Livestock breeding feed physical property non-contact rapid acquisition and model construction system and method

Abstract

The invention discloses a livestock breeding feed physical property non-contact rapid acquisition and model construction system, which comprises an image acquisition module, an image processing module and a model construction module, wherein the image acquisition module is used for acquiring original image data including two-dimensional color images and three-dimensional depth information of feed particles based on a high-precision camera; the processing control module processes and analyzes original image data based on an image processing algorithm and an evaluation model and extracts physical parameters of feed particles, and the data interaction module is used for transmitting an analysis result to a production management system in real time to complete closed-loop feedback control of production parameters. The visualization module carries out interaction operation and information display based on a visualization user interface; through a non-contact image analysis technology, the feed processing quality can be monitored online in real time, and accurate regulation and control of a production process are directly guided through an intelligent evaluation model to meet individual requirements of different animals, so that the uniformity, palatability and digestibility of feed products are remarkably improved, and the equipment and environment-friendly treatment cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of feed processing technology, and in particular to a system and method for rapid, non-contact acquisition and model building of feed properties for livestock farming. Background Technology

[0002] Modern animal husbandry is developing towards large-scale and intensive operations, and feed particle size directly affects animal nutrient absorption, health, and production performance; different livestock (such as piglets and dairy cows) and growth stages have significantly different requirements for particle size.

[0003] Current mainstream detection methods rely on traditional screening methods, which are inefficient (requiring 15-30 minutes per sample) and have large errors (±8%-12%). They cannot accurately meet the above-mentioned refined requirements and have problems such as strong reliance on manual labor and difficulty in data traceability. Foreign automated equipment (such as high-end imaging systems) are expensive (over 500,000 yuan) and are not suitable for China's high dust and high humidity environment. Domestic technologies are mostly still in the laboratory stage and lack industrial solutions.

[0004] The urgent need for industrial upgrading necessitates the development of a closed-loop system for online particle size detection and real-time feedback to accurately match the refined requirements of animal nutrition. Therefore, this invention proposes a system and method for rapid, non-contact acquisition and model construction of livestock feed properties to address the problems existing in the prior art. Summary of the Invention

[0005] To address the aforementioned problems, the present invention aims to propose a non-contact rapid acquisition and modeling system and method for livestock feed properties. This system and method utilizes non-contact image analysis technology to achieve rapid and automated detection of feed particle properties, effectively overcoming the limitations of low efficiency and delayed results in traditional methods.

[0006] To achieve the objectives of this invention, the following technical solution is provided: a system for rapid, contactless acquisition and model building of livestock feed properties, comprising an image acquisition module, a processing and control module, a data interaction module, and a visualization module. The image acquisition module acquires raw image data of feed pellets, including two-dimensional color images and three-dimensional depth information, using a high-precision camera. The processing and control module processes and analyzes the raw image data based on image processing algorithms and evaluation models to extract the physical property parameters of the feed pellets. The data interaction module transmits the analysis results to the production management system in real time to complete closed-loop feedback control of production parameters. The visualization module provides interactive operation and information display based on a visual user interface.

[0007] Further improvements are made in the following aspects: The processing control module specifically performs preprocessing, binarization, and contour detection on the original image data, identifies and separates individual feed particles, and then calculates the geometric feature parameters of each feed particle, including area, perimeter, roundness, and aspect ratio. Based on user-preset multi-condition classification rules, the identified feed particles are classified and statistically analyzed according to the formula A=Σm nX n *F n Calculate the overall assessment score A for feed properties, where X n F is the weighting coefficient for the nth physical property index. n Let m be the standardized score of the nth physical property index, and m be the total number of physical property indexes.

[0008] Further improvements include: the preprocessing includes image grayscale conversion, histogram equalization, median-based noise reduction, and contrast enhancement. Image grayscale conversion uses a weighted method to convert the original color image into a single-channel grayscale image; histogram equalization redistributes the pixel intensity values ​​of the grayscale image to cover a wider grayscale range; noise reduction sets the grayscale value of each pixel to the median of the grayscale values ​​of all pixels within a certain neighborhood window; and contrast enhancement expands the dynamic range of the image grayscale levels through linear or nonlinear transformations, with an adjustment range of 50%-200%.

[0009] Further improvements are made in the following ways: the binarization is to use the Otsu thresholding method or the adaptive thresholding method to convert the grayscale image into a binary image containing only black and white, and to segment it into foreground and background, supplemented by morphological opening or closing operations to optimize the particle contour; the contour detection specifically involves finding and extracting the pixel sequence representing the object boundary in the binary image.

[0010] The further improvement is that the matching logic of the classification rule requires that all conditions must be met; when classifying particles in batches, the first matching priority principle is adopted. When a particle matches all the conditions of a rule, it is marked as the category corresponding to that rule, and the subsequent rule checks are stopped.

[0011] Further improvements are made in that the physical properties include particle size and particle size distribution, density, morphological and structural characteristics, mechanical strength, moisture content, and mixing uniformity.

[0012] A further improvement is that the closed-loop feedback control specifically adjusts the speed of the feed grinder or the pressure of the pellet mill when the comprehensive evaluation score A or the proportion of a specific particle category exceeds a preset threshold.

[0013] Further improvements are made in that the visual user interface is specifically used to provide real-time image display, visual parameter adjustment, customizable classification rules, particle distribution curve display, and abnormal alarm functions.

[0014] A method for constructing a system for rapid, contactless acquisition and model building of livestock feed properties includes the following steps:

[0015] Step 1: Acquire the original image of the feed sample using the image acquisition module;

[0016] Step 2: The original image is preprocessed by grayscale conversion, noise reduction and contrast enhancement through the processing control module. The preprocessed image is then binarized and segmented to extract the outline of the feed particles.

[0017] Step 3: Filter valid particles based on the contour area threshold and calculate the geometric feature parameters of each particle;

[0018] Step 4: Apply the preset classification rules to automatically classify the particles and count the quantity and proportion of each category. Then, based on the extracted physical property parameters and classification statistics, calculate the comprehensive evaluation score A.

[0019] Step 5: Output the evaluation results through the data interaction module, and use the results to implement feedback control on the production process.

[0020] A further improvement is made in the following way: before acquiring the original image in step one, the system is calibrated using a standard particle sample of known size to establish the conversion relationship between pixel size and actual physical size, and an error compensation model is constructed to correct the measured value.

[0021] The beneficial effects of this invention are as follows: This invention, through non-contact image analysis technology, can monitor feed processing quality online in real time, and directly guide the precise control of production processes through intelligent evaluation models to meet the personalized needs of different animals, thereby significantly improving the uniformity, palatability, and digestibility of feed products, reducing equipment and environmental treatment costs, and effectively reducing labor input and improving production efficiency. It provides key technical support for ensuring livestock nutrition and promoting healthy breeding, and promotes the intelligent and refined development of feed production. Attached Figure Description

[0022] Figure 1 This is a system architecture diagram of the present invention.

[0023] Figure 2 This is a flowchart of the detection process of the present invention. Detailed Implementation

[0024] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0025] according to Figure 1 and Figure 2As shown in the figure, this embodiment provides a system for rapid acquisition and model building of livestock feed properties without contact, including an image acquisition module, a processing and control module, a data interaction module, and a visualization module.

[0026] The image acquisition module uses a high-precision camera to acquire raw image data of feed pellets, including two-dimensional color images and three-dimensional depth information;

[0027] The Azure Kinect DK depth camera is selected as the core hardware, featuring high resolution (12 megapixels), high frame rate (30fps), and depth sensing capabilities. It is adapted to the high dust and strong vibration environment of feed production (through IP65 protection modification) and solves the problem of traditional cameras not being able to fully identify irregular particles. The camera, light source, and conveyor belt linkage components are integrated into the detection terminal, which supports docking with the production line PLC system to realize the full automation of sampling (collecting once every five seconds), imaging, and analysis, with a response time of ≤0.5 seconds, meeting the requirements of online real-time detection.

[0028] The processing and control module processes and analyzes the raw image data based on image processing algorithms and evaluation models, and extracts the physical property parameters of the feed pellets.

[0029] The processing control module specifically preprocesses, binarizes, and performs contour detection on the raw image data, identifies and separates individual feed particles, and then calculates the geometric feature parameters of each feed particle, including area, perimeter, roundness, and aspect ratio. Based on user-preset multi-condition classification rules, it classifies and statistically analyzes the identified feed particles according to the formula A=Σm nX. n *F n Calculate the overall assessment score A for feed properties, where X n F is the weighting coefficient for the nth physical property index. n Let m be the standardized score of the nth physical property index, and m be the total number of physical property indexes.

[0030] Preprocessing includes image grayscale conversion, histogram equalization, median-based noise reduction, and contrast enhancement;

[0031] Image grayscale conversion uses a weighted method to convert the original color image into a single-channel grayscale image, simplifying the data dimensions and reducing the computational load for subsequent processing (such as filtering and segmentation).

[0032] Histogram equalization redistributes the pixel intensity values ​​of a grayscale image to cover a wider grayscale range and improves global contrast (e.g., making dark grain details clearer).

[0033] Noise reduction involves setting the gray value of each pixel to the median of the gray values ​​of all pixels within a neighborhood window (e.g., point noise caused by dust in feed production). Gaussian filtering can also be used to suppress Gaussian noise. This method uses a Gaussian kernel to perform convolution with the image, replacing the value of each pixel with the weighted average of its neighboring pixels. The weights follow a Gaussian distribution (with the highest weight at the center), thus smoothing the image and blurring details to ensure that the preprocessed image is both clean and retains key features (particle edges, shapes).

[0034] Contrast enhancement expands the dynamic range of an image's grayscale levels through linear or non-linear transformations, with an adjustment range of 50%-200%, finely adjusting the differences in brightness and darkness of the image (such as making the boundary between grains and the background more prominent).

[0035] Binarization uses the Otsu thresholding method or adaptive thresholding method to convert a grayscale image into a binary image containing only black and white, and segments it into foreground and background, supplemented by morphological opening or closing operations to optimize the grain contour.

[0036] Contour detection specifically involves finding and extracting a sequence of pixels representing the boundaries of an object in a binary image. A commonly used algorithm is the Suzuki 85 boundary tracking algorithm, which traverses the image and, when it encounters a foreground pixel, tracks its boundary according to certain rules (such as clockwise direction) until it returns to the starting point, thus obtaining a closed contour. Each contour consists of a series of point coordinates.

[0037] The matching logic of the classification rules is that all conditions must be met; when classifying particles in batches, the first matching priority principle is adopted. When a particle matches all the conditions of a rule, it is marked as the category corresponding to that rule, and the checking of subsequent rules stops.

[0038] The rules are specifically composed of user-defined classification categories (such as "qualified particles", "extra-large particles", "fine powder", "irregular particles" etc.) and one or more judgment conditions. Each condition is built based on the geometric or morphological parameters of the particles, allowing users to flexibly set them according to specific business needs.

[0039] The physical properties include particle size and particle size distribution, density, morphology and structural characteristics, mechanical strength, moisture content, and mixing uniformity, with contribution values ​​of 0.25, 0.25, 0.1, 0.1, 0.15, and 0.15, respectively. The average particle size (geometric mean particle size), particle size distribution range (whether it is too concentrated or dispersed), and fine powder ratio (<0.5 mm) are also considered. Morphology and structural characteristics include the length, diameter, aspect ratio, surface roughness, and integrity of powder / granules / tablets / expanded materials. Mechanical strength includes hardness, compressive strength, crush resistance, and pulverization rate (after transportation and mixing). Mixing uniformity is used to evaluate whether components of different particle sizes and densities are prone to stratification.

[0040] The data interaction module is used to transmit the analysis results to the production management system in real time to complete the closed-loop feedback control of production parameters; specifically, when the comprehensive evaluation score A or the proportion of a specific particle category exceeds the preset threshold, the speed of the feed grinder or the pressure of the pellet mill is adjusted.

[0041] The visualization module provides interactive operations and information display based on the visual user interface. Specifically, the visual user interface is used to provide functions such as real-time image display, visual parameter adjustment, custom classification rules, particle distribution curve display, and abnormal alarm (such as triggering an alarm when the proportion of particles exceeding the threshold is >5%).

[0042] It also includes data management, using a MySQL database to store test data (including time, batch, particle parameters, etc.), supports historical data query, trend analysis and quality traceability, and has a data storage capacity of ≥1 million records.

[0043] A method for constructing a system for rapid, contactless acquisition and model building of livestock feed properties includes the following steps:

[0044] Step 1: Use standard particle samples of known size to calibrate the system, establish the conversion relationship between pixel size and actual physical size, and build an error compensation model to correct the measured values. Then, acquire the original image of the feed sample through the image acquisition module.

[0045] Step 2: The original image is preprocessed by grayscale conversion, noise reduction and contrast enhancement through the processing control module. The preprocessed image is then binarized and segmented to extract the outline of the feed particles.

[0046] Step 3: Filter valid particles based on the contour area threshold and calculate the geometric feature parameters of each particle.

[0047] Step 4: Apply the preset classification rules to automatically classify the particles and count the quantity and proportion of each category. Then, based on the extracted physical property parameters and classification statistics, calculate the comprehensive evaluation score A.

[0048] Step 5: Output the evaluation results through the data interaction module, and use the results to implement feedback control on the production process.

[0049] The measurement experiment was conducted according to the above method, and the results were verified. The data were compared with those measured manually using traditional methods. The samples were divided into 7 groups for measurement, as shown in the table below:

[0050]

[0051] Analysis: The experimental measurements deviated little from the manual measurements, indicating good accuracy.

[0052] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for rapid, contactless acquisition and model building of livestock feed properties, characterized in that: The system includes an image acquisition module, a processing and control module, a data interaction module, and a visualization module. The image acquisition module uses a high-precision camera to acquire raw image data of feed pellets, including two-dimensional color images and three-dimensional depth information. The processing and control module processes and analyzes the raw image data based on image processing algorithms and evaluation models to extract the physical property parameters of the feed pellets. The data interaction module is used to transmit the analysis results to the production management system in real time to complete the closed-loop feedback control of production parameters. The visualization module provides interactive operation and information display based on a visual user interface.

2. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 1, characterized in that: The processing control module specifically preprocesses, binarizes, and performs contour detection on the original image data, identifies and separates individual feed particles, and then calculates the geometric feature parameters of each feed particle, including area, perimeter, roundness, and aspect ratio. Based on user-preset multi-condition classification rules, it classifies and statistically analyzes the identified feed particles according to the formula A=ΣmnX. n *F n Calculate the overall assessment score A for feed properties, where X n F is the weighting coefficient for the nth physical property index. n Let m be the standardized score of the nth physical property index, and m be the total number of physical property indexes.

3. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 2, characterized in that: The preprocessing includes image grayscale conversion, histogram equalization, median-based denoising, and contrast enhancement. Image grayscale conversion uses a weighted method to convert the original color image into a single-channel grayscale image. Histogram equalization redistributes the pixel intensity values ​​of the grayscale image to cover a wider grayscale range. Denoising sets the grayscale value of each pixel to the median of the grayscale values ​​of all pixels within a neighborhood window of that pixel. Contrast enhancement expands the dynamic range of the image's grayscale levels through linear or nonlinear transformations, with an adjustment range of 50%-200%.

4. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 2, characterized in that: The binarization process involves using the Otsu thresholding method or adaptive thresholding method to convert a grayscale image into a binary image containing only black and white, and segmenting it into foreground and background. Morphological opening or closing operations are then used to optimize the particle contours. Contour detection specifically involves finding and extracting a sequence of pixels representing the boundaries of an object in the binary image.

5. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 2, characterized in that: The matching logic of the classification rules is that all conditions must be met; when classifying particles in batches, the first matching priority principle is adopted. When a particle matches all the conditions of a rule, it is marked as the category corresponding to that rule, and the subsequent rule checks are stopped.

6. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 2, characterized in that: The physical properties include particle size and particle size distribution, density, morphological and structural characteristics, mechanical strength, moisture content, and mixing uniformity.

7. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 1, characterized in that: The closed-loop feedback control specifically adjusts the speed of the feed grinder or the pressure of the pellet mill when the comprehensive evaluation score A or the proportion of a specific particle category exceeds a preset threshold.

8. The system for rapid, contactless acquisition and model building of livestock feed properties according to claim 1, characterized in that: The visual user interface is specifically used to provide real-time image display, visual parameter adjustment, customizable classification rules, particle distribution curve display, and abnormal alarm functions.

9. A method for constructing a system for rapid, contactless acquisition and model building of livestock feed properties according to any one of claims 1-8, characterized in that, Includes the following steps: Step 1: Acquire the original image of the feed sample using the image acquisition module; Step 2: The original image is preprocessed by grayscale conversion, noise reduction and contrast enhancement through the processing control module. The preprocessed image is then binarized and segmented to extract the outline of the feed particles. Step 3: Filter valid particles based on the contour area threshold and calculate the geometric feature parameters of each particle; Step 4: Apply the preset classification rules to automatically classify the particles and count the quantity and proportion of each category. Then, based on the extracted physical property parameters and classification statistics, calculate the comprehensive evaluation score A. Step 5: Output the evaluation results through the data interaction module, and use the results to implement feedback control on the production process.

10. The method for constructing a contactless rapid acquisition and model building system for livestock feed properties according to claim 9, characterized in that: Before acquiring the original image in step one, the system is calibrated using standard particle samples of known size to establish the conversion relationship between pixel size and actual physical size, and an error compensation model is constructed to correct the measured value.

Images