CN-121980315-A - Dedicated intelligent free-range monitoring system of ecological chicken raising

CN121980315ACN 121980315 ACN121980315 ACN 121980315ACN-121980315-A

Abstract

The invention discloses an intelligent free-range monitoring system special for ecological chicken raising, which comprises an image processing module, a behavior recognition module, an environment acquisition module and a fusion calculation module, wherein the image processing module comprises a plurality of groups of image acquisition units used for acquiring original image data, the image processing module is used for carrying out image denoising, illumination correction and background modeling processing on the original image data and outputting a standardized image frame sequence used for feature extraction, the behavior recognition module is used for recognizing the standardized image frame sequence based on a pre-trained multi-scale convolutional neural network structure to obtain a recognition result aiming at a chicken flock, the environment acquisition module is used for acquiring climate environment parameters, harmful environment parameters and light environment parameters in a chicken flock environment, and the fusion calculation module is used for carrying out time sequence synchronization and data fusion analysis on the recognition result and the parameters acquired by the environment acquisition module.

Inventors

ZHANG XI
SUN YUEHUA
SUN JIANXIANG
GUO XIAOHUI

Assignees

甘肃畜牧工程职业技术学院

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (9)

1. Dedicated intelligent free-range monitoring system of ecological chicken raising, characterized by comprising: The system comprises an image processing module, a behavior recognition module, an environment acquisition module and a fusion calculation module; The image processing module comprises a plurality of groups of image acquisition units which are arranged at different positions of the farm and are used for acquiring original image data of chicken flocks in the farm, and the image processing module is used for carrying out image denoising, illumination correction and background modeling processing on the original image data and outputting a standardized image frame sequence for feature extraction; The behavior recognition module is used for recognizing the standardized image frame sequence based on a pre-trained multi-scale convolutional neural network structure to obtain a recognition result aiming at the chicken flock, wherein the recognition result comprises quantity distribution, motion trail, behavior category and abnormal gesture information of the chicken flock in the cultivation field; The environment acquisition module comprises a plurality of sensor nodes and a wireless transmission unit and is used for acquiring climate environment parameters, harmful environment parameters and light environment parameters in the environment of the farm; The fusion calculation module is used for carrying out time sequence synchronization and data fusion analysis on the identification result and the parameters acquired by the environment acquisition module, and sending out early warning information when the abnormal state in the farm is analyzed.
2. The intelligent free-range monitoring system special for ecological chicken raising according to claim 1, wherein the image acquisition unit comprises a combined installation structure of a visible light camera and a thermal imager, the combined installation structure is provided with an integrated bracket and an adjustable angle rotating mechanism, and the visible light camera has a night infrared light supplementing function and is used for day and night continuous image acquisition and heat map acquisition of a body temperature abnormal region.
3. The intelligent free-range monitoring system for ecological chickens according to claim 1, wherein the image processing module further comprises an image timing synchronization unit for performing timestamp alignment and view registration processing on the raw image data acquired by the image acquisition unit to obtain a bimodal image of a corresponding frame in the standardized image frame sequence.
4. The intelligent free-range monitoring system dedicated to ecological chickens according to claim 1, wherein the image processing module further comprises a foreground extraction unit and a target screening unit; The foreground extraction unit is used for calling a pre-constructed sample library to generate a background model of the original image data, and removing data corresponding to the background model from the original image data to obtain foreground data; The target screening unit is used for carrying out smoothing treatment and hole filling on the foreground data, and removing non-chicken interference objects from the foreground data according to the target area range, the aspect ratio and the number of connected areas in the foreground.
5. The intelligent free-range monitoring system dedicated for ecological chicken raising as claimed in claim 4, wherein said foreground extraction unit is specifically configured to: invoking standard scene images without chicken states, which are stored in different positions of the farm in the sample library, and performing probability modeling on pixel brightness and color distribution of the farm based on the standard scene images by using a Gaussian mixture model to generate a background model; For each frame of image in the original image data, calculating the difference value of pixels in the image and pixels in the background model, marking the pixels with the calculated difference value not exceeding a preset threshold value as background pixels, and removing data corresponding to the background pixels from the original image data to obtain foreground data.
6. The intelligent free-range monitoring system dedicated to ecological chickens according to claim 4, wherein the target screening unit is further configured to: extracting an outer boundary contour of a candidate target reserved according to the target area range, the aspect ratio and the number of connected regions; generating a fitting contour of the candidate target based on the outer boundary contour by adopting a polygonal curve fitting algorithm; and calculating a smoothness index of the fitting contour, and determining that the candidate target is a non-chicken interference object and removing the candidate target from the foreground data under the condition that the smoothness index is lower than a preset threshold.
7. The intelligent free-range monitoring system for ecological chickens according to claim 1, wherein the behavior recognition module is specifically configured to: carrying out multi-scale convolution feature extraction on each frame image in the standardized image frame sequence to obtain a special sheet image of each frame image; generating a plurality of rectangular detection frames on the feature map of each frame of image by using an anchor frame regression algorithm, calculating classification probabilities of two types of chicken and non-chicken in each rectangular detection frame, and obtaining pixel coordinates of the upper left corner and the lower right corner of each detection frame; Outputting the rectangular detection frame with the classification probability larger than a preset threshold value as a candidate chicken object, inputting a local image area corresponding to the rectangular detection frame of the candidate chicken object into a classification network adopting a MobileNet structure, and outputting a behavior category confidence score of the candidate chicken object; and in each frame of image, taking the behavior type with the highest confidence score of the candidate chicken object as the recognition result of the behavior type in the frame of image.
8. The intelligent free-range monitoring system special for ecological chicken raising according to claim 1, wherein the fusion calculation module is deployed in an edge gateway device, and a calculation chip with a central processing unit, a neural network processing unit and a field programmable gate array heterogeneous acceleration unit is integrated in the edge gateway device, and is used for performing rapid fusion analysis and preliminary classification judgment on the identification result and the parameters acquired by the environment acquisition module.
9. The intelligent free-range monitoring system for ecological chickens according to claim 8, further comprising a cloud platform communicatively connected to the edge gateway device, wherein the edge gateway device uploads the results of the fusion analysis to the cloud platform via cellular or LoRa communication, and performs chicken flock behavioral trend analysis, recognition model online iteration, and/or remote parameter issuing operations on the cloud platform.

Description

Dedicated intelligent free-range monitoring system of ecological chicken raising Technical Field The invention relates to the field of pattern recognition based on image analysis, in particular to an intelligent free-range monitoring system special for ecological chicken raising. Background In an ecological free-range scene, the traditional chicken flock monitoring usually depends on manual observation or a single sensing means (such as environmental temperature and humidity sensing), so that the behavior state and the spatial distribution of the chicken flock are difficult to cover, the problems of monitoring blind areas and response lag exist, and particularly, accurate information cannot be acquired under the conditions of illumination change and night. The Chinese patent CN120298143A discloses a cage chicken abnormal behavior detection early warning system based on image analysis, and the technical scheme comprises image sequence acquisition, target detection and behavior pattern matching, and is used for identifying abnormal behaviors of chicken flocks in cages and giving early warning. However, the system focuses on static individual behaviors in cages, lacks analysis of spatial distribution, aggregation behaviors and environmental parameter correlation of chicken flocks in a free-range environment, and is difficult to realize comprehensive monitoring of the population activity state in an ecological free-range. Disclosure of Invention The invention provides an intelligent free-range monitoring system special for ecological chicken raising, which aims to solve the technical problem that the comprehensive monitoring of the group activity state in an ecological free-range farm is difficult to realize in the prior art. In order to solve the technical problems, the invention provides the following technical scheme: the invention provides an intelligent free-range monitoring system special for ecological chicken raising, which comprises the following components: The system comprises an image processing module, a behavior recognition module, an environment acquisition module and a fusion calculation module; The image processing module comprises a plurality of groups of image acquisition units which are arranged at different positions of the farm and are used for acquiring original image data of chicken flocks in the farm, and the image processing module is used for carrying out image denoising, illumination correction and background modeling processing on the original image data and outputting a standardized image frame sequence for feature extraction; The behavior recognition module is used for recognizing the standardized image frame sequence based on a pre-trained multi-scale convolutional neural network structure to obtain a recognition result aiming at the chicken flock, wherein the recognition result comprises quantity distribution, motion trail, behavior category and abnormal gesture information of the chicken flock in the cultivation field; The environment acquisition module comprises a plurality of sensor nodes and a wireless transmission unit and is used for acquiring climate environment parameters, harmful environment parameters and light environment parameters in the environment of the farm; The fusion calculation module is used for carrying out time sequence synchronization and data fusion analysis on the identification result and the parameters acquired by the environment acquisition module, and sending out early warning information when the abnormal state in the farm is analyzed. In an alternative embodiment, the image acquisition unit comprises a combined installation structure of a visible light camera and a thermal imager, the combined installation structure is provided with an integrated bracket and an angle-adjustable rotating mechanism, and the visible light camera has a night infrared light supplementing function and is used for performing day and night continuous image acquisition and heat map acquisition of a body temperature abnormal region. In an optional embodiment, the image processing module further includes an image timing synchronization unit, configured to perform timestamp alignment and perspective registration processing on the raw image data acquired by the image acquisition unit, to obtain a bimodal image of a corresponding frame in the standardized image frame sequence. In an alternative embodiment, the image processing module further includes a foreground extraction unit and a target screening unit; The foreground extraction unit is used for calling a pre-constructed sample library to generate a background model of the original image data, and removing data corresponding to the background model from the original image data to obtain foreground data; The target screening unit is used for carrying out smoothing treatment and hole filling on the foreground data, and removing non-chicken interference objects from the foreground data according to the target area range, the a