CN112862757A - Weight evaluation system based on computer vision technology and implementation method - Google Patents
Weight evaluation system based on computer vision technology and implementation method Download PDFInfo
- Publication number
- CN112862757A CN112862757A CN202110047484.5A CN202110047484A CN112862757A CN 112862757 A CN112862757 A CN 112862757A CN 202110047484 A CN202110047484 A CN 202110047484A CN 112862757 A CN112862757 A CN 112862757A
- Authority
- CN
- China
- Prior art keywords
- animal
- image
- weight
- camera
- ear tag
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000011156 evaluation Methods 0.000 title claims abstract description 26
- 238000005516 engineering process Methods 0.000 title claims abstract description 20
- 241001465754 Metazoa Species 0.000 claims abstract description 99
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims description 28
- 230000005540 biological transmission Effects 0.000 claims description 20
- 230000011218 segmentation Effects 0.000 claims description 17
- 238000005259 measurement Methods 0.000 claims description 14
- 238000003709 image segmentation Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000037396 body weight Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 241000894007 species Species 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000002146 bilateral effect Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 125000001475 halogen functional group Chemical group 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000013178 mathematical model Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 241000283690 Bos taurus Species 0.000 description 3
- 241000282887 Suidae Species 0.000 description 3
- 230000001276 controlling effect Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000005303 weighing Methods 0.000 description 3
- 241000287828 Gallus gallus Species 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 235000013330 chicken meat Nutrition 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 241000272517 Anseriformes Species 0.000 description 1
- 241001494479 Pecora Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000008571 general function Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 244000144972 livestock Species 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 244000144977 poultry Species 0.000 description 1
- 235000013594 poultry meat Nutrition 0.000 description 1
- 238000003825 pressing Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses a weight evaluation system based on computer vision technology and an implementation method thereof. Secondly, the invention obtains parameters by processing the animal body through images, then fits with the known weight data to obtain a mathematical model and stores the mathematical model in a database, and when in practical application, the system identifies the type of the animal body and calls a related model to calculate and obtain the estimated weight data of the animal, thereby effectively solving the problem that the traditional weight scale obtains the weight of the animal, and because the individual animal is not completely controllable, the operation is more complicated, the invention can realize contactless animal weight evaluation, reduce the dependence on the traditional weight scale, and has reliable prediction result and high efficiency.
Description
Technical Field
The invention relates to the technical field of animal body weight measurement, in particular to a weight evaluation system based on a computer vision technology and an implementation method.
Background
In the feeding process of animals such as farm pigs, cattle and sheep, the animals are weighed by adopting an electronic scale mode at present. The electronic scale weighing process animal does not cooperate, wastes time and energy, and the operation is comparatively loaded down with trivial details. The electronic scale is easily corroded by dirt and has a limited service life. Meanwhile, results need to be recorded manually, animal numbers are correlated, and errors are prone to occurring.
Disclosure of Invention
The invention aims to provide a weight evaluation system based on a computer vision technology and an implementation method thereof, and the weight evaluation system can effectively solve the problems of uncooperative animals, time and labor waste and complex operation in the weighing process of the traditional electronic scale by adopting a mode of taking pictures to obtain the weight.
In order to achieve the purpose, the invention provides the following technical scheme:
a weight evaluation system based on computer vision technology and an implementation method thereof comprise a central processing unit, a memory, a wireless transmission module, an ear tag module and a distance sensor; the central processing unit, the memory, the wireless transmission module, the ear tag module and the distance sensor are integrated in a handheld shell device, and a display screen, a camera and a key for controlling operation are arranged on the shell device;
the display screen is used for displaying the shot pictures, and the calculation result and the human-computer interaction related to the user setting; the camera is used for shooting an animal picture; the distance sensor is used for detecting the distance between the camera and the animal when the animal picture is shot so as to correct the actual size of the shot animal conveniently; the wireless transmission module is used for uploading the measurement result to a cloud server or used for downloading equipment parameters and updating a firmware program; the memory is used for storing the measurement result to the local equipment; the ear tag module is used for reading the unique identification code of the animal and correlating the unique identification code with the measurement result; the central processing unit is used as a core processing unit of the whole system and is respectively connected with the memory, the wireless transmission module, the ear tag module, the distance sensor, the display screen, the camera and the key, and the weight estimation is carried out through an animal image acquired by the camera by using an algorithm.
The invention provides another technical scheme: a weight evaluation system based on computer vision technology and an implementation method of the weight evaluation system based on computer vision technology comprise the following steps:
s1: taking pictures and measuring distance: a user triggers a camera to shoot an animal image through a key and displays the animal image on a display screen; meanwhile, the distance sensor measures the distance of the tested animal and displays the distance on the display screen;
s2: image segmentation: inputting the image obtained in the step S1 into a Mask-RCNN neural network which is trained to obtain an animal image after binary segmentation and an animal category;
s3: extracting animal parameters: extracting three-dimensional structure information of the animal including parameters of body length, chest circumference and body height according to the binarized image, and calculating actual values of the parameters of body length, chest circumference and body height according to the distance measured by S1;
s4: inputting the actual animal parameter values obtained in the step S3 into a pre-trained-RBF neural network according to the animal type to obtain a weight estimation value of the shot animal;
s5: scanning an ear tag of the animal through an ear tag module to obtain a unique identification code of the animal, and associating the unique identification code with the weight value obtained in S4;
s6: the result obtained by the S5 is stored in the system through the memory, so that the subsequent viewing is convenient; meanwhile, the measurement result can be transmitted to the cloud server through the wireless transmission module, and large-scale data automatic management is facilitated.
Further, the preprocessing step of the animal image taken by the camera in S1 is as follows:
the method comprises the steps of firstly realizing graying of an image, then improving the definition and contrast of the image by utilizing image histogram equalization, then realizing image denoising by utilizing bilateral filtering, and then solving the problems of halo artifacts and reduction of hue and color saturation by utilizing an image adaptive enhancement algorithm based on Retinex theory, thereby increasing the image usability.
Furthermore, in S2, a Mask-RCNN neural network is used, and a branch network is added on the basis of the fast-RCNN, so that the target detection is realized, and simultaneously, the target instance is segmented and identified.
Further, in S3 and S4, by establishing a regression model for weight estimation, the RBF neural network is used to obtain the relationship between the body size parameter of the species and the weight according to the body size parameter extracted from the image.
Further, the specific method in S6 includes:
building a database: by renting a server, building a database on the server, and putting historical data into the database, historical estimation records of a user can be put into the database in real time, and the server can also access the database in real time to continuously perfect a weight estimation model;
uploading the picture to a server to obtain the estimated weight: the user uploads the picture of the measured animal through a handheld terminal consisting of a central processing unit, a memory, a wireless transmission module, an ear tag module, a distance sensor, a display screen, a camera and a key, and the server outputs the estimated weight of the measured animal in the picture.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the weight evaluation system based on the computer vision technology and the implementation method, the central processing unit, the memory, the wireless transmission module, the ear tag module, the distance sensor, the display screen, the camera and the key for controlling operation can be integrally arranged in the shell equipment which can be held by hand, and the weight evaluation system has the advantages of small and light handheld equipment, convenience in carrying and operation and the like.
2. According to the weight evaluation system based on the computer vision technology and the implementation method, the distance sensor is used, the distance of the shot animal can be automatically measured, and the operation of size calibration is avoided.
3. According to the weight evaluation system based on the computer vision technology and the implementation method, the Mask-RCNN neural network is used for carrying out target segmentation on the animal image, the target segmentation precision is improved, meanwhile, the animal species are automatically identified, and the animal type selection is avoided.
4. According to the weight evaluation system based on the computer vision technology and the implementation method thereof, the animal ear tag is used as the unique identification code of the read animal and is automatically associated with the estimated animal weight, so that manual input errors are avoided, and large-scale data automatic management is facilitated.
5. According to the weight evaluation system based on the computer vision technology and the implementation method, the mode that the handheld device shoots to estimate the weight of the animal is adopted, various problems caused by the fact that the animal is not matched can be effectively avoided, compared with the traditional method, the handheld device is used, resources are saved in time and space, the weight evaluation system is light, fast and effective, and production efficiency can be greatly improved.
Drawings
FIG. 1 is a system block diagram of the present invention;
FIG. 2 is an overall framework diagram of Mask-RCNN of the present invention;
FIG. 3 is a diagram illustrating the effect of using RolAlign to process misalignment between a mask and an object in an original image according to the present invention;
fig. 4 is a schematic diagram of the bin of division 7 × 7 of the present invention;
FIG. 5 is a diagram of the bilinear interpolation algorithm performed on each bin of FIG. 4 in accordance with the present invention;
FIG. 6 is a graph of the results of the algorithm of FIG. 5 in accordance with the present invention;
FIG. 7 is a diagram of an RBF neural network architecture according to the present invention;
FIG. 8 is a flow chart of the RBF neural network algorithm of the present invention.
In the figure: 1. a central processing unit; 2. a memory; 3. a wireless transmission module; 4. an ear tag module; 5. a distance sensor; 6. a display screen; 7. a camera; 8. and (6) pressing a key.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, in the embodiment of the present invention: a weight evaluation system based on computer vision technology and its implementation method, including central processing unit 1, memorizer 2, wireless transmission module 3, ear tag module 4 and distance sensor 5; the central processing unit 1, the memory 2, the wireless transmission module 3, the ear tag module 4 and the distance sensor 5 are integrally arranged in the shell equipment which can be held by hands, and a display screen 6, a camera 7 and a key 8 for controlling operation are arranged on the shell equipment.
In the above embodiment, the display screen 6 is used for displaying the shot picture, and the calculation result and the human-computer interaction related to the user setting; the camera 7 is used for shooting an animal picture; the distance sensor 5 is used for detecting the distance between the camera 7 and the animal when the animal picture is shot so as to conveniently correct the actual size of the shot animal; the wireless transmission module 3 is used for uploading the measurement result to a cloud server, or is used for downloading equipment parameters and updating a firmware program; the memory 2 is used for storing the measurement result to the local equipment; the ear tag module 4 is used for reading the unique identification code of the animal and correlating the unique identification code with the measurement result; the central processing unit 1 is used as a core processing unit of the whole system, is respectively connected with the memory 2, the wireless transmission module 3, the ear tag module 4, the distance sensor 5, the display screen 6, the camera 7 and the key 8, and estimates the weight of the animal image acquired by the camera 7 by using an algorithm.
In order to better explain the above embodiments, the present invention further provides a weight evaluation system based on computer vision technology and an implementation method thereof, comprising the following steps:
the first step is as follows: taking pictures and measuring distance: a user triggers the camera 7 to shoot an animal image through the key 8 and displays the animal image on the display screen 6; meanwhile, the distance sensor 5 measures the distance of the tested animal and displays the distance on the display screen 6;
the second step is that: image segmentation: inputting the image obtained in the first step into a trained Mask-RCNN neural network to obtain an animal image subjected to binary segmentation and an animal category;
the third step: extracting animal parameters: extracting three-dimensional structure information of the animal including parameters of body length, chest circumference and body height according to the binarized image, and calculating actual values of the parameters of body length, chest circumference and body height according to the distance measured in the first step;
the fourth step: inputting the actual parameter values of the animals obtained in the third step into a pre-trained RBF neural network according to the animal types to obtain weight estimation values of the shot animals;
the fifth step: scanning an ear tag of the animal through the ear tag module 4 to obtain a unique identification code of the animal, and associating the unique identification code with the weight value obtained in the fourth step;
and a sixth step: the result obtained in the fifth step is stored in the system through the memory 2, so that the subsequent checking is convenient; meanwhile, the measurement result can be transmitted to a cloud server through the wireless transmission module 3, so that large-scale data automatic management is facilitated.
In the first step, in the image analysis, the quality of the image directly affects the subsequent processing effect on the image; when the camera 7 is used for shooting an animal image, the image can be interfered by environmental factors such as illumination, noise and the like, even if the same animal has a large difference in the image collected under different environments at the same shooting angle, and the non-contact shooting weight estimation faces to different animals under different environments, the influence of the environment on the image difference is more remarkable, so that the invention needs to carry out image preprocessing on the collected image after deletion and selection to eliminate irrelevant information in the image, recover useful real information, enhance the detectability of relevant information and simplify the estimation data to the maximum extent, thereby ensuring the reliability of the subsequent process according to the weight of the image; the pretreatment steps are as follows:
the method comprises the steps of firstly realizing graying of an image, then improving the definition and contrast of the image by utilizing image histogram equalization, then realizing image denoising by utilizing bilateral filtering, and then solving the problems of halo artifacts and reduction of hue and color saturation by utilizing an image adaptive enhancement algorithm based on Retinex theory, thereby increasing the image usability.
In the second step, the image segmentation is a part which is crucial to image recognition and computer vision, and the image segmentation is not correctly recognized without correct segmentation; the image segmentation has great significance to the non-contact photographing weight estimation technology, and can be embodied in two aspects; in a first aspect, identification of animal species is achieved using accurate image segmentation; in the second aspect, in the process of obtaining weight pre-evaluation by using images, parameters related to weight need to be obtained from the images, and accurate acquisition of certain parameters is established on the premise of accurate segmentation of the images, for example, the acquisition of the body length of a pig depends on the accurate segmentation of the outline of the pig; the measurement of the length of a person's leg must rely on segmenting the image of the leg; the current image segmentation methods are mainly divided into several categories: a threshold-based segmentation method, a region-based segmentation method, an edge-based segmentation method, a segmentation method based on a specific theory, and the like; the method utilizes the Mask-RCNN neural network, and divides and identifies the target example while realizing target detection by adding a branch network on the basis of the fast-RCNN.
In the embodiment, the Mask-RCNN is an example segmentation framework, and the method adds a Mask branch on the basis of the fast-RCNN and performs pixel-level segmentation and classification on the target while detecting the target; and the ROI align is used for replacing the ROI posing in the fast-RCNN, so that the problem of region mismatching is solved, and the method has higher precision and speed.
The overall framework of Mask-RCNN is shown in FIG. 2 and is composed of 4 parts, namely, Backbone, RPN (Region pro-visual network, RPN), ROI align and Classifier.
The Backbone is used for extracting a Feature map (Feature map) of an input picture, and the Feature map is used as the input of a subsequent RPN and a full connection layer; the RPN is used for generating a candidate Region (Region probability), and performing class possibility judgment and frame regression (Bounding box regression) operation on each feature Region; extracting the Proposal feature maps by the ROI align by collecting the input feature map and the candidate region, and taking the Proposal feature maps as the input of a subsequent full-connection layer to judge the target category; the Classifier calculates the category of the candidate region by using the Proposal feature maps, and simultaneously performs frame regression again to accurately position the detection frame and generate a mask for the target.
In the third step, although the parameter selection of the invention is based on the two-dimensional image, the three-dimensional structure information of the animal, such as body length, chest circumference, body height and the like, can be extracted from the two-dimensional image; the method comprises the steps of acquiring images, wherein the distances between a photographing device and an animal are different, the positions of photographing objects are different, and the area of a reference system in the images is changed, so that a known fixed-size reference object is required to be selected to calculate corresponding parameters.
The establishment of the weight estimation model comprises the following steps:
building a database: by renting a server, building a database on the server, and putting historical data into the database, historical estimation records of a user can be put into the database in real time, and the server can also access the database in real time to continuously perfect a weight estimation model;
uploading the picture to a server to obtain the estimated weight: the user uploads the picture of the measured animal through the handheld terminal consisting of the central processing unit 1, the memory 2, the wireless transmission module 3, the ear tag module 4, the distance sensor 5, the display screen 6, the camera 7 and the key 8, and the server outputs the estimated weight of the measured animal in the picture.
In order to further explain the invention better, the following specific application examples are provided for the Mask-RCNN neural network:
the Mask-RCNN neural network and the related algorithm are applied to aspects of example segmentation, target detection, key point detection and the like, so that example individuals in the photos are identified and the outlines are separated.
(1) Example segmentation:
the target individual is an individual with single color and clear outline in the picture, and because the test condition is that the individual is independently photographed vertically, the conditions of overlapping, curling and the like of the individual to be detected do not need to be considered, and Mask-RCNN is not needed to be used for realizing gesture recognition; therefore, the preprocessed picture to be detected can be directly input, and the animal outline can be directly extracted from the picture by utilizing the self-carried recognition function of the Mask-RCNN; the same is true for the identification of other species; for example, the identification of livestock such as pigs, cattle and the like and poultry such as chickens and ducks and the like can obtain results such as pigs, cattle, chickens and the like; if the Mask-RCNN library does not contain the outline identification of the species to be detected, the identified model library can be manually expanded; performing feature extraction on the training image by using a convolutional neural network, and establishing a model through deep learning so as to achieve the purpose of expansion; different deep learning networks have different characteristic extraction effects on different objects, a proper deep learning network is selected as a backbone network during migration, when the same data is used for training, the accuracy of the network using ResNet101 is higher than that of the network using ResNet50 and the like, and the algorithm of Mask-RCNN is high-speed, simple, convenient, high in accuracy, simple, visual and easy to use.
(2) Error rate calculation:
in order to quantitatively evaluate the segmentation effect, a target individual manually scratched in Photoshop software is selected as an evaluation reference, the error rate is used for evaluating the accuracy of a target image by aiming at the algorithm, and the evaluation result is used as an improvement standard, so that the operation methods and related algorithms of photographing, image preprocessing and image segmentation are perfected, and the experimental error is reduced.
(3) Application of Mask-RCNN algorithm:
Mask-RCNN FCN (Mask bridge) is added on the basis of fast-RCNN to generate a Mask of an object, and RolAlign is used for processing the problem that the Mask is not aligned with the object in the original image.
The pixel values of the fixed four point coordinates are obtained by the RolAlign through bilinear interpolation, so that discontinuous operations become continuous, the situation that huge errors are generated due to twice rounding in the original RolPooling is avoided, and the errors are reduced when the RolAlign returns to the original image, as shown in fig. 3.
The method comprises the following three steps:
the method comprises the following steps: bin division 7 × 7, as shown in fig. 4;
step two: performing bilinear interpolation on each bin to obtain four points, as shown in fig. 5;
step three: the final 7 x 7 ROI was obtained by max poling after the interpolation was completed, as shown in fig. 6.
To further explain the invention, the following specific application examples are also provided for the RBF (radial basis) neural network:
after the data are obtained, a mathematical model is established by regression fitting through methods such as a neural network and the like, a calculation formula of the animal weight is obtained, a system stores relevant models and deep learning, parameters are stored in a database, the model is built based on an RBF neural network, the relation between body size data and the weight is learned, and therefore the body mass of various animals is predicted; the specific method comprises the following steps:
(1) RBF network architecture analysis:
an RBF neural network, under the condition that the hidden nodes are sufficient, through sufficient learning, any nonlinear function can be approached by any precision, and the RBF neural network has the optimal approximation capability of a general function, in addition, the RBF neural network has higher convergence speed and strong anti-noise and repair capabilities, and is distinguished from other neural networks by the advantages of small calculated amount, high learning speed, difficult falling into local parts, and the like in the aspect of solving the problem of collinearity of animal body scale data; its network structure is not complicated, only has two layers: hidden layer and output layer, the network model of which is shown in fig. 7:
the structural hidden layer is a neuron using RBF as an activation function, wherein the common activation function of the RBF hidden layer is a Gaussian function:
the output layer adopts neurons of a linear function to perform linear classification, and data is converted into a high-dimensional space so as to be linearly separable in the high-dimensional space; between neurons isIn the random weight calculation method, the weight of the first layer is set as the transposition of the input matrix P, and the weight and the bias of the second layer are reversely deduced through the input parameter T and the output result of the first layer.
(2) RBF network algorithm application:
the RBF neural network in the general case has an unknown quantity: the central vector ui, the constant sigma in the gaussian function, the output layer weight W, and the learning of the algorithm adopt a lazy RBF method, the flow of which is shown in FIG. 8:
specifically, it can be described as follows:
1) searching a central vector ui by using a kmeans algorithm;
2) calculating sigma by KNN (K nearest neighbor) rule
3) W can be obtained by the least square method.
The RBF neural network applied by the invention is a three-layer static feedforward neural network, the number of hidden layer units, namely the structure of the network, can be adaptively adjusted in a training stage according to the specific problem of research, and the applicability of the network is good; secondly, the RBF neural network has stronger input and output mapping functions, has the characteristic of unique optimal approximation, has no local minimum problem, and has good classification capability and high convergence speed in the learning process by theoretically proving that the RBF neural network in the forward network is the optimal network for completing the mapping function.
In summary, the following steps: according to the weight evaluation system and the implementation method based on the computer vision technology, parameters are obtained by processing the animal body through images, then a mathematical model is obtained by fitting with known weight data and stored in a database, the system identifies the type of the animal body and calls a related model during actual application, and rough weight data are obtained through calculation, so that the problem that the animal body weight is obtained by a traditional weight scale can be effectively solved, and the problem that the operation is complicated due to the fact that the animal individual is not completely controllable is solved, and the weight evaluation system has a high application value in the current market; has the following practicability: non-contact animal weight estimation can be realized; has the advantages of high efficiency: the traditional method is avoided, the labor amount is reduced, and the operation is simplified; the method has the following reliability: the dependence on the traditional weighing scale is reduced, and the prediction result is reliable; has the innovation: and (3) providing an image weight estimation model based on computer vision and deep learning.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be able to cover the technical solutions and the inventive concepts of the present invention within the technical scope of the present invention.
Claims (6)
1. A weight evaluation system based on computer vision technology and an implementation method thereof are characterized by comprising a central processing unit (1), a memory (2), a wireless transmission module (3), an ear tag module (4) and a distance sensor (5); the central processing unit (1), the memory (2), the wireless transmission module (3), the ear tag module (4) and the distance sensor (5) are integrated in a handheld shell device, and a display screen (6), a camera (7) and a key (8) for controlling operation are arranged on the shell device;
the display screen (6) is used for displaying the shot pictures, and the calculation result and the human-computer interaction related to the user setting; the camera (7) is used for shooting an animal picture; the distance sensor (5) is used for detecting the distance between the camera (7) and the animal when the animal picture is shot so as to conveniently correct the actual size of the shot animal; the wireless transmission module (3) is used for uploading the measurement result to a cloud server, or is used for downloading equipment parameters and updating a firmware program; the memory (2) is used for storing the measurement results to a local device; the ear tag module (4) is used for reading the unique identification code of the animal and correlating the unique identification code with the measurement result; the central processing unit (1) is used as a core processing unit of the whole system and is respectively connected with the memory (2), the wireless transmission module (3), the ear tag module (4), the distance sensor (5), the display screen (6), the camera (7) and the key (8), and the weight of the animal image acquired by the camera (7) is estimated by using an algorithm.
2. A method of implementing a computer vision based animal weight assessment system as claimed in claim 1, comprising the steps of:
s1: taking pictures and measuring distance: a user triggers the camera (7) to shoot an animal image through the key (8) and displays the animal image on the display screen (6); meanwhile, the distance sensor (5) measures the distance of the tested animal and displays the distance on the display screen (6);
s2: image segmentation: inputting the image obtained in the step S1 into a Mask-RCNN neural network which is trained to obtain an animal image after binary segmentation and an animal category;
s3: extracting animal parameters: extracting three-dimensional structure information of the animal including parameters of body length, chest circumference and body height according to the binarized image, and calculating actual values of the parameters of body length, chest circumference and body height according to the distance measured by S1;
s4: inputting the actual animal parameter values obtained in the step S3 into a pre-trained-RBF neural network according to the animal type to obtain a weight estimation value of the shot animal;
s5: scanning an ear tag of the animal through an ear tag module (4), acquiring a unique identification code of the animal, and associating the unique identification code with the weight value obtained in S4;
s6: the result obtained by S5 is stored in the system through the memory (2) so as to be convenient for subsequent viewing; meanwhile, the measurement result can be transmitted to the cloud server through the wireless transmission module (3), so that large-scale data automatic management is facilitated.
3. The system for assessing the weight based on computer vision technology and the method for implementing the same as claimed in claim 2, wherein the step of preprocessing the image of the animal captured by the camera (7) in S1 is as follows:
the method comprises the steps of firstly realizing graying of an image, then improving the definition and contrast of the image by utilizing image histogram equalization, then realizing image denoising by utilizing bilateral filtering, and then solving the problems of halo artifacts and reduction of hue and color saturation by utilizing an image adaptive enhancement algorithm based on Retinex theory, thereby increasing the image usability.
4. The system of claim 2, wherein in step S2, a Mask-RCNN neural network is used, and a branch network is added on the basis of fast-RCNN to detect the target and segment and identify the target instance.
5. The system of claim 2, wherein the relationship between the body size parameters of the species and the body weight is obtained according to the body size parameters extracted from the images by establishing a regression model for estimating the body weight in S3 and S4 and using an RBF neural network.
6. The system for assessing weight based on computer vision technology and the method for implementing the same as claimed in claim 2, wherein the specific method in S6 includes:
building a database: by renting a server, building a database on the server, and putting historical data into the database, historical estimation records of a user can be put into the database in real time, and the server can also access the database in real time to continuously perfect a weight estimation model;
uploading the picture to a server to obtain the estimated weight: the user uploads the picture of the tested animal through the handheld terminal consisting of the central processing unit (1), the memory (2), the wireless transmission module (3), the ear tag module (4), the distance sensor (5), the display screen (6), the camera (7) and the keys (8), and the server outputs the weight evaluation of the tested animal in the picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110047484.5A CN112862757A (en) | 2021-01-14 | 2021-01-14 | Weight evaluation system based on computer vision technology and implementation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110047484.5A CN112862757A (en) | 2021-01-14 | 2021-01-14 | Weight evaluation system based on computer vision technology and implementation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112862757A true CN112862757A (en) | 2021-05-28 |
Family
ID=76005732
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110047484.5A Pending CN112862757A (en) | 2021-01-14 | 2021-01-14 | Weight evaluation system based on computer vision technology and implementation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112862757A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113487107A (en) * | 2021-07-28 | 2021-10-08 | 华南农业大学 | Large animal weight automatic evaluation method, system and medium based on multilayer radial basis network |
| CN113901600A (en) * | 2021-09-13 | 2022-01-07 | 杭州大杰智能传动科技有限公司 | Automatic monitoring control method and system for lifting load balance of intelligent tower crane |
| CN114092973A (en) * | 2022-01-19 | 2022-02-25 | 北京探感科技股份有限公司 | Animal feature detection equipment and method based on visual identification |
| CN114431687A (en) * | 2022-02-18 | 2022-05-06 | 江苏中科能凯夫空调有限公司 | Electric heating regulation and control platform based on signal processing |
| CN115273155A (en) * | 2022-09-28 | 2022-11-01 | 成都大熊猫繁育研究基地 | Method and system for identifying pandas through portable equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017034776A1 (en) * | 2015-08-24 | 2017-03-02 | Illinois Tool Works Inc. | Multifunction livestock measurement station |
| US20190075756A1 (en) * | 2017-09-11 | 2019-03-14 | FarmIn Technologies | Systems, methods, and apparatuses for animal weight monitoring and management |
| CN109655019A (en) * | 2018-10-29 | 2019-04-19 | 北方工业大学 | Cargo volume measurement method based on deep learning and three-dimensional reconstruction |
| KR20190068266A (en) * | 2017-12-08 | 2019-06-18 | 김정구 | System for measuring weight of livestocks using image analysis and method using the same |
| US20200202511A1 (en) * | 2018-12-21 | 2020-06-25 | Neuromation, Inc. | System and method to analyse an animal's image for market value determination |
| KR102131560B1 (en) * | 2019-05-27 | 2020-07-07 | 주식회사 일루베이션 | Wearable type livestock weighing apparatus and a livestock weighing method using the same |
-
2021
- 2021-01-14 CN CN202110047484.5A patent/CN112862757A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017034776A1 (en) * | 2015-08-24 | 2017-03-02 | Illinois Tool Works Inc. | Multifunction livestock measurement station |
| US20190075756A1 (en) * | 2017-09-11 | 2019-03-14 | FarmIn Technologies | Systems, methods, and apparatuses for animal weight monitoring and management |
| KR20190068266A (en) * | 2017-12-08 | 2019-06-18 | 김정구 | System for measuring weight of livestocks using image analysis and method using the same |
| CN109655019A (en) * | 2018-10-29 | 2019-04-19 | 北方工业大学 | Cargo volume measurement method based on deep learning and three-dimensional reconstruction |
| US20200202511A1 (en) * | 2018-12-21 | 2020-06-25 | Neuromation, Inc. | System and method to analyse an animal's image for market value determination |
| KR102131560B1 (en) * | 2019-05-27 | 2020-07-07 | 주식회사 일루베이션 | Wearable type livestock weighing apparatus and a livestock weighing method using the same |
Non-Patent Citations (2)
| Title |
|---|
| MIKEL GJERGJI等: "Deep Learning Techniques for Beef Cattle Body Weight Prediction", 《2020 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)》 * |
| 张凯: "基于计算机视觉的育肥猪体重估测分析研究", 《中国优秀博硕士学位论文全文数据库（硕士）农业科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113487107A (en) * | 2021-07-28 | 2021-10-08 | 华南农业大学 | Large animal weight automatic evaluation method, system and medium based on multilayer radial basis network |
| CN113487107B (en) * | 2021-07-28 | 2024-04-12 | 华南农业大学 | Automatic large animal weight assessment method, system and medium based on multilayer radial basis network |
| CN113901600A (en) * | 2021-09-13 | 2022-01-07 | 杭州大杰智能传动科技有限公司 | Automatic monitoring control method and system for lifting load balance of intelligent tower crane |
| CN113901600B (en) * | 2021-09-13 | 2023-06-02 | 杭州大杰智能传动科技有限公司 | Automatic monitoring and controlling method and system for lifting load balance of intelligent tower crane |
| CN114092973A (en) * | 2022-01-19 | 2022-02-25 | 北京探感科技股份有限公司 | Animal feature detection equipment and method based on visual identification |
| CN114431687A (en) * | 2022-02-18 | 2022-05-06 | 江苏中科能凯夫空调有限公司 | Electric heating regulation and control platform based on signal processing |
| CN115273155A (en) * | 2022-09-28 | 2022-11-01 | 成都大熊猫繁育研究基地 | Method and system for identifying pandas through portable equipment |
| CN115273155B (en) * | 2022-09-28 | 2022-12-09 | 成都大熊猫繁育研究基地 | Method and system for identifying pandas through portable equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112862757A (en) | Weight evaluation system based on computer vision technology and implementation method | |
| CN108960245B (en) | Tire mold character detection and recognition method, device, equipment and storage medium | |
| CN111780764B (en) | Visual positioning method and device based on visual map | |
| CN106886216B (en) | Robot automatic tracking method and system based on RGBD face detection | |
| US7706601B2 (en) | Object posture estimation/correlation system using weight information | |
| CN109003390A (en) | A kind of commodity recognition method, self-service machine and computer readable storage medium | |
| WO2019200735A1 (en) | Livestock feature vector acquisition method, apparatus, computer device and storage medium | |
| CN111368766B (en) | Deep learning-based cow face detection and recognition method | |
| CN110569837A (en) | Method and device for optimizing damage detection result | |
| CN111415339B (en) | Image defect detection method for complex texture industrial product | |
| CN111126393A (en) | Vehicle appearance refitting judgment method and device, computer equipment and storage medium | |
| JP2010157093A (en) | Motion estimation device and program | |
| CN109145752A (en) | For assessing the method, apparatus, equipment and medium of object detection and track algorithm | |
| CN112200056A (en) | Face living body detection method and device, electronic equipment and storage medium | |
| CN113128518B (en) | Sift mismatch detection method based on twin convolution network and feature mixing | |
| CN111723688A (en) | Human body action recognition result evaluation method and device and electronic equipment | |
| JP6893812B2 (en) | Object detector | |
| CN116434266A (en) | Automatic extraction and analysis method for data information of medical examination list | |
| CN114743224B (en) | Animal husbandry livestock body temperature monitoring method and system based on computer vision | |
| CN115527050A (en) | Image feature matching method, computer device and readable storage medium | |
| CN114821035A (en) | Distance parameter identification method for infrared temperature measurement equipment of power equipment | |
| KR20210031444A (en) | Method and Apparatus for Creating Labeling Model with Data Programming | |
| WO2022171267A1 (en) | System, method, and computer executable code for organism quantification | |
| CN109034125A (en) | Pedestrian detection method and system based on scene complexity | |
| CN116758589B (en) | Cattle face recognition method for processing gesture and visual angle correction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210528 |