Image recognition system of breeding chicken manure conveyor belt

Abstract

The invention discloses an image recognition system for a chicken manure breeding conveyor belt, which comprises a conveyor belt, a network camera for acquiring chicken manure images and gateway equipment for processing image data, and is characterized in that the conveyor belt is divided into a plurality of detection units according to the number of cages; each detection unit corresponds to a chicken manure output area of one coop; the network camera is used for continuously photographing the conveyor belt at a fixed frequency and uploading the photographed images to the gateway equipment; and an edge calculation module is arranged in the gateway equipment and is used for identifying the starting point and the ending point of the effective part of each conveyor belt, so that an image area needing to be processed is determined. According to the invention, through automatic photographing, image splicing, separation and AI identification, real-time monitoring and anomaly detection of the chicken manure conveyor belt are realized, the chicken manure anomaly detection efficiency is improved, and intelligent support is provided for breeding management.

Description

Technical Field

[0001] This invention relates to the field of poultry farming technology, and in particular to an image recognition system for a chicken manure conveyor belt. Background Technology

[0002] In poultry farming, timely cleaning and monitoring of chicken manure are crucial for preventing the spread of disease. Traditional methods of cleaning chicken manure mainly rely on manual inspection, which is inefficient and prone to missed detections. Existing technologies have attempted to use automated equipment for cleaning, but their ability to identify abnormalities in chicken manure is limited. For example, some systems rely solely on scheduled cleaning and lack real-time monitoring capabilities, making it difficult to detect anomalies promptly. With the development of artificial intelligence and image recognition technologies, the use of AI technology for automated monitoring of chicken manure conveyor belts is gradually showing potential, but further optimization is still needed in practical applications to improve recognition accuracy and efficiency. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention proposes an image recognition system for chicken manure conveyor belts. By capturing real-time images of chicken manure on the conveyor belt and performing multi-dimensional analysis, the system optimizes the recognition accuracy of anomalies by combining color and regional features. Simultaneously, the recognition results are uploaded to the cloud to achieve remote monitoring and local early warning functions.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: An image recognition system for a chicken manure conveyor belt includes a conveyor belt, a network camera for capturing images of chicken manure, and a gateway device for processing image data. The system is characterized in that the conveyor belt is divided into several detection units based on the number of cages, with each detection unit corresponding to the chicken manure output area of ​​one cage; the network camera continuously captures images of the conveyor belt at a fixed frequency and uploads the captured images to the gateway device; the gateway device has a built-in edge computing module used to identify the start and end points of the effective portion of each conveyor belt, thereby determining the image area to be processed; the gateway device also includes a stitching module for stitching multiple captured images into a complete conveyor belt image; the gateway device further includes a segmentation module for dividing the stitched complete image into equidistant segments based on the number of cages, forming chicken manure images corresponding to each detection unit; the gateway device also includes a recognition module that classifies and identifies abnormal points in the chicken manure images based on AI algorithms.

[0005] In the image recognition system for chicken manure conveyor belts of this invention, the recognition module optimizes color recognition of soft-shelled eggs, bloody feces, and green feces in chicken manure images using the HSV color model. Specifically, the HSV model replaces the traditional RGB model, separating hue, saturation, and brightness information to reduce the impact of lighting changes on color recognition. Through training with a large amount of sample data, color threshold ranges for soft-shelled eggs, bloody feces, and green feces are set. When the color value of a certain area in the image falls within the corresponding range, that area is marked as a suspected anomaly. Furthermore, the recognition module also performs color enhancement processing on the image using histogram equalization to highlight target color features.

[0006] In the image recognition system for a chicken manure conveyor belt of this invention, the recognition module further optimizes the region recognition for blank areas without manure caused by dead chickens. First, the shape, size, and edge features of the blank areas are analyzed. For example, the blank areas under dead chickens are usually irregular in shape and have a specific area range. Second, a region segmentation algorithm is used to divide the conveyor belt image into multiple small regions, and regions that match the characteristics of blank areas under dead chickens are selected based on the features of the blank areas. Finally, a comprehensive judgment is made by combining color features and region features. If a region matches both the characteristics of a blank area and the color of the surrounding environment, it is determined that a dead chicken exists in that region.

[0007] In the image recognition system for a chicken manure conveyor belt of the present invention, the gateway device further includes an integration module for performing correlation analysis on the results of color recognition and region recognition; the integration module marks the identified suspected abnormal points and generates integrated recognition results; the gateway device uploads the integrated recognition results to a cloud server through a wireless communication module, and simultaneously provides real-time warning prompts on a local display terminal.

[0008] In the image recognition system for chicken manure conveyor belts of this invention, the network camera is fixed to a bracket directly above the conveyor belt via a threaded connection. The height of the bracket is adjustable to accommodate conveyor belts of different widths. The lens of the network camera adopts a wide-angle design, covering the entire width of the conveyor belt to ensure that a complete image of the chicken manure is captured. The network camera integrates a supplementary light, which automatically adjusts its brightness through a light sensor to adapt to shooting needs under different lighting conditions.

[0009] In the image recognition system for a chicken manure conveyor belt of this invention, the gateway device is connected to a cloud server via wired or wireless means. The cloud server stores historical recognition data and provides a remote access interface. The edge computing module of the gateway device adopts an embedded processor with a built-in dedicated algorithm library to support real-time image processing. The display terminal of the gateway device adopts a touch screen design with an intuitive operation interface, making it easy for staff to view recognition results and adjust parameters.

[0010] In the image recognition system for the chicken manure conveyor belt of this invention, the surface of the conveyor belt is specially treated to have anti-slip and easy-to-clean properties; baffles are provided on both sides of the conveyor belt to prevent chicken manure from overflowing; the drive motor of the conveyor belt is controlled by a frequency converter to dynamically adjust the running speed according to the amount of chicken manure.

[0011] A method for monitoring a chicken manure conveyor belt, characterized in that the monitoring method includes the following steps: S1: Take continuous photos of the conveyor belt at a fixed frequency using the network camera, and upload the captured images to the gateway device; S2: The edge computing module identifies the start and end points of the effective portion of each conveyor belt to determine the image area that needs to be processed; S3: The stitching module stitches together multiple captured images into a complete conveyor belt image; S4: The segmentation module divides the spliced ​​complete image into equal intervals according to the number of cages to form a chicken manure image corresponding to each detection unit; S5: The recognition module classifies and identifies abnormal points in the chicken manure image based on AI algorithms, including color recognition optimization and region recognition optimization. S6: The integration module performs correlation analysis on the results of color recognition and region recognition to generate an integrated recognition result; S7: The integrated identification results are uploaded to the cloud server through the wireless communication module, and real-time warning prompts are displayed on the local display terminal.

[0012] In the monitoring method for chicken manure conveyor belt of this invention, the specific steps of color recognition optimization in S5 are as follows: S5.1: The HSV color model is adopted instead of the traditional RGB color model, separating hue, saturation and lightness information; S5.2: Through training with a large amount of sample data, set the color threshold range for soft-shelled eggs, bloody feces, and green feces; S5.3: When the color value of a certain region in an image falls within the corresponding range, the region is marked as a suspected anomaly. S5.4: Histogram equalization is used to enhance the color of the image and highlight the color features of the target.

[0013] In the monitoring method for chicken manure conveyor belts of this invention, the specific steps of area identification and optimization in S5 are as follows: S5.5: Analyze the shape, size, and edge features of the blank, feces-free area created under the dead chicken; S5.6: The conveyor belt image is divided into multiple small regions using a region segmentation-based algorithm; S5.7: Select areas that match the characteristics of the blank area under the dead chicken based on the features of the blank area; S5.8: Combine color features and area features for comprehensive judgment. If an area meets the characteristics of a blank area and matches the color of the surrounding environment, then it is determined that there is a dead chicken in the area.

[0014] In the monitoring method for chicken manure conveyor belt of this invention, the specific steps of real-time early warning in S7 are as follows: S7.1: Mark the location of suspected anomalies graphically on the local display terminal; S7.2: Alert staff by sounding an alarm or flashing lights; S7.3: Record the time, location, and type of the anomaly and generate a log file for later querying.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves real-time monitoring and anomaly detection of chicken manure conveyor belts through automatic photography, image stitching, segmentation, and AI recognition, thereby improving the efficiency of chicken manure anomaly detection and providing intelligent support for breeding management.

[0016] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall structure of the image recognition system for the chicken manure conveyor belt of the present invention.

[0018] Figure 2 This is a functional block diagram of the gateway device in this invention.

[0019] Figure 3 This is a flowchart of the monitoring method of the present invention.

[0020] Figure 4 This is a schematic diagram of the structure of the chicken manure conveyor belt and network camera of the present invention.

[0021] The attached diagram lists the components represented by each number as follows: 1. Conveyor belt; 2. Webcam. Detailed Implementation

[0022] Please see Figures 1-4 In this embodiment of the invention, an image recognition system for a chicken manure conveyor belt includes a conveyor belt 1, a network camera 2 for acquiring images of chicken manure, and a gateway device for processing image data. The conveyor belt 1 is divided into several detection units according to the number of cages, and each detection unit corresponds to the chicken manure output area of ​​a chicken cage. The network camera 2 takes continuous pictures of the conveyor belt 1 at a fixed frequency and uploads the captured images to the gateway device. The gateway device has a built-in edge computing module, splicing module, segmentation module, recognition module, integration module, and wireless communication module.

[0023] Conveyor belt 1 is one of the core components of the entire system. Its surface is specially treated to provide anti-slip and easy-to-clean properties. Conveyor belt 1 has baffles on both sides to prevent chicken manure from overflowing. Its drive motor's speed is controlled by a frequency converter, dynamically adjusting the operating speed according to the amount of chicken manure. Conveyor belt 1 is divided into several detection units based on the number of cages, with each detection unit corresponding to the chicken manure output area of ​​one cage. This division ensures accurate location of the chicken manure situation within each detection unit during subsequent image processing. Network camera 2 is fixed to a bracket directly above conveyor belt 1 via a threaded connection. The bracket's height is adjustable to accommodate conveyor belt 1 of varying widths. Network camera 2 features a wide-angle lens that covers the entire width of conveyor belt 1, ensuring complete capture of the chicken manure image. Furthermore, network camera 2 integrates a supplementary light, whose brightness is automatically adjusted by a light sensor to adapt to shooting needs under different lighting conditions.

[0024] Network camera 2 continuously captures images of conveyor belt 1 at a fixed frequency and uploads the images to the gateway device. The gateway device has a built-in edge computing module, which first identifies the start and end points of the effective portion of each conveyor belt 1 to determine the image area to be processed. This process is achieved by analyzing the boundary features of conveyor belt 1, such as detecting the position of the baffles on both sides of conveyor belt 1 or defining the effective area based on the color and texture features of the conveyor belt 1 surface. Subsequently, the stitching module stitches the multiple captured images into a complete image of conveyor belt 1. During the stitching process, the stitching module aligns and merges the overlapping areas between images to ensure that the stitched image is seamless and without obvious stitching marks. After stitching is completed, the segmentation module divides the stitched complete image into equal intervals according to the number of cages, forming a chicken manure image corresponding to each detection unit. The segmentation module works based on the width of conveyor belt 1 and the number of detection units, extracting each detection unit from the complete image by calculating its width range.

[0025] The recognition module uses AI algorithms to classify and identify anomalies in chicken manure images, including color recognition optimization and region recognition optimization. The specific steps for color recognition optimization are as follows: First, the HSV color model is used instead of the traditional RGB color model to separate hue, saturation, and brightness information. The HSV model can reduce the impact of lighting changes on color recognition, thereby improving recognition accuracy. Second, through training with a large amount of sample data, color threshold ranges are set for soft-shelled eggs, bloody feces, and green feces. When the color value of a certain region in the image falls within the corresponding range, that region is marked as a suspected anomaly. In addition, the recognition module also performs color enhancement processing on the image through histogram equalization to highlight target color features. Histogram equalization adjusts the grayscale distribution of the image, making the colors of the target regions more vivid, thus facilitating subsequent recognition operations.

[0026] The region recognition optimization focuses on analyzing the shape, size, and edge features of the blank, feces-free areas created by dead chickens. First, the recognition module analyzes the shape, size, and edge features of the blank areas; for example, the blank areas under dead chickens are typically irregular in shape and have a specific area range. Second, a region segmentation algorithm is used to divide the conveyor belt image into multiple smaller regions, and regions matching the characteristics of blank areas under dead chickens are selected based on their features. Finally, a comprehensive judgment is made by combining color and region features. If a region matches both the characteristics of a blank area and the color of its surroundings, it is determined that a dead chicken exists in that region. This comprehensive judgment method effectively avoids false positives and improves the accuracy of recognition.

[0027] The integration module correlates and analyzes the results of color recognition and region recognition to generate a unified recognition result. First, the integration module marks identified suspected anomalies and records their location, type, and time information. Then, the integration module uploads the unified recognition result to the cloud server via a wireless communication module, while simultaneously providing real-time alerts on the local display terminal. The wireless communication module supports wired or wireless connection to the cloud server, which stores historical recognition data and provides a remote access interface for users to view and analyze data at any time. The local display terminal features a touchscreen design with an intuitive interface, facilitating easy viewing of recognition results and parameter adjustments. On the local display terminal, the location of suspected anomalies is marked graphically and alerted to staff through audible alarms or flashing lights. Furthermore, the system records the time, location, and type information of the anomalies and generates log files for later retrieval.

[0028] The monitoring method of the present invention includes the following steps: First, a network camera 2 continuously takes pictures of the conveyor belt 1 at a fixed frequency and uploads the captured images to a gateway device. Second, an edge computing module identifies the start and end points of the effective portion of each conveyor belt 1 to determine the image area to be processed. Then, a stitching module stitches multiple captured images into a complete image of the conveyor belt 1, and a segmentation module divides the stitched image into equal intervals according to the number of cages, forming a chicken manure image corresponding to each detection unit. Next, a recognition module classifies and identifies abnormal points in the chicken manure image based on an AI algorithm, including color recognition optimization and region recognition optimization. Then, an integration module performs correlation analysis on the results of color recognition and region recognition to generate an integrated recognition result. Finally, a wireless communication module uploads the integrated recognition result to a cloud server and simultaneously provides real-time warning prompts on a local display terminal.

[0029] In practical applications, the system and method of this invention can significantly improve the efficiency and accuracy of chicken manure management. For example, in large-scale farms, conveyor belt 1 needs to process a large amount of chicken manure daily. Traditional manual inspection methods are not only time-consuming and labor-intensive, but also prone to missing anomalies. This invention, through automated image recognition technology, can complete the scanning and analysis of the entire conveyor belt 1 in a short time and accurately locate anomalies. Furthermore, the system's remote monitoring function allows managers to view the recognition results at any time via a cloud server and take timely measures to resolve anomalies. This efficient management model not only improves the operational efficiency of the farm but also reduces economic losses caused by anomalies.

[0030] In terms of hardware configuration, the gateway device uses an embedded processor with a built-in dedicated algorithm library to support real-time image processing. The selection of the embedded processor must consider its computing power and power consumption to ensure the system maintains stability and efficiency during long-term operation. Furthermore, the gateway device's display terminal features a touchscreen design with an intuitive interface, facilitating easy viewing of recognition results and parameter adjustments. The display terminal's interface includes an image display area, an anomaly marking area, and a parameter setting area. Users can quickly switch between different function modules and view detailed recognition results and historical records via the touchscreen.

[0031] In terms of software implementation, the various functional modules of this invention work collaboratively to complete image recognition and anomaly detection tasks. The edge computing module is responsible for preliminary image processing, including identifying the effective area of ​​conveyor belt 1 and determining the processing range. The stitching module and segmentation module are responsible for image stitching and segmentation, respectively, ensuring that the image of each detection unit can be accurately extracted. The recognition module implements color recognition and region recognition based on AI algorithms, and improves recognition accuracy through the HSV color model and region segmentation algorithm. The integration module performs correlation analysis on the recognition results and generates the final recognition report. The wireless communication module is responsible for data transmission, ensuring that the recognition results can be uploaded to the cloud server and local display terminal in a timely manner.

[0032] The system and method of this invention perform excellently in practical applications, especially maintaining high recognition accuracy under complex lighting conditions. For example, in low-light conditions such as cloudy days or nighttime, the supplementary light of the network camera 2 can automatically adjust its brightness to ensure the quality of the captured images. Furthermore, the application of the HSV color model enables the system to accurately identify target color features under different lighting conditions, thereby reducing the false positive rate. Region recognition optimization further improves the ability to identify blank areas under dead chickens by combining shape, size, and edge features. This multi-dimensional recognition approach allows the system to cope with various complex farming environments and meet the actual needs of users.

[0033] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention is further explained below in conjunction with a specific application scenario.

[0034] In actual operation, farm staff first start conveyor belt 1 and then continuously photograph it at a fixed frequency using webcam 2. Webcam 2's lens covers the entire width of conveyor belt 1, and its integrated supplementary lighting automatically adjusts the light intensity based on the ambient brightness detected by a light sensor, thus ensuring image quality. For example, in low-light conditions such as cloudy days or nighttime, the supplementary lighting automatically increases brightness, while during the day it appropriately decreases brightness to avoid overexposure. This adaptive supplementary lighting mechanism effectively solves the image acquisition problem under different lighting conditions, providing high-quality raw data for subsequent image processing.

[0035] After the images captured by network camera 2 are uploaded to the gateway device, the edge computing module begins preliminary image processing. By analyzing the positions of the baffles on both sides of conveyor belt 1 and the color and texture features of the conveyor belt 1 surface, the edge computing module accurately locates the effective area of ​​each conveyor belt 1. For example, the boundary features of the baffles typically appear as high-contrast straight lines, while the color and texture of the conveyor belt 1 surface are uniform; these features are used as key criteria for defining the effective area. In this way, the system can quickly eliminate invalid areas, reduce the amount of data processed subsequently, and thus improve overall efficiency.

[0036] Subsequently, the stitching module seamlessly stitches together multiple captured images. Since there may be some overlap during the capture of images by network camera 2, the stitching module matches the grayscale values ​​and texture features of the overlapping areas to ensure that the stitched image has no obvious stitching artifacts. For example, in the overlapping area of ​​two images, the stitching module prioritizes pixels with grayscale values ​​closer to the average value for fusion, thus ensuring a smooth transition in the stitched image. After stitching is complete, the segmentation module divides the complete image into equidistant segments based on the number of cages, forming chicken manure images corresponding to each detection unit. This process is based on precise calculations of the total width of conveyor belt 1 and the number of detection units to ensure that the image of each detection unit can be accurately extracted.

[0037] The recognition module performs anomaly classification on the chicken manure images of each detection unit. In color recognition optimization, the module uses the HSV color model instead of the traditional RGB model to separate hue, saturation, and brightness information. For example, when the hue value of a detected area falls within the threshold range for bloody feces, and its saturation and brightness meet the set standards, that area is marked as a suspected bloody feces anomaly. Furthermore, the recognition module also performs color enhancement processing on the image through histogram equalization to highlight the color features of the target area. For example, in a certain detection unit, the color of green feces may appear dull due to insufficient lighting; histogram equalization adjusts the grayscale distribution to make the color of the target area more vivid, thus facilitating subsequent recognition operations.

[0038] In the region recognition optimization, the recognition module analyzes the shape, size, and edge features of the blank areas created by the dead chicken. For example, the blank areas under a dead chicken are usually irregular in shape, and their area ranges within a certain threshold. The recognition module segments the conveyor belt image into multiple small regions using a region segmentation algorithm and filters out regions that match the characteristics of blank areas with dead chickens. Subsequently, the system combines color features and region features for a comprehensive judgment. For example, if a region matches both the shape and area characteristics of a blank area and its color matches the surrounding environment, then it is determined that a dead chicken exists in that region. This multi-dimensional recognition method significantly improves the accuracy of the system.

[0039] The integration module correlates and analyzes the results of color recognition and region recognition, generating an integrated recognition result. For example, when blood feces and a dead chicken blank area are present simultaneously in a detection unit, the integration module records the location, type, and time information of these two anomalies and uploads it to the cloud server via the wireless communication module. Simultaneously, the local display terminal graphically marks the location of the anomalies and alerts staff through audible alarms or flashing lights. For instance, when blood feces are detected, the display terminal marks the corresponding detection unit's image with a red marker and emits a short alarm sound; when a dead chicken is detected, it is marked with a yellow marker and continuously flashes a light.

[0040] In terms of hardware configuration, the gateway device uses an embedded processor with a built-in dedicated algorithm library, supporting real-time image processing. For example, the high-performance computing capability of the embedded processor ensures that the system can complete the entire process from image acquisition to anomaly identification in a short time, while its low-power design guarantees stability during long-term operation. In addition, the gateway device's touchscreen display terminal provides an intuitive operating interface, allowing users to quickly switch between image display areas, anomaly marking areas, and parameter setting areas via the touchscreen, and view detailed recognition results and historical records.

[0041] In terms of software implementation, the various functional modules of this invention work together to ensure the efficient completion of image recognition and anomaly detection tasks. For example, the edge computing module is responsible for preliminary image processing, including identifying the effective area of ​​conveyor belt 1 and determining the processing range; the stitching module and the segmentation module are responsible for image stitching and segmentation, respectively, to ensure that the image of each detection unit can be accurately extracted; the recognition module realizes color recognition and region recognition based on AI algorithms, and improves recognition accuracy through the HSV color model and region segmentation algorithm; the integration module performs correlation analysis on the recognition results and generates the final recognition report; the wireless communication module is responsible for data transmission, ensuring that the recognition results can be uploaded to the cloud server and local display terminal in a timely manner.

[0042] By combining the above steps and principles, the system and method of this invention demonstrate excellent performance in practical applications. For example, in large-scale farms, conveyor belt 1 needs to process a large amount of chicken manure daily. Traditional manual inspection methods are not only time-consuming and labor-intensive, but also prone to missing anomalies. However, this invention, through automated image recognition technology, can complete the scanning and analysis of the entire conveyor belt 1 in a short time and accurately locate anomalies. Furthermore, the system's remote monitoring function allows managers to view the recognition results at any time via a cloud server and take timely measures to resolve anomalies. This efficient management model not only improves the operational efficiency of the farm but also reduces economic losses caused by anomalies.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. An image recognition system for a chicken manure conveyor belt (1), characterized in that, The device includes a conveyor belt (1), a network camera (2) for collecting images of chicken manure, and a gateway device for processing image data. The conveyor belt (1) is divided into several detection units according to the number of cages. Each detection unit corresponds to the chicken manure output area of ​​a chicken cage. The network camera (2) takes continuous pictures of the conveyor belt (1) at a fixed frequency and uploads the captured images to the gateway device. The gateway device has a built-in edge computing module, splicing module, segmentation module, recognition module, integration module and wireless communication module. The edge computing module is used to identify the start and end points of the effective part of each conveyor belt (1) to determine the image area to be processed. The stitching module is used to stitch multiple captured images into a complete image of the conveyor belt (1). The segmentation module is used to divide the stitched complete image into equal intervals according to the number of cages to form a chicken manure image corresponding to each detection unit. The recognition module classifies and recognizes abnormal points in the chicken manure image based on AI algorithm. The integration module is used to perform correlation analysis on the results of color recognition and region recognition to generate an integrated recognition result. The wireless communication module is used to upload the integrated recognition result to the cloud server.

2. The image recognition system for a chicken manure conveyor belt according to claim 1, characterized in that, The recognition module optimizes the color recognition of soft-shelled eggs, bloody droppings, and green droppings in chicken droppings images using the HSV color model. It separates hue, saturation, and brightness information and sets color threshold ranges for soft-shelled eggs, bloody droppings, and green droppings. When the color value of a certain area in the image falls into the corresponding range, the area is marked as a suspected anomaly.

3. The image recognition system for a chicken manure conveyor belt according to claim 2, characterized in that, The recognition module performs color enhancement processing on the image through histogram equalization to highlight the target color features.

4. The image recognition system for a chicken manure conveyor belt according to claim 3, characterized in that, The recognition module optimizes the region recognition for the blank area without feces generated by the dead chicken, analyzes the shape, size and edge features of the blank area, and uses a region segmentation algorithm to divide the conveyor belt (1) image into multiple small regions, and selects the regions that meet the characteristics of the blank area under the dead chicken.

5. The image recognition system for a chicken manure conveyor belt according to claim 4, characterized in that, The recognition module combines color features and region features for comprehensive judgment. If a region meets both the characteristics of a blank region and the color of the surrounding environment, it is determined that there is a dead chicken in that region.

6. The image recognition system for a chicken manure conveyor belt according to claim 5, characterized in that, The network camera (2) is fixed to the bracket directly above the conveyor belt (1) by a threaded connection. The height of the bracket is adjustable to accommodate conveyor belts (1) of different widths. The lens of the network camera (2) adopts a wide-angle design to cover the entire width of the conveyor belt (1). The network camera (2) has an integrated fill light and automatically adjusts the brightness through a light sensor.

7. The image recognition system for a chicken manure conveyor belt according to claim 6, characterized in that, The gateway device is connected to the cloud server via wired or wireless means. The cloud server stores historical identification data and provides a remote access interface. The display terminal of the gateway device adopts a touch screen design.

8. The image recognition system for a chicken manure conveyor belt according to claim 7, characterized in that, The surface of the conveyor belt (1) is specially treated to have anti-slip and easy-to-clean properties. Baffles are provided on both sides of the conveyor belt (1) to prevent chicken manure from overflowing. The drive motor of the conveyor belt (1) is controlled by a frequency converter to dynamically adjust the running speed.

9. The image recognition system for a chicken manure conveyor belt according to any one of claims 1-8, characterized in that, It also includes a monitoring method for chicken manure conveyor belts, which includes the following steps: S1: Take continuous photos of the conveyor belt (1) at a fixed frequency using a network camera (2) and upload the captured images to the gateway device; S2: Identify the start and end points of the effective portion of each conveyor belt (1) through the edge computing module to determine the image area that needs to be processed; S3: The stitching module stitches together multiple captured images into a complete conveyor belt (1) image; S4: The segmentation module divides the stitched complete image into equal intervals according to the number of cages to form the chicken manure image corresponding to each detection unit; S5: The recognition module classifies and identifies abnormal points in chicken manure images based on AI algorithms, including color recognition optimization and region recognition optimization; S6: The integration module performs correlation analysis on the results of color recognition and region recognition to generate integrated recognition results; S7: The integrated recognition results are uploaded to the cloud server via the wireless communication module and real-time warning prompts are displayed on the local terminal.

10. The image recognition system for a chicken manure conveyor belt according to claim 9, characterized in that, The specific steps for color recognition optimization in S5 are as follows: S5.1: The HSV color model is adopted to replace the traditional RGB color model to separate hue, saturation and lightness information; S5.2: Set the color threshold range for soft-shelled eggs, bloody feces, and green feces through training with a large amount of sample data; S5.3: When the color value of a certain region in an image falls within the corresponding range, that region is marked as a suspected anomaly. S5.4: Histogram equalization is used to enhance the color of the image to highlight the target color features.

Images