CN111652283A - Vegetable and fruit identification method, settlement equipment and use method - Google Patents

Vegetable and fruit identification method, settlement equipment and use method Download PDF

Info

Publication number
CN111652283A
CN111652283A CN202010373783.3A CN202010373783A CN111652283A CN 111652283 A CN111652283 A CN 111652283A CN 202010373783 A CN202010373783 A CN 202010373783A CN 111652283 A CN111652283 A CN 111652283A
Authority
CN
China
Prior art keywords
layer
vegetable
fruit
neural network
vegetables
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
Application number
CN202010373783.3A
Other languages
Chinese (zh)
Inventor
董元发
严华兵
舒正涛
朱成龙
郭盼
王田
舒现维
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Three Gorges University CTGU
Original Assignee
China Three Gorges University CTGU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Three Gorges University CTGU filed Critical China Three Gorges University CTGU
Priority to CN202010373783.3A priority Critical patent/CN111652283A/en
Publication of CN111652283A publication Critical patent/CN111652283A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/68—Food, e.g. fruit or vegetables

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a vegetable and fruit identification method, settlement equipment and a using method, wherein the vegetable and fruit identification method comprises the following steps: the method comprises the following steps: establishing a vegetable and fruit image data set in advance according to a large number of specified vegetable and fruit pictures; step two: preprocessing the acquired image; step three: after processing, a deep learning method is adopted to build a vegetable and fruit classification convolution neural network model; step four: the convolutional neural network training model can be used for automatically identifying the types of vegetables and fruits through training. The equipment effectively improves the weighing and settlement efficiency of vegetables and fruits, and can be suitable for weighing and settlement of most vegetables and fruits.

Description

Vegetable and fruit identification method, settlement equipment and use method
Technical Field
The invention belongs to the technical field of machine vision, and particularly relates to a vegetable and fruit identification method and automatic settlement equipment.
Background
With the development of society and economy, the development of computer technology is very rapid, wherein, the artificial intelligence technology is also rapidly developed, and various intelligent products based on the artificial intelligence technology gradually appear in the visual field of people. In addition, the deep learning technology plays an excellent role in the field of image recognition as a mature technology of artificial intelligence. The technology breaks through the constraint of the traditional manufacturing, fully utilizes image resources, improves the automation degree and saves manpower and material resources.
At present, after a customer selects to bag vegetables and fruits for selling in various large comprehensive supermarkets, medium and small supermarkets, retail stores and vegetable and fruit exclusive-selling stores, the customers are handed to a weigher to weigh, prick and stick price labels. The process is repeated mechanically, and during the shopping peak period, a weigher should take no time, so that the labor intensity is high. In addition, the low weighing efficiency also causes the queuing phenomenon of customers, so that the customer experience is poor and the supermarket benefit is reduced. There is also an independent settlement device in the market, and customers can settle accounts by scanning commodity bar codes by themselves, but the premise is that commodities are provided with price bar codes, and the price bar codes do not help scattered vegetables and fruits except for part of vegetables and fruits which are packed in advance. In summary, at present, no equipment capable of automatically weighing and settling accounts for bulk commodities such as vegetables and fruits exists, so that the selling efficiency is improved, the labor intensity of workers is reduced, and the labor cost investment is reduced.
Disclosure of Invention
The invention overcomes the defects of the prior art, and provides a vegetable and fruit identification method and settlement equipment, which effectively improve the vegetable and fruit weighing and settlement efficiency and can be suitable for weighing and settlement of most vegetables and fruits.
In order to achieve the technical features, the invention is realized as follows: a vegetable and fruit identification method comprises the following steps:
the method comprises the following steps: establishing a vegetable and fruit image data set in advance according to a large number of specified vegetable and fruit pictures;
step two: preprocessing the acquired image;
step three: after processing, a deep learning method is adopted to build a vegetable and fruit classification convolution neural network model;
step four: the convolutional neural network model can be used for automatically identifying the types of vegetables and fruits through training.
The convolutional neural network model is subjected to model training by putting pictures in batches, and the convolutional neural network model extracts features through a bottom layer and further extracts features at a deeper level to finally obtain the classification of the target;
the convolutional neural network model is based on LeNet basis and adopts a convolutional neural network model consisting of 2 convolutional layers, 2 pooling layers, 2 full-connection layers based on modified linear activation and an output layer.
The convolutional layer in the convolutional neural network model is a feature extraction layer, the input of the convolutional layer is from an input layer or a pooling layer, and in the convolutional layer, each neuron is only partially connected with the neuron in the input of the previous layer;
each feature map in the convolutional layer has a convolution kernel corresponding to the feature map, and the feature maps are the same as the convolution kernel in size, each feature map of the convolutional layer is obtained by performing convolution on the feature map input by the previous layer through different convolution kernels, then adding a bias after corresponding elements are accumulated, and finally activating a function;
assuming now that the l-th layer is a convolutional layer, the formula for the jth feature map in the convolutional layer is:
Figure BDA0002479364040000021
in the formula:
Figure BDA0002479364040000022
j output, M, representing the current l layerjMeans that first, from the characteristic diagram of the l-1 layer, selection is madeSeveral of which constitute the set of the ith input feature map,
Figure BDA0002479364040000023
the ith feature map representing the l-1 st layer,
Figure BDA0002479364040000024
a convolution kernel representing the l-th layer,
Figure BDA0002479364040000025
denotes the jth bias of the ith layer and f denotes the activation function.
After the pooling layer follows the convolutional layer, the pooling layer performs down-sampling processing on the feature map of the previous layer, and represents the extraction process of statistical information; the introduction of the pooling layer is only to perform dimension reduction processing on the input feature maps, does not change the number of the feature maps and has invariance;
let it be assumed that down denotes pooling operations,
Figure BDA0002479364040000026
indicating that any one of the characteristic diagrams has a multiplicative basis of the characteristic diagram,
Figure BDA0002479364040000027
and representing an additive base, calculating an expression formula of a certain feature graph in the pooling layer:
Figure BDA0002479364040000028
in order to make the value in which the response is large become larger while suppressing other neurons with small feedback, a local response normalization layer that creates a competitive mechanism for local neuron activity is proposed, which also strengthens the generalization ability of the model, the formula is calculated:
Figure BDA0002479364040000029
in the formula: k, N, alpha and beta are hyper-parameters which need to be selected in advance, a, b and N respectively represent the output of the convolution layer where the LRN is located, the output after regularization and the number of convolution kernels of one plane of the convolution layer, and a is also used as the input of the LRN; the settlement device comprises a machine body; the device is internally provided with a processor, a memory and computer hardware equipment.
The output layer of the convolutional neural network is a classifier layer, a Softmax regression classifier, a Sigmoid output unit and a radial basis function output unit are commonly used, and a calculation formula of a Softmax function is as follows:
Figure BDA0002479364040000031
in the formula: z represents the combination of all output nodes, zjRepresenting j output nodes, σ (z), in all combinationsjIndicating the occupation probability value of the j output node in all combinations.
The activation function introduces a nonlinear part into the whole network, the expression capacity of the network is enhanced, a relu function is adopted as the activation function, the relu activation function has the advantages of fast increase of the descending speed, linear non-saturation, simple calculation and the like, and the properties of the relu function and a calculation formula are as follows:
Figure BDA0002479364040000032
the preparation of the data set needs a large number of sample pictures, so that the characteristics learned by a neural network are more, the generated model is better, if the data set is insufficient, the over-fitting or under-fitting condition of the model can be caused to a great extent, and according to a data enhancement mode, the data sample size of the data set of the algorithm is expanded by adopting a horizontal rotation mode, a horizontal mirror image mode, a vertical mirror image mode, a salt and pepper noise adding mode and a Gaussian noise adding mode.
The vegetable and fruit recognition and settlement equipment designed by the vegetable and fruit recognition method comprises a machine body, wherein a microcomputer processor is installed in the machine body, a vegetable and fruit neural network training model is stored in the microcomputer processor, a display screen for displaying the type, quality, unit price and payment sum of vegetables and fruits is arranged on the machine body, and the display screen is arranged in front of the machine body so as to facilitate the browsing of customers; the back end face of the machine body is fixedly provided with a camera and an annular light source which are used for collecting commodity images through an upright post, and the camera is connected with the image input end of the microcomputer processor through a video line; and the signal output end of the microcomputer processor is connected with the display screen and outputs the result identified by the vegetable and fruit classification convolutional neural network model.
The top of the machine body is provided with an electronic scale which comprises a weighing tray and a computing system arranged on a microcomputer processor, and a leveling screw for leveling the equipment is arranged below the machine body;
price bar code printer is still installed to the lateral part of organism, prints the bar code after calculating the price to export the label from, the printer can change and beat printing paper, beat printing paper and install in the printer storehouse.
The use method of the vegetable and fruit identifying and settling equipment comprises the following steps:
step 1: the method comprises the following steps that supermarket staff record the type and price information of vegetables and fruits sold in a supermarket in advance to obtain equipment, and the supermarket staff is started;
step 2: after a customer selects required bulk vegetables and fruits to put into a bag, placing the vegetables and fruits on a weighing tray;
step 3: after the vegetables and fruits are placed, the annular light source is turned on to supplement light, the camera automatically captures vegetables and fruits pictures, and meanwhile, the vegetable and fruit quality weighing is completed;
step 4: the image information and the quality information of the vegetables and fruits are transmitted to a microcomputer processor, the commodity type is identified through a vegetable and fruit classification convolutional neural network model, the identified fruit and vegetable type information is transmitted to a computing system, and then the payment amount is automatically calculated according to unit price and quality;
step 5: the information of the type, unit price, quality and total amount of the vegetables and fruits is displayed on the electronic screen, meanwhile, the price bar code printer prints out the price label, and a customer can tear off the price label by himself and paste the price label on a vegetable and fruit packaging belt, namely, the automatic vegetable and fruit settlement process is completed.
The invention has the following beneficial effects:
1. by adopting the vegetable and fruit identification method and the settlement equipment, the equipment effectively improves the vegetable and fruit weighing and settlement efficiency, and can be suitable for weighing and settlement of most vegetables and fruits.
2. By adopting the vegetable and fruit classification convolution neural network model with the structure, a vegetable and fruit image data set is established in advance according to a large number of specified vegetable and fruit images, the images are preprocessed, a deep learning technology is adopted after the preprocessing, the vegetable and fruit classification convolution neural network model is established, the network model can autonomously identify the vegetable and fruit types through training, and the identification accuracy is more than 99%.
Drawings
The invention is further illustrated by the following figures and examples.
Fig. 1 is a three-dimensional view of the vegetable and fruit identifying and settling device of the present invention.
Fig. 2 is a flow chart of the method for using the vegetable and fruit identification and settlement device of the present invention.
FIG. 3 is a diagram of a network structure of a detection model according to the present invention.
FIG. 4 is a display interface of the terminal test result and system settlement according to the present invention.
In the figure: display screen 1, camera 2, annular light source 3, stand 4, tray 5 of weighing, organism 6, printer storehouse 7, leveling screw 8, price bar code printer 9.
Detailed Description
Embodiments of the present invention will be further described with reference to the accompanying drawings.
Example 1:
referring to fig. 1-4, a method for identifying vegetables and fruits comprises the following steps:
the method comprises the following steps: establishing a vegetable and fruit image data set in advance according to a large number of specified vegetable and fruit pictures;
step two: preprocessing the acquired image;
step three: after processing, a deep learning method is adopted to build a vegetable and fruit classification convolution neural network model;
step four: the convolutional neural network training model can be used for automatically identifying the types of vegetables and fruits through training.
By adopting the vegetable and fruit identification method, the automatic identification of the fruits and the vegetables can be realized, and the automatic settlement of the fruits and the vegetables in the supermarket is further improved, so that the selling efficiency is accelerated, the labor intensity of staff is reduced, and the labor cost input is reduced.
Furthermore, the convolutional neural network model is subjected to model training by putting pictures in batches, and the convolutional neural network model extracts features through a bottom layer and further extracts features at a deeper level to finally obtain the classification of the target; the convolutional neural network model is based on LeNet basis and adopts a convolutional neural network model consisting of 2 convolutional layers, 2 pooling layers, 2 full-connection layers based on modified linear activation and an output layer.
Further, the convolutional neural network training model is characterized by comprising the following steps:
Figure BDA0002479364040000051
further, a convolutional layer in the convolutional neural network model is a feature extraction layer, the input of the convolutional layer is derived from an input layer or a pooling layer, and in the convolutional layer, each neuron is only partially connected with a neuron in the input of the previous layer; each feature map in the convolutional layer has a convolution kernel corresponding to the feature map, and the feature maps are the same as the convolution kernel in size, each feature map of the convolutional layer is obtained by performing convolution on the feature map input by the previous layer through different convolution kernels, then adding a bias after corresponding elements are accumulated, and finally activating a function;
assuming now that the l-th layer is a convolutional layer, the formula for the jth feature map in the convolutional layer is:
Figure BDA0002479364040000061
in the formula:
Figure BDA0002479364040000062
j output, M, representing the current l layerjThe method comprises selecting a plurality of sets forming the I-th input feature map from the feature maps of the I-1 layer,
Figure BDA0002479364040000063
the ith feature map representing the l-1 st layer,
Figure BDA0002479364040000064
a convolution kernel representing the l-th layer,
Figure BDA0002479364040000065
denotes the jth bias of the ith layer and f denotes the activation function.
Furthermore, after the pooling layer follows the convolutional layer, the pooling layer performs down-sampling processing on the feature map of the previous layer, which represents the extraction process of the statistical information; the introduction of the pooling layer is only to perform dimension reduction processing on the input feature maps, does not change the number of the feature maps and has invariance;
down denotes a pooling operation and is the operation of,
Figure BDA0002479364040000066
indicating that any one of the characteristic diagrams has a multiplicative basis of the characteristic diagram,
Figure BDA0002479364040000067
and representing an additive base, calculating an expression formula of a certain feature graph in the pooling layer:
Figure BDA0002479364040000068
in order to make the value in which the response is large become larger while suppressing other neurons with small feedback, a local response normalization layer that creates a competitive mechanism for local neuron activity is proposed, which also strengthens the generalization ability of the model, the formula is calculated:
Figure BDA0002479364040000069
in the formula: k, N, alpha and beta are hyper-parameters which need to be selected in advance, a, b and N respectively represent the output of the convolution layer where the LRN is located, the output after regularization and the number of convolution kernels of one plane of the convolution layer, and a is also used as the input of the LRN; the settlement device comprises a machine body; the device is internally provided with a processor, a memory and computer hardware equipment.
Further, an output layer of the convolutional neural network is a classifier layer, a Softmax regression classifier, a Sigmoid output unit and a radial basis function output unit are commonly used, and a calculation formula of a Softmax function is as follows:
Figure BDA0002479364040000071
in the formula: z represents the combination of all output nodes, zjRepresenting j output nodes, σ (z), in all combinationsjIndicating the occupation probability value of the j output node in all combinations.
Furthermore, the activation function introduces a nonlinear part into the whole network, so that the expression capacity of the network is enhanced, a relu function is adopted as the activation function, the relu activation function has the advantages of rapid increase of the dropping speed, linear non-saturation, simple calculation and the like, and the properties of the relu function and a calculation formula are as follows:
Figure BDA0002479364040000072
furthermore, a large number of sample pictures are needed for preparing the data set, so that the characteristics learned by the neural network are more, the generated model is better, if the data set is insufficient, the over-fitting or under-fitting condition of the model can be caused to a great extent, and according to a data enhancement mode, the data set of the algorithm adopts a mode of expanding the data sample size through horizontal rotation, horizontal mirror image, vertical mirror image, addition of salt and pepper noise and Gaussian noise.
Example 2:
the vegetable and fruit recognition and settlement equipment designed by the vegetable and fruit recognition method comprises a machine body 6, wherein a microcomputer processor is installed in the machine body 6, a vegetable and fruit neural network training model is stored in the microcomputer processor, a display screen 1 for displaying the type, quality, unit price and payment sum of vegetables and fruits is arranged on the machine body 6, and the display screen 1 is arranged in front of the machine body 6 so as to facilitate the browsing of customers; the back end face of the machine body 6 is fixedly provided with a camera 2 and an annular light source 3 which are used for collecting commodity images through an upright post 4, and the camera 2 is connected with the image input end of a microcomputer processor through a video line; and the signal output end of the microcomputer processor is connected with the display screen 1 and outputs the result identified by the vegetable and fruit classification convolutional neural network model.
Further, the top of the machine body 6 of the vegetable and fruit identification and settlement equipment is provided with an electronic scale which comprises a weighing tray 5 and a computing system arranged on a microcomputer processor, and a leveling screw 8 for leveling the equipment is arranged below the machine body 6;
further, a price bar code printer 9 is installed on the side of the machine body 6, after the price is calculated, a bar code is printed, and a label is output from an outlet, the printer can replace printing paper, and the printing paper is installed on the printer bin 7.
Example 3:
the use method of the vegetable and fruit identifying and settling equipment comprises the following steps:
step 1: the method comprises the following steps that supermarket staff record the type and price information of vegetables and fruits sold in a supermarket in advance to obtain equipment, and the supermarket staff is started;
step 2: after a customer selects the required bulk vegetables and fruits to put into a bag, the vegetables and fruits are placed on a weighing tray 5;
step 3: after the vegetables and fruits are placed, the annular light source 2 is turned on to supplement light, the camera 2 automatically captures vegetables and fruits pictures, and meanwhile, the vegetable and fruit quality weighing is completed;
step 4: the image information and the quality information of the vegetables and fruits are transmitted to a microcomputer processor, the commodity type is identified through a vegetable and fruit classification convolutional neural network model, the identified fruit and vegetable type information is transmitted to a computing system, and then the payment amount is automatically calculated according to unit price and quality;
step 5: the information of the type, unit price, quality and total amount of the vegetables and fruits is displayed on the electronic screen 1, meanwhile, the price bar code printer 9 prints out the price label, and a customer can tear off the price label by himself and paste the price label on a vegetable and fruit packaging belt, namely, the automatic vegetable and fruit settlement process is completed.

Claims (10)

1. A vegetable and fruit identification method is characterized by comprising the following steps:
the method comprises the following steps: establishing a vegetable and fruit image data set in advance according to a large number of specified vegetable and fruit pictures;
step two: preprocessing the acquired image;
step three: after processing, a deep learning method is adopted to build a vegetable and fruit classification convolution neural network model;
step four: the convolutional neural network model can be used for automatically identifying the types of vegetables and fruits through training.
2. The method for identifying fruit and vegetable according to claim 1, wherein: the convolutional neural network model is subjected to model training by putting pictures in batches, and the convolutional neural network model extracts features through a bottom layer and further extracts features at a deeper level to finally obtain the classification of the target;
the convolutional neural network model is based on LeNet basis and adopts a convolutional neural network model consisting of 2 convolutional layers, 2 pooling layers, 2 full-connection layers based on modified linear activation and an output layer.
3. The method of claim 2, wherein the fruit and vegetable recognition method comprises: the convolutional layer in the convolutional neural network model is a feature extraction layer, the input of the convolutional layer is from an input layer or a pooling layer, and in the convolutional layer, each neuron is only partially connected with the neuron in the input of the previous layer;
each feature map in the convolutional layer has a convolution kernel corresponding to the feature map, and the feature maps are the same as the convolution kernel in size, each feature map of the convolutional layer is obtained by performing convolution on the feature map input by the previous layer through different convolution kernels, then adding a bias after corresponding elements are accumulated, and finally activating a function;
assuming now that the l-th layer is a convolutional layer, the formula for the jth feature map in the convolutional layer is:
Figure FDA0002479364030000011
in the formula:
Figure FDA0002479364030000012
j output, M, representing the current l layerjThe method comprises selecting a plurality of sets forming the I-th input feature map from the feature maps of the I-1 layer,
Figure FDA0002479364030000013
the ith feature map representing the l-1 st layer,
Figure FDA0002479364030000014
a convolution kernel representing the l-th layer,
Figure FDA0002479364030000015
denotes the jth bias of the ith layer and f denotes the activation function.
4. The method of claim 2, wherein the fruit and vegetable recognition method comprises: after the pooling layer follows the convolutional layer, the pooling layer performs down-sampling processing on the feature map of the previous layer, and represents the extraction process of statistical information; the introduction of the pooling layer is only to perform dimension reduction processing on the input feature maps, does not change the number of the feature maps and has invariance;
let it be assumed that down denotes pooling operations,
Figure FDA0002479364030000016
indicating that any one of the characteristic diagrams has a multiplicative basis of the characteristic diagram,
Figure FDA0002479364030000017
and representing an additive base, calculating an expression formula of a certain feature graph in the pooling layer:
Figure FDA0002479364030000021
in order to make the value in which the response is large become larger while suppressing other neurons with small feedback, a local response normalization layer that creates a competitive mechanism for local neuron activity is proposed, which also strengthens the generalization ability of the model, the formula is calculated:
Figure FDA0002479364030000022
in the formula: k, N, alpha and beta are hyper-parameters which need to be selected in advance, a, b and N respectively represent the output of the convolution layer where the LRN is located, the output after regularization and the number of convolution kernels of one plane of the convolution layer, and a is also used as the input of the LRN; the settlement device comprises a machine body; the device is internally provided with a processor, a memory and computer hardware equipment.
5. The method of claim 2, wherein the fruit and vegetable recognition method comprises: the output layer of the convolutional neural network is a classifier layer, a Softmax regression classifier, a Sigmoid output unit and a radial basis function output unit are commonly used, and a calculation formula of a Softmax function is as follows:
Figure FDA0002479364030000023
in the formula: z represents the combination of all output nodes, zjRepresenting j output nodes, σ (z), in all combinationsjIndicating the occupation probability value of the j output node in all combinations.
6. The method of claim 3, wherein the fruit and vegetable recognition method comprises: the activation function introduces a nonlinear part into the whole network, the expression capacity of the network is enhanced, a relu function is adopted as the activation function, the relu activation function has the advantages of fast increase of the descending speed, linear non-saturation, simple calculation and the like, and the properties of the relu function and a calculation formula are as follows:
Figure FDA0002479364030000024
7. the method for identifying fruit and vegetable according to claim 1, wherein: the preparation of the data set needs a large number of sample pictures, so that the characteristics learned by a neural network are more, the generated model is better, if the data set is insufficient, the over-fitting or under-fitting condition of the model can be caused to a great extent, and according to a data enhancement mode, the data sample size of the data set of the algorithm is expanded by adopting a horizontal rotation mode, a horizontal mirror image mode, a vertical mirror image mode, a salt and pepper noise adding mode and a Gaussian noise adding mode.
8. The vegetable and fruit identification settlement device designed by the vegetable and fruit identification method of any one of claims 1 to 7, wherein: the fruit and vegetable classifying neural network model comprises a machine body (6), wherein a microcomputer processor is installed in the machine body (6), a fruit and vegetable classifying neural network model is stored in the microcomputer processor, a display screen (1) for displaying the type, quality, unit price and payment sum of the fruits and vegetables is arranged on the machine body (6), and the display screen (1) is arranged in front of the machine body (6) so as to facilitate the browsing of a customer; the back end face of the machine body (6) is fixedly provided with a camera (2) and an annular light source (3) which are used for collecting commodity images through an upright post (4), and the camera (2) is connected with the image input end of a microcomputer processor through a video line; and the signal output end of the microcomputer processor is connected with the display screen (1) and outputs the result identified by the vegetable and fruit classification convolutional neural network model.
9. The vegetable and fruit identification and settlement apparatus according to claim 8, wherein: the top of the machine body (6) is provided with an electronic scale, the electronic scale comprises a weighing tray (5) and a computing system arranged on a microcomputer processor, and a leveling screw (8) for leveling equipment is arranged below the machine body (6);
price bar code printer (9) are still installed to the lateral part of organism (6), print the bar code after calculating the price to export the label from the export, printer can change printing paper, and printing paper is installed in printer storehouse (7).
10. The method for using the equipment for identifying and settling the vegetables and fruits as claimed in any one of claims 8 to 9, which comprises the steps of:
step 1: the method comprises the following steps that supermarket staff record the type and price information of vegetables and fruits sold in a supermarket in advance to obtain equipment, and the supermarket staff is started;
step 2: after a customer selects the vegetables and fruits in bulk to be bagged, placing the vegetables and fruits on a weighing tray (5);
step 3: after the vegetables and fruits are placed, the annular light source (2) is turned on to supplement light, the camera (2) automatically captures vegetables and fruits pictures, and meanwhile, the vegetable and fruit quality weighing is completed;
step 4: the image information and the quality information of the vegetables and fruits are transmitted to a microcomputer processor, the commodity type is identified through a vegetable and fruit classification convolutional neural network model, the identified fruit and vegetable type information is transmitted to a computing system, and then the payment amount is automatically calculated according to unit price and quality;
step 5: the information of the type, unit price, quality and total amount of the vegetables and fruits is displayed on the electronic screen (1), meanwhile, the price bar code printer (9) prints out the price label, and a customer can tear down the price label by himself and paste the price label on a vegetable and fruit packaging belt, namely, the automatic vegetable and fruit settlement process is completed.
CN202010373783.3A 2020-05-06 2020-05-06 Vegetable and fruit identification method, settlement equipment and use method Pending CN111652283A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010373783.3A CN111652283A (en) 2020-05-06 2020-05-06 Vegetable and fruit identification method, settlement equipment and use method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010373783.3A CN111652283A (en) 2020-05-06 2020-05-06 Vegetable and fruit identification method, settlement equipment and use method

Publications (1)

Publication Number Publication Date
CN111652283A true CN111652283A (en) 2020-09-11

Family

ID=72346634

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010373783.3A Pending CN111652283A (en) 2020-05-06 2020-05-06 Vegetable and fruit identification method, settlement equipment and use method

Country Status (1)

Country Link
CN (1) CN111652283A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112863081A (en) * 2021-01-04 2021-05-28 西安建筑科技大学 Device and method for automatic weighing, classifying and settling vegetables and fruits
CN113984169A (en) * 2021-10-26 2022-01-28 南京智凝人工智能研究院有限公司 Intelligent electronic scale system based on visual analysis
CN114399619A (en) * 2022-01-14 2022-04-26 南京苏胜天信息科技有限公司 Machine vision image recognition system and processing method thereof
CN116402671A (en) * 2023-06-08 2023-07-07 北京万象创造科技有限公司 Sample coding image processing method for automatic coding system

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010018348A2 (en) * 2008-08-14 2010-02-18 Azdine Bahou Device for identifying and automatically paying for agri-food or pharmaceutical products through optical recognition
CN107767590A (en) * 2016-08-16 2018-03-06 深圳仪普生科技有限公司 Automatic identification commercialization bar code electronic scale and Automatic identification method
CN107784305A (en) * 2017-09-29 2018-03-09 中国农业科学院农业环境与可持续发展研究所 Facilities vegetable disease recognition method and device based on convolutional neural networks
CN108537994A (en) * 2018-03-12 2018-09-14 深兰科技（上海）有限公司 View-based access control model identifies and the intelligent commodity settlement system and method for weight induction technology
CN108921642A (en) * 2018-06-04 2018-11-30 北京小轮科技有限公司 One kind intelligent fruits and vegetables checkout apparatus based on computer vision and method
CN109190476A (en) * 2018-08-02 2019-01-11 福建工程学院 A kind of method and device of vegetables identification
CN110580450A (en) * 2019-08-12 2019-12-17 西安理工大学 traffic sign identification method based on convolutional neural network
CN110647941A (en) * 2019-09-25 2020-01-03 绍兴数鸿科技有限公司 Vegetable identification method and equipment based on convolutional neural network
CN112863081A (en) * 2021-01-04 2021-05-28 西安建筑科技大学 Device and method for automatic weighing, classifying and settling vegetables and fruits

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010018348A2 (en) * 2008-08-14 2010-02-18 Azdine Bahou Device for identifying and automatically paying for agri-food or pharmaceutical products through optical recognition
CN107767590A (en) * 2016-08-16 2018-03-06 深圳仪普生科技有限公司 Automatic identification commercialization bar code electronic scale and Automatic identification method
CN107784305A (en) * 2017-09-29 2018-03-09 中国农业科学院农业环境与可持续发展研究所 Facilities vegetable disease recognition method and device based on convolutional neural networks
CN108537994A (en) * 2018-03-12 2018-09-14 深兰科技（上海）有限公司 View-based access control model identifies and the intelligent commodity settlement system and method for weight induction technology
CN108921642A (en) * 2018-06-04 2018-11-30 北京小轮科技有限公司 One kind intelligent fruits and vegetables checkout apparatus based on computer vision and method
CN109190476A (en) * 2018-08-02 2019-01-11 福建工程学院 A kind of method and device of vegetables identification
CN110580450A (en) * 2019-08-12 2019-12-17 西安理工大学 traffic sign identification method based on convolutional neural network
CN110647941A (en) * 2019-09-25 2020-01-03 绍兴数鸿科技有限公司 Vegetable identification method and equipment based on convolutional neural network
CN112863081A (en) * 2021-01-04 2021-05-28 西安建筑科技大学 Device and method for automatic weighing, classifying and settling vegetables and fruits

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
曾平平，李林升 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112863081A (en) * 2021-01-04 2021-05-28 西安建筑科技大学 Device and method for automatic weighing, classifying and settling vegetables and fruits
CN113984169A (en) * 2021-10-26 2022-01-28 南京智凝人工智能研究院有限公司 Intelligent electronic scale system based on visual analysis
CN114399619A (en) * 2022-01-14 2022-04-26 南京苏胜天信息科技有限公司 Machine vision image recognition system and processing method thereof
CN116402671A (en) * 2023-06-08 2023-07-07 北京万象创造科技有限公司 Sample coding image processing method for automatic coding system
CN116402671B (en) * 2023-06-08 2023-08-15 北京万象创造科技有限公司 Sample coding image processing method for automatic coding system

Similar Documents

Publication Publication Date Title
CN111652283A (en) Vegetable and fruit identification method, settlement equipment and use method
US11151427B2 (en) Method and apparatus for checkout based on image identification technique of convolutional neural network
CN107767590A (en) Automatic identification commercialization bar code electronic scale and Automatic identification method
Rocha et al. Automatic fruit and vegetable classification from images
CN111814614A (en) Intelligent object-identifying electronic scale weighing method and system
Bai et al. Products-10k: A large-scale product recognition dataset
CN106340137A (en) Automatic weighing machine based on deep learning and program control method thereof
CN111222389B (en) Analysis method and system for commodity on super commodity shelf
Zeeshan et al. Fruit classification system using multiclass support vector machine classifier
CN108960260A (en) A kind of method of generating classification model, medical image image classification method and device
CN110909698A (en) Electronic scale recognition result output method, system, device and readable storage medium
CN111814660A (en) Image identification method, terminal equipment and storage medium
Ragesh et al. Deep learning based automated billing cart
WO2021179138A1 (en) Method and system for analyzing commodities on supermarket shelf
Burkapalli et al. TRANSFER LEARNING: INCEPTION-V3 BASED CUSTOM CLASSIFICATION APPROACH FOR FOOD IMAGES.
CN112232334B (en) Intelligent commodity selling identification and detection method
WO2021048813A1 (en) Scale and method for the automatic recognition of a product
Yuhui et al. An automatic recognition method of fruits and vegetables based on depthwise separable convolution neural network
Priya et al. Real time fruits quality detection with the help of artificial intelligence
CN114282818A (en) Bright area electronic tag control method for maximizing sales
Darapaneni et al. Banana Sub-Family Classification and Quality Prediction using Computer Vision
Satria et al. ConFruit: Effective Fruit Classification Using CNN Algorithm
Kishore et al. Fruits and vegetables classification using progressive resizing and transfer learning
Haritha et al. A Novel Method of Billing System Using Deep Learning
Alodat Deep Transfer Learning and Intelligent Item Packing in Retail Management

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