning methods are difficult to adapt to the cleaning needs of waterfowl of different sizes, thus affecting the welfare and health of waterfowl. In particular, traditional wing cleaning methods mostly rely on manual or simple spraying, which is not only inefficient, but also easily leads to waste of water resources and pollution. Therefore, how to improve the wing cleaning effect in the cage model without affecting the welfare of waterfowl has become an urgent problem to be solved.

[0005] In order to solve the above problems, researchers have designed automatic feather cleaning methods and developed corresponding devices, but there are still some problems. The methods for feather cleaning of waterfowl are different for different breeding modes (floor breeding or cage breeding).

[0006] CN209234668U discloses an automatic shower suitable for flat-raised waterfowl, which allows waterfowl to shower freely and automatically shuts down after leaving, thereby improving the ability of waterfowl to adapt to dry-raised environments and improving production performance such as down quality. However, this method requires breeders to drive waterfowl into the shower room, which is likely to cause stress reactions in waterfowl. CN112655592A discloses a cage system for laying ducks without a water surface combined with intermittent spraying, which can realize regular waterfowl feather cleaning in cage mode, but lacks an intelligent recognition algorithm to detect whether waterfowl feathers need to be cleaned, and also lacks an intelligent control system to dynamically adjust the atomization intensity, time and water temperature according to the individual status of waterfowl. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a caged waterfowl intelligent pollution inspection and feather cleaning method and system, which can dynamically adjust the atomization intensity, spraying time and water temperature according to the individual status of the waterfowl, thereby minimizing the waste of water resources and pollution emissions during the cleaning process, while meeting the physiological needs of the waterfowl, ensuring the stability of their production performance and reproductive performance, comprehensively improving animal welfare, and realizing efficient utilization of resources.

[0008] In order to achieve the above-mentioned purpose of the invention, the technical solution adopted by the present invention is: a method for intelligent pollution inspection and feather cleaning of caged waterfowl, comprising the following steps: S1: Move the feed cart with core processing module, HD camera and sprinkler to obtain QR code labels and waterfowl data in breeding cages through HD camera; S2: The core processing module removes the cage net in the image based on the defense algorithm and restores the blocked part; S3: Based on the restored image, the improved MDA-DepLabV3+ model is used to detect the pollution degree of the QR code label and waterfowl feathers, and the stain area ratio is obtained by segmenting the waterfowl area and the stain area; S4: The core processing module compares the proportion of the stain area with the set threshold to obtain a cleaning instruction, and activates the telescopic rod to move the nozzle to the top of the breeding cage according to the cleaning instruction; S5: Use the nozzle to execute the corresponding cleaning instructions, and use the flow sensor and temperature sensor equipped at the nozzle to monitor the relevant data in real time, and feed back the relevant data to the core processing module; S6: Determine whether the feed cart has moved to the end of the breeding cage. If yes, proceed to step S7; otherwise, return to step S1 and continue to move the feed cart. S7: Start the manure-water collection conveyor belt to remove the waterfowl manure, and use the manure-water separation device to dehydrate the collected wet manure and then transfer it to the fermentation tank for fermentation treatment to convert it into organic fertilizer; S8: Use the intelligent detection of waterfowl feather stains and the adaptive clean feather management platform to uniformly manage and display the farms.

[0009] Furthermore, the core processing module in S1 adopts a parallel scheduling mechanism of multithreading and thread pool as a control strategy; The parallel scheduling mechanism of the multi-threading and thread pool is specifically as follows: the core processing module allocates a control thread to each high-definition camera and the corresponding nozzle control unit, including a camera thread and a nozzle control thread, and the camera thread and the nozzle control thread are uniformly managed by the thread pool built into the core processing module.

[0010] Furthermore, the defense algorithm in S2 performs the following operations: S21: Perform semantic segmentation on the original image and obtain the semantic map through the Unet network. The formula is:

[0011] in, To obtain the semantic graph, For the Unet network, is the original image; S22: Perform bitwise operation on the original image according to the semantic map, and remove the part marked as the breeding cage net area in the semantic map from the original image by performing mask operation, so as to obtain the image after the breeding cage net is removed. The formula is:

[0012] in, To remove the image after the breeding cage net, and is a pixel; S23: Based on the image after removing the breeding cage net, the Pix2pixHD network is used to restore the image after removing the breeding cage net to the real image and restore the occluded part. The formula is:

[0013] in, For the restored image, For Pix2pixHD Network.

[0014] Furthermore, S3 includes the following sub-steps: S31: Based on the restored image, the improved MDA-DepLabV3+ model is used to identify the QR code label, and the Pyzbar library is used to decode it and extract the breeding cage number; S32: Use the improved MDA-DepLabV3+ model to segment the waterfowl area and the stain area, and obtain the stain area ratio. The formula is:

[0015] in, is the stain area ratio, is the area of ​​the stain, is the total area of ​​waterfowl areas.

[0016] Furthermore, the improved MDA-DepLabV3+ model uses the deep separable convolution DSConv to replace the traditional convolution Conv, uses the MobileNetV2 network to replace the original backbone network, and introduces the dynamic void spatial pyramid pool ASPP module.

[0017] Furthermore, the waterfowl feather stain intelligent detection and adaptive feather cleaning management platform in S8 includes: Waterfowl feather stain cleaning report: The waterfowl feather stain cleaning report is used to record the cleaning operation information of each waterfowl feather, including waterfowl code, cleaning time, pollution degree, breeding cage number, cleaning water flow, spraying time, spraying water temperature, water usage, energy consumption, cleaning effect evaluation and related pictures generated during the cleaning process; Group waterfowl cleaning performance report: The group waterfowl cleaning performance report is used to display the cleaning effect of waterfowl groups, including cleaning time distribution, average water flow, cleaning time and group water resource usage, and provides cleaning efficiency and effect trend analysis to help optimize group cleaning solutions; User feedback report: The user feedback report is used to upload feedback information through the mini program, including information about incomplete feather cleaning and equipment failure. The platform automatically generates a feedback report to help administrators deal with problems and improve system service quality; Equipment alarm table: The equipment alarm table is used to monitor the equipment status in real time, automatically generate equipment fault alarm table, and remind the administrator to repair in time, including equipment type, fault description and occurrence time; Operation management module: The operation management module is used to provide operation management functions, including threshold management, equipment management, farm management and waterfowl file management; Report and visualization display platform: The report and visualization display platform has the function of generating various reports, supports detailed statistics and analysis of cleaning data, equipment status, and warning information, and provides a variety of visualization display methods.

[0018] The technical solution also adopted by the present invention is: a system for intelligent pollution inspection and feather cleaning method of caged waterfowl, the system comprising a plurality of breeding cages of the same structure, and the plurality of breeding cages are connected in parallel, the breeding cages are provided with a drinking water pipe, a QR code label, a feed trough, an egg collecting trough, a manure collecting conveyor belt and a roller, the drinking water pipe is located inside the breeding cage, the feed trough is located outside the breeding cage, the QR code label is attached to the feed trough, the egg collecting trough is located below the feed trough, the manure collecting conveyor belt is located below each breeding cage, and the roller is located at the tail of the manure collecting conveyor belt; The system also includes a feed cart, which is located directly in front of the feed trough of the multi-layer breeding cage. A core processing module is mounted in the middle of the feed cart. The core processing module is connected to multiple high-definition cameras and multiple nozzles, and the nozzles are connected to the feed cart through telescopic rods.

[0019] Furthermore, the telescopic rod is driven by an electric push rod or a pneumatic cylinder, and a water quality filtration unit is integrated inside the nozzle, including a replaceable filter element and a water pressure stabilizer.

[0020] Furthermore, both sides of the manure-water collecting conveyor belt are rolled up to form a sealed enclosure structure, and the manure-water collecting conveyor belt is driven by a motor for transportation. The end of the manure-water collecting conveyor belt is equipped with a manure scraper and an adjustable roller, which can press out excess water in the wet manure.

[0021] The beneficial effects of the present invention are: (1) Multi-layer caged waterfowl feather stain intelligent detection and adaptive feather cleaning device: This includes the use of core processing modules, QR code labels, waterfowl feather intelligent detection devices, adaptive feather cleaning devices, feed carts, manure treatment and other multi-device collaborative deployment strategies to ensure accuracy and efficiency in the process of caged waterfowl feather stain detection and cleaning.

[0022] (2) Retractable and anti-clogging nozzle design: The nozzle is connected to the telescopic rod, which can control the forward and backward movement of the nozzle. The nozzle moves left and right with the material cart, which can make the cleaning more comprehensive and save costs. The nozzle is equipped with a water quality filtration system to effectively prevent scale and impurities from depositing, avoid nozzle clogging, and improve the stability and service life of the equipment.

[0023] (3) Data monitoring and feedback: The nozzle is equipped with a flow sensor and a temperature sensor, which enables the system to record water flow and temperature.

[0024] (4) Modular design: This system adopts a modular design, which can flexibly adjust the camera and nozzle configuration according to the actual scale and environmental requirements of the farm. Through a core processing module, combined with multi-threading and thread pool technology, it can realize the simultaneous control of multiple cameras and nozzles, and can perform personalized and autonomous cleaning of waterfowl. In addition, the system has good scalability, can adapt to breeding environments of different sizes, and is easy to upgrade and maintain later.

[0025] (5) Intelligent removal of breeding cage nets: A multi-stage defense method based on the combination of Unet semantic segmentation network and Pix2pixHD generates an adversarial network, which improves the accuracy of removing cage net occlusions. Especially in complex images, it can effectively remove cage nets and restore image details of the occluded parts, providing an accurate basis for subsequent waterfowl feather stain analysis.

[0026] (6) Intelligent identification of pollution: The present invention improves DepLabV3+ and obtains MDA-DepLabV3+ which is suitable for this scenario. It improves the accuracy of identifying stains on QR code labels and waterfowl feathers, and can accurately judge the degree of pollution of waterfowl feathers, especially in complex environments.

[0027] (7) Adaptive cleaning control: Based on the real-time monitoring of pollution levels, the system can dynamically adjust the water flow, spraying time and water temperature, and implement personalized cleaning plans. Waterfowl with different pollution levels can obtain the best cleaning effect, thereby improving the cleaning effect and resource utilization.

[0028] (8) Separation and collection of waterfowl feces and water: The roller separation technology adopted by the present invention can efficiently separate waterfowl feces from washing water. The collected feces can be centrally processed and converted into organic fertilizer, while the washing water can be recycled or enter the sewage treatment system after preliminary filtration, thus realizing the recycling of water resources and environmental protection.

[0029] (9) Through the communication strategy of the waterfowl feather intelligent detection module, the adaptive feather cleaning module, the information storage module, and the information analysis module, the cleaning efficiency is greatly improved, the labor cost is reduced, and the health management level of waterfowl is improved, while ensuring the accuracy and traceability of the cleaning process.

[0030] (10) Intelligent data recording and management: The system records the degree of waterfowl pollution, cleaning frequency, water usage and energy consumption data in real time and uploads them to the backend database to support data analysis and optimize cleaning strategies to improve management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The present invention is a flow chart of a method for intelligent dirt inspection and feather cleaning of caged waterfowl.

[0032] Figure 2This is a module diagram of an intelligent method for detecting dirt and cleaning feathers of caged waterfowl according to the present invention.

[0033] Figure 3 The present invention is a top view schematic diagram of the system structure of an intelligent dirt inspection and feather cleaning method for caged waterfowl.

[0034] Figure 4 The present invention is a schematic side view of the system structure of an intelligent pollution inspection and feather cleaning method for caged waterfowl.

[0035] Among them: 1. High-definition camera; 2. Feed cart; 3. Nozzle; 4. Telescopic rod; 5. Breeding cage; 6. QR code label; 7. Feed trough; 8. Egg collection trough; 9. Core processing module; 10. Manure and water collection conveyor belt; 11. Roller; 12. Drinking water pipe. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0037] Embodiment 1, as Figure 1 and Figure 2 As shown, a method for intelligent dirt inspection and feather cleaning of caged waterfowl comprises the following steps: S1: Move the feed cart with core processing module, HD camera and sprinkler to obtain QR code labels and waterfowl data in breeding cages through HD camera; S2: The core processing module removes the cage net in the image based on the defense algorithm and restores the blocked part; S3: Based on the restored image, the improved MDA-DepLabV3+ model is used to detect the pollution degree of the QR code label and waterfowl feathers, and the stain area ratio is obtained by segmenting the waterfowl area and the stain area; S4: The core processing module compares the proportion of the stain area with the set threshold to obtain a cleaning instruction, and activates the telescopic rod to move the nozzle to the top of the breeding cage according to the cleaning instruction; S5: Use the nozzle to execute the corresponding cleaning instructions, and use the flow sensor and temperature sensor equipped at the nozzle to monitor the relevant data in real time, and feed back the relevant data to the core processing module; S6: Determine whether the feed cart has moved to the end of the breeding cage. If yes, proceed to step S7; otherwise, return to step S1 and continue to move the feed cart. S7: Start the manure-water collection conveyor belt to remove the waterfowl manure, and use the manure-water separation device to dehydrate the collected wet manure and then transfer it to the fermentation tank for fermentation treatment to convert it into organic fertilizer; S8: Use the intelligent detection of waterfowl feather stains and the adaptive clean feather management platform to uniformly manage and display the farms.

[0038] The core processing module in S1 adopts a parallel scheduling mechanism of multi-threading and thread pool as a control strategy to ensure that the system still has high responsiveness and high stability when facing multi-channel input and output tasks; The parallel scheduling mechanism of the multi-thread and thread pool is as follows: the core processing module allocates a control thread to each high-definition camera and the corresponding nozzle control unit, including a camera thread and a nozzle control thread, which are uniformly managed by the thread pool built into the core processing module. The camera thread is used to process image acquisition, feather stain recognition and positioning, and the nozzle control thread is used to control the spraying action (water pressure, water volume, time, etc.) according to the recognition results.

[0039] This mechanism can not only process the cleaning processes of multiple waterfowl in parallel, avoiding the delay caused by single-threaded serial execution, but also supports real-time scheduling and thread load balancing, greatly improving the stability and processing efficiency of the system.

[0040] In order to prevent the occlusion problem of breeding cage nets, which affects the subsequent judgment of waterfowl feather stains, the present invention proposes a defense algorithm for removing breeding cage nets from images. According to the proposed defense algorithm, the image is processed through three steps to remove the influence of breeding cage nets in the image and restore the clear image of the object behind the breeding cage nets. Specifically, Unet is used for semantic segmentation, bitwise operation is used to remove the breeding cage net area, and Pix2pixHD is used to generate adversarial networks to restore the occluded part.

[0041] The defense algorithm in S2 performs the following operations: S21: Perform semantic segmentation on the original image and obtain the semantic map through the Unet network. The formula is:

[0042] in, To obtain the semantic graph, For the Unet network, is the original image; S22: Perform bitwise operation on the original image according to the semantic map, and remove the part marked as the breeding cage net area in the semantic map from the original image by performing mask operation, so as to obtain the image after the breeding cage net is removed. The formula is:

[0043] in, To remove the image after the breeding cage net, and is a pixel; Indicates the breeding cage area, Indicates non-breeding cage area; S23: Based on the image after removing the breeding cage net, the Pix2pixHD network is used to restore the image after removing the breeding cage net to the real image and restore the occluded part. The formula is:

[0044] in, For the restored image, For Pix2pixHD Network.

[0045] The core of image restoration is to use a generative adversarial network (GAN), specifically Pix2pixHD. Pix2pixHD is an image-to-image translation network suitable for tasks such as image restoration and denoising. It restores the image after removing the cage net to a real image through adversarial training of a generator and a discriminator. The generator tries to generate realistic images, and the discriminator evaluates the authenticity of the generated images. Through multiple iterative training, the generator learns how to repair the missing parts in the image, so that the restored image is restored to a state close to the real one.

[0046] The S3 includes the following sub-steps: S31: Based on the restored image, the improved MDA-DepLabV3+ model is used to identify the QR code label, and the Pyzbar library is used to decode it and extract the breeding cage number; S32: Use the improved MDA-DepLabV3+ model to segment the waterfowl area and the stain area, and obtain the stain area ratio. The formula is:

[0047] in, is the stain area ratio, is the area of ​​the stain, is the total area of ​​waterfowl areas.

[0048] The improved MDA-DepLabV3+ model uses the deep separable convolution DSConv to replace the traditional convolution Conv, uses the MobileNetV2 network to replace the original backbone network, and introduces the dynamic void spatial pyramid pool ASPP module.

[0049] Among them, by using depthwise separable convolution, the amount of calculation and parameters can be significantly reduced, thereby improving the inference speed and reducing the burden on the device. As a lightweight convolutional neural network, MobileNetV2's low computational complexity and high efficiency are very suitable for real-time processing in resource-constrained environments and are suitable for embedded devices in breeding environments. The ASPP module effectively captures multi-scale information in the image, helping the model to identify stains of different scales, especially in the feather area. The details have been significantly improved, and smaller pollution areas can be identified.

[0050] The training and evaluation of the MDA-DepLabV3+ model includes the following steps: First, image data containing QR code labels and waterfowl stains in different scenarios were collected, and the images were accurately annotated using the Labelme annotation tool. Then, the dataset was divided into training set, validation set, and test set in a ratio of 7:2:1. In order to ensure the training effect, the size of all images was uniformly adjusted to 640x640 pixels. During the training process, MDA-DepLabV3+ was used as the model architecture and 200 rounds of training were performed. After each round of training, the model was evaluated using the validation set, and the best network training weights were saved based on the verification results. Finally, the generalization ability of the model was evaluated using the test set to ensure that the model performed stably in practical applications and avoid overfitting. This training process helps to improve the accuracy and robustness of the model in stain detection tasks. At the same time, through the evaluation of the validation set and the test set, it is ensured that the model has good generalization ability and can work stably in different waterfowl feather stain scenarios.

[0051] In this embodiment, the high-definition camera communicates with the core processing module via USB to detect the contamination level of the QR code label and the corresponding waterfowl feathers in the breeding cage. The specific process is as follows: First of all, it is necessary to accurately identify the degree of contamination of QR code labels and waterfowl in complex scenarios. In addition, a core processing module needs to detect multiple images at the same time and upload the detection data, so the inference speed of the model needs to be improved. According to the requirements, the improved model MDA-DepLabV3+ is used to identify QR code labels and waterfowl.

[0052] The improved model first identifies the QR code label and decodes it through the Pyzbar library in the Python program to extract the cage number. Based on the result of the QR code recognition, the program saves the corresponding cage number and detects the waterfowl identified in the cage. Next, the degree of contamination of the waterfowl feathers will be evaluated.

[0053] The improved MDA-DepLabV3+ model segments the waterfowl area and the stain area. To ensure the accuracy of recognition, the feed cart will hold one cage for a period of time each time. After extracting the stain area, the proportion of the stain area is calculated.

[0054] Next, based on the customized stain ratio threshold in the operation management of the waterfowl feather stain intelligent detection and adaptive feather cleaning management system platform, , and , divides waterfowl stains into the following three categories:

[0055] In order to ensure the accuracy of the cleaning process and the health management of waterfowl, the core processing module will be based on the degree of contamination of each waterfowl. C Generate cleaning instructions. The cleaning instructions include: starting the telescopic rod to extend the nozzle into the top of the cage, and selecting the appropriate water flow pressure, cleaning time and water flow temperature. This step ensures that the core processing module can transmit the correct pollution information to the adaptive feather cleaning device (including water flow sensor, temperature sensor and telescopic rod) based on the results of the stain detection.

[0056] Next, the core processing module generates different control signals based on the pollution level information received. Specifically, the pollution level information determines the cleaning operation parameters of the adaptive feather cleaning device: Light pollution: Instruct the adaptive feather cleaning device to start low water pressure and short time spraying, and keep the water temperature at normal temperature (for example 20-25°C).

[0057] Moderate pollution: indicates moderate water pressure, moderate spraying time, and moderately elevated water temperature (e.g. 25-30°C).

[0058] Severe soiling: Indicates high water pressure, long spraying time, and increased water temperature (e.g. 30-40°C) to enhance cleaning effect.

[0059] During this process, the cleaning instructions include not only the adjustment of water flow, spraying time and water temperature, but also the operation of starting the telescopic rod. The core processing module will start the telescopic rod through instructions and extend the nozzle into the top of the cage to ensure that the feather area of ​​the waterfowl can be covered for cleaning.

[0060] The adaptive feather cleaning device monitors the spraying water volume and water temperature in real time through flow sensors and temperature sensors, and feeds relevant data back to the core processing module. The system records data such as waterfowl pollution degree, cleaning frequency, water consumption and energy consumption, and uploads these data to the backend database for storage and analysis. Through data analysis, the system can optimize the cleaning strategy, improve overall management efficiency, and ensure that the waterfowl feather cleaning process is efficient and accurate.

[0061] The S8 waterfowl feather stain intelligent detection and adaptive feather cleaning management platform includes: Waterfowl feather stain cleaning report: The waterfowl feather stain cleaning report is used to record the cleaning operation information of each waterfowl feather, including waterfowl code, cleaning time, pollution degree, breeding cage number, cleaning water flow, spraying time, spraying water temperature, water usage, energy consumption, cleaning effect evaluation and related pictures generated during the cleaning process. These data can help administrators monitor the cleaning status of each waterfowl in real time, evaluate the cleaning effect, and provide a basis for subsequent cleaning strategies; Group waterfowl cleaning performance report: The group waterfowl cleaning performance report is used to display the cleaning effect of waterfowl groups, including cleaning time distribution, average water flow, cleaning time and group water resource usage, and provides cleaning efficiency and effect trend analysis to help optimize group cleaning solutions; User feedback report: The user feedback report is used to upload feedback information through the mini program, including information about incomplete feather cleaning and equipment failure. The platform automatically generates a feedback report to help administrators deal with problems and improve system service quality; Equipment alarm table: The equipment alarm table is used to monitor the equipment status in real time, automatically generate equipment fault alarm table, and remind the administrator to repair in time, including equipment type, fault description and occurrence time; Operation management module: The operation management module is used to provide operation management functions, including threshold management, equipment management, farm management and waterfowl file management. Managers can set various management thresholds (such as water flow threshold, cleaning time threshold, temperature threshold, etc.) according to actual operational needs. The farm management function helps administrators to divide the farm into different areas, set cleaning parameters and equipment for different areas, and ensure that waterfowl in different areas receive the most suitable cleaning conditions. At the same time, the waterfowl file management function can record the cleaning history, health status, etc. of each waterfowl, which is convenient for administrators to conduct long-term tracking and management; Report and visualization display platform: The report and visualization display platform has the function of generating various reports, supports detailed statistics and analysis of cleaning data, equipment status, and early warning information, and provides a variety of visualization display methods to facilitate users to view and interpret data and grasp the system operation status in real time.

[0062] Embodiment 2, as Figure 3 and Figure 4As shown, a system for an intelligent pollution inspection and feather cleaning method for caged waterfowl, the system comprises a plurality of breeding cages 5 of the same structure, and the plurality of breeding cages 5 are connected in parallel, the breeding cages 5 are provided with a drinking pipe 12, a QR code label 6, a feed trough 7, an egg collecting trough 8, a manure collecting conveyor belt 10 and a roller 11, the drinking pipe 12 is located inside the breeding cage 5, the feed trough 7 is located outside the breeding cage 5, the QR code label 6 is pasted on the feed trough 7, the egg collecting trough 8 is located below the feed trough 7, the manure collecting conveyor belt 10 is located below each breeding cage 5, and the roller 11 is located at the tail of the manure collecting conveyor belt 10; The system also includes a feed cart 2, which is located directly in front of the feed trough 7 of the multi-layer breeding cage 5. A core processing module 9 is mounted in the middle of the feed cart 2. The core processing module 9 is connected to multiple high-definition cameras 1 and multiple nozzles 3. The nozzles 3 are connected to the feed cart 2 via a telescopic rod 4.

[0063] The telescopic rod 4 is driven by an electric push rod or a pneumatic cylinder, and a water quality filtering unit is integrated inside the nozzle 3, including a replaceable filter element and a water pressure stabilizer.

[0064] The two sides of the manure-collecting water conveyor belt 10 are rolled up to form a sealed enclosure structure, which can effectively prevent the overflow of waterfowl feces and spray water during transportation, ensuring a clean environment. The manure-collecting water conveyor belt 10 is transported by a motor drive. The end of the manure-collecting water conveyor belt 10 is equipped with a manure scraper, which can effectively remove the feces on the surface of the conveyor belt, and an adjustable roller 11, which can press out the excess water in the wet manure, so that the water seeping from the spray water and waterfowl feces can be effectively introduced into the collection ditch and enter the subsequent water resource treatment system. In the process of collecting manure, a special "wet-dry separation unit" is set up, which can separate the collected wet manure from excess water and reduce the moisture content of the manure. The wet manure can be further dehydrated through the dry-wet separation device, or it can be guided to a fermentation tank for fermentation treatment and converted into organic fertilizer.

[0065] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the invention.

Claims

1. An intelligent method for cleaning feathers of caged waterfowl, characterized in that: The following steps are involved: S1: Move the feed cart with core processing module, HD camera and sprinkler to obtain QR code labels and waterfowl data in breeding cages through HD camera; S2: The core processing module removes the cage net in the image based on the defense algorithm and restores the blocked part; S3: Based on the restored image, the improved MDA-DepLabV3+ model is used to detect the pollution degree of the QR code label and waterfowl feathers, and the stain area ratio is obtained by segmenting the waterfowl area and the stain area; S4: The core processing module compares the proportion of the stain area with the set threshold to obtain a cleaning instruction, and activates the telescopic rod to move the nozzle to the top of the breeding cage according to the cleaning instruction; S5: Use the nozzle to execute the corresponding cleaning instructions, and use the flow sensor and temperature sensor equipped at the nozzle to monitor the relevant data in real time, and feed back the relevant data to the core processing module; S6: Determine whether the feed cart has moved to the end of the breeding cage. If yes, proceed to step S7; otherwise, return to step S1 and continue to move the feed cart. S7: Start the manure-water collection conveyor belt to remove the waterfowl manure, and use the manure-water separation device to dehydrate the collected wet manure and then transfer it to the fermentation tank for fermentation treatment to convert it into organic fertilizer; S8: Use the intelligent detection of waterfowl feather stains and the adaptive clean feather management platform to uniformly manage and display the farms.

2. The intelligent pollution inspection and feather cleaning method for caged waterfowl according to claim 1 is characterized in that: The core processing module in S1 adopts a parallel scheduling mechanism of multi-threading and thread pool as a control strategy; The parallel scheduling mechanism of the multi-threading and thread pool is specifically as follows: the core processing module allocates a control thread to each high-definition camera and the corresponding nozzle control unit, including a camera thread and a nozzle control thread, and the camera thread and the nozzle control thread are uniformly managed by the thread pool built into the core processing module.

3. The intelligent pollution inspection and feather cleaning method for caged waterfowl according to claim 1, characterized in that: The defense algorithm in S2 performs the following operations: S21: Perform semantic segmentation on the original image and obtain the semantic map through the Unet network. The formula is: in, To obtain the semantic graph, For the Unet network, is the original image; S22: Perform bitwise operation on the original image according to the semantic map, and remove the part marked as the breeding cage net area in the semantic map from the original image by performing mask operation, so as to obtain the image after the breeding cage net is removed. The formula is: in, To remove the image after the breeding cage net, and is a pixel; S23: Based on the image after removing the breeding cage net, the Pix2pixHD network is used to restore the image after removing the breeding cage net to the real image and restore the occluded part. The formula is: in, For the restored image, For Pix2pixHD Network.

4. The intelligent pollution inspection and feather cleaning method for caged waterfowl according to claim 1, characterized in that: The S3 includes the following sub-steps: S31: Based on the restored image, the improved MDA-DepLabV3+ model is used to identify the QR code label, and the Pyzbar library is used to decode it and extract the breeding cage number; S32: Use the improved MDA-DepLabV3+ model to segment the waterfowl area and the stain area, and obtain the stain area ratio. The formula is: in, is the stain area ratio, is the area of ​​the stain, is the total area of ​​waterfowl areas.

5. The intelligent feather cleaning method for caged waterfowl according to claim 1, characterized in that: The improved MDA-DepLabV3+ model uses the deep separable convolution DSConv to replace the traditional convolution Conv, uses the MobileNetV2 network to replace the original backbone network, and introduces the dynamic void spatial pyramid pool ASPP module.

6. The intelligent feather cleaning method for caged waterfowl according to claim 1, characterized in that: The S8 waterfowl feather stain intelligent detection and adaptive feather cleaning management platform includes: Waterfowl feather stain cleaning report: The waterfowl feather stain cleaning report is used to record the cleaning operation information of each waterfowl feather, including waterfowl code, cleaning time, pollution degree, breeding cage number, cleaning water flow, spraying time, spraying water temperature, water usage, energy consumption, cleaning effect evaluation and related pictures generated during the cleaning process; Group waterfowl cleaning performance report: The group waterfowl cleaning performance report is used to display the cleaning effect of waterfowl groups, including cleaning time distribution, average water flow, cleaning time and group water resource usage, and provides cleaning efficiency and effect trend analysis to help optimize group cleaning solutions; User feedback report: The user feedback report is used to upload feedback information through the mini program, including information about incomplete feather cleaning and equipment failure. The platform automatically generates a feedback report to help administrators deal with problems and improve system service quality; Equipment alarm table: The equipment alarm table is used to monitor the equipment status in real time, automatically generate equipment fault alarm table, and remind the administrator to repair in time, including equipment type, fault description and occurrence time; Operation management module: The operation management module is used to provide operation management functions, including threshold management, equipment management, farm management and waterfowl file management; Report and visualization display platform: The report and visualization display platform has the function of generating various reports, supports detailed statistics and analysis of cleaning data, equipment status, and warning information, and provides a variety of visualization display methods.

7. A system for intelligent pollution inspection and feather cleaning of caged waterfowl according to any one of claims 1 to 6, characterized in that: The system comprises a plurality of breeding cages (5) of the same structure, and the plurality of breeding cages (5) are connected in parallel, and the breeding cages (5) are provided with a drinking pipe (12), a QR code label (6), a feed trough (7), an egg collecting trough (8), a manure collecting conveyor belt (10) and a roller (11), wherein the drinking pipe (12) is located inside the breeding cage (5), the feed trough (7) is located outside the breeding cage (5), the QR code label (6) is attached to the feed trough (7), the egg collecting trough (8) is located below the feed trough (7), the manure collecting conveyor belt (10) is located below each breeding cage (5), and the roller (11) is located at the tail of the manure collecting conveyor belt (10); The system further comprises a feed cart (2), the feed cart (2) being located directly in front of a feed trough (7) of a multi-layer breeding cage (5), a core processing module (9) being mounted in the middle of the feed cart (2), the core processing module (9) being connected to a plurality of high-definition cameras (1) and a plurality of nozzles (3), and the nozzles (3) being connected to the feed cart (2) via a telescopic rod (4).

8. The system of the intelligent pollution inspection and feather cleaning method for caged waterfowl according to claim 7, characterized in that: The telescopic rod (4) is driven by an electric push rod or a pneumatic cylinder, and a water quality filtering unit is integrated inside the nozzle (3), including a replaceable filter element and a water pressure stabilizer.

9. The system of the intelligent pollution inspection and feather cleaning method for caged waterfowl according to claim 7, characterized in that: The two sides of the manure-water collecting conveyor belt (10) are rolled up to form a sealed enclosure structure. The manure-water collecting conveyor belt (10) is driven by a motor for transportation. The end of the manure-water collecting conveyor belt (10) is equipped with a manure scraper and an adjustable roller (11) that can press out excess water in wet manure.

Images