CN-122023658-A - System and method for non-contact rapid acquisition and model construction of physical properties of livestock breeding feed

CN122023658ACN 122023658 ACN122023658 ACN 122023658ACN-122023658-A

Abstract

The invention discloses a non-contact rapid acquisition and model construction system for physical properties of livestock breeding feeds, which comprises an image acquisition module for acquiring raw image data of feed particles, wherein the raw image data comprises two-dimensional color images and three-dimensional depth information, a processing control module for processing and analyzing the raw image data and extracting physical property parameters of the feed particles, and a data interaction module for transmitting analysis results to a production management system in real time to complete closed-loop feedback control of production parameters, and a visualization module for performing interactive operation and information display.

Inventors

LEI KAIDONG
LI YUECHAO
LIU XIANGLONG
ZHANG SHIQI
GUO ZHENZHEN
LI KUNPENG

Assignees

山西农业大学

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The system is characterized by comprising an image acquisition module, a processing control module, a data interaction module and a visualization module, wherein the image acquisition module acquires raw image data of feed particles, including two-dimensional color images and three-dimensional depth information, based on a high-precision camera, the processing control module processes and analyzes the raw image data based on an image processing algorithm and an evaluation model and extracts physical parameters of the feed particles, the data interaction module is used for transmitting analysis results to a production management system in real time to complete closed-loop feedback control of production parameters, and the visualization module performs interaction operation and information display based on a visual user interface.
2. The system for quickly acquiring and constructing a model without contact on physical properties of livestock feed according to claim 1, wherein the processing control module is specifically configured to perform preprocessing, binarization and contour detection on raw image data, identify and separate individual feed particles, then calculate geometric characteristic parameters including area, perimeter, circularity and aspect ratio of each feed particle, classify and count the identified feed based on a multi-condition classification rule preset by a user, calculate a comprehensive evaluation score a of physical properties of the feed according to a formula a=Σmnx n *F n , wherein X n is a weight coefficient of an nth physical property index, F n is a standardized score of the nth physical property index, and m is a total number of physical property indexes.
3. The system for quickly acquiring and constructing the model without contact on the physical properties of the livestock feed according to claim 2, wherein the preprocessing comprises image graying, histogram equalization, noise reduction processing based on median filtering and contrast enhancement, wherein the image graying is to convert a color image original image into a single-channel gray image by adopting a weighting method, the histogram equalization is to redistribute pixel intensity values of the gray image to cover a wider gray range, the noise reduction processing is to set the gray value of each pixel point as the median value of gray values of all pixel points in a certain neighborhood window of the point, the contrast enhancement is to expand the dynamic range of the gray level of the image by linear or nonlinear transformation, and the adjustment range is 50% -200%.
4. The system for quickly acquiring and constructing the model without contact on the physical properties of the livestock feed according to claim 2, wherein the binarization is to convert a gray level image into a binary image only comprising black and white colors by adopting an Otsu threshold method or an adaptive threshold method, divide the binary image into a foreground and a background, and optimize the particle contour by assisting with morphological open operation or closed operation, and the contour detection is to find and extract a pixel point sequence representing the boundary of an object in the binary image.
5. The system for quickly acquiring and constructing the model without contact on the physical properties of the livestock feed according to claim 2, wherein the matching logic of the classification rule is that all conditions must be met, when the particles are classified in batches, the first-time matching priority principle is adopted, and when one particle matches all conditions of a certain rule, the particle is marked as the category corresponding to the rule, and the subsequent rule inspection is stopped.
6. The system for quickly obtaining and constructing a model without contact for physical properties of livestock feed according to claim 2, wherein the physical properties comprise particle size and particle size distribution, density, morphology and structural characteristics, mechanical strength, water content and mixing uniformity.
7. The system for the non-contact rapid acquisition and model construction of physical properties of livestock feed according to claim 1, wherein the closed loop feedback control is specifically to adjust the feed grinder speed or the granulator pressure when the overall evaluation score A or the specific particle classification ratio exceeds a preset threshold.
8. The system for non-contact rapid acquisition and model construction of livestock feed physical properties of claim 1, wherein the visual user interface is specifically used for providing functions of image real-time display, parameter visual adjustment, classification rule customization, particle distribution curve display and abnormality alarm.
9. The method for constructing a system for the contactless rapid acquisition and model construction of the physical properties of livestock raising feeds according to any one of claims 1 to 8, characterized by comprising the steps of: Step one, acquiring an original image of a feed sample through an image acquisition module; Step two, preprocessing the original image through a processing control module, performing binarization segmentation on the preprocessed image, and extracting the outline of the feed particles; step three, screening effective particles based on a contour area threshold value, and calculating geometric characteristic parameters of each particle; Step four, applying a preset classification rule to automatically classify particles, counting the number and the duty ratio of each class, and then calculating a comprehensive evaluation score A based on the extracted physical parameters and the classification statistics result; and fifthly, outputting an evaluation result through the data interaction module, and performing feedback control on the production process based on the result.
10. The method for constructing a system for the non-contact rapid acquisition and model construction of physical properties of livestock feed according to claim 9, wherein the step one is to calibrate the system by using standard particle samples of known dimensions, establish a conversion relationship between pixel dimensions and actual physical dimensions, and construct an error compensation model to correct the measured values before the step one is to acquire the original images.

Description

System and method for non-contact rapid acquisition and model construction of physical properties of livestock breeding feed Technical Field The invention relates to the technical field of feed processing, in particular to a system and a method for non-contact rapid acquisition and model construction of livestock breeding feed physical properties. Background The modern animal husbandry is developed to large scale and intensive, the granularity of the feed directly influences the nutrition absorption, health and production performance of animals, and the granularity requirements of different livestock (such as piglets and cows) and growth stages are obviously different. The current mainstream detection means rely on the traditional screening method, the efficiency is low (15-30 minutes are needed for a single sample), the error is large (+ -8% -12%), the above-mentioned refined requirements cannot be matched accurately, the problems of strong artificial dependence, difficult data tracing and the like exist, foreign automation equipment (such as a high-end imaging system) is high in price (more than 50 ten thousand yuan), and is not suitable for the domestic high-dust and high-humidity environment, the domestic technology is mostly stopped in a laboratory stage, and an industrial solution is lacking. The industrial upgrading requirement is urgent, and a closed-loop system for realizing granularity online detection and real-time feedback is needed to accurately match the animal nutrition refinement requirement, so that the invention provides a system and a method for quickly acquiring physical properties of livestock breeding feeds in a non-contact manner and constructing a model to solve the problems in the prior art. Disclosure of Invention Aiming at the problems, the invention aims to provide a system and a method for quickly acquiring and constructing a model of livestock breeding feed physical properties in a non-contact way, the system and the method for quickly acquiring and constructing the model of the physical properties of the livestock breeding feed in a non-contact way realize the quick and automatic detection of the physical properties of the feed particles by a non-contact image analysis technology, and effectively overcome the limitations of low efficiency and lagging result of the traditional method. The invention aims at realizing the technical scheme that the system for quickly acquiring and constructing the physical properties of the livestock breeding feed in a non-contact manner comprises an image acquisition module, a processing control module, a data interaction module and a visualization module, wherein the image acquisition module acquires raw image data of feed particles, which comprises a two-dimensional color image and three-dimensional depth information, based on a high-precision camera, the processing control module processes and analyzes the raw image data based on an image processing algorithm and an evaluation model and extracts physical property parameters of the feed particles, the data interaction module is used for transmitting analysis results to a production management system in real time to complete closed-loop feedback control of production parameters, and the visualization module performs interaction operation and information display based on a visual user interface. The processing control module is used for preprocessing, binarizing and contour detecting the original image data, identifying and separating single feed particles, then calculating geometric characteristic parameters of each feed particle including area, perimeter, circularity and length-width ratio, classifying and counting the identified feed based on a multi-condition classification rule preset by a user, and calculating comprehensive evaluation score A of feed physical properties according to a formula A=ΣmnX n*Fn, wherein X n is a weight coefficient of an nth physical property index, F n is a standardized score of the nth physical property index, and m is the total number of physical property indexes. The preprocessing comprises image graying, histogram equalization, noise reduction processing based on median filtering and contrast enhancement, wherein the image graying is to convert a color image original image into a single-channel gray image by adopting a weighting method, the histogram equalization is to redistribute pixel intensity values of the gray image to cover a wider gray scale range, the noise reduction processing is to set the gray scale value of each pixel point to be the median value of gray scale values of all pixel points in a certain neighborhood window of the point, the contrast enhancement is to expand the dynamic range of the gray scale of the image by linear or nonlinear transformation, and the adjustment range is 50% -200%. The method is further improved in that the binarization is to convert a gray level image into a binary image only containing black and white colors by adop