TWI703319B - Fruit sweetness detecting device and detecting method combined with artificial intelligence and spectral detection - Google Patents

Abstract

A fruit sweetness detecting device and a detecting method combined with artificial intelligence and spectral detection are disclosed. A detection platform is provided for disposing a fruit. An near infrared detecting device generates an infrared beam to illuminate the fruit and receive the reflected beam. A processing device compares the sweetness database corresponding to the fruit variety according to a set of wavelength data in the reflected light beam to determine the sweetness data of the fruit, wherein the processing device uses the artificial intelligence framework to train and establish a sweetness database corresponding to the variety of the fruit. The identification code printing machine is configured to print an identification code on the sticker according to a printing command of the processing device, wherein the identification code corresponds to the information, so that the user can scan the information of the identification code to at least obtain the sweetness of the fruit.

Description

Fruit sweetness detection device and detection method combining artificial intelligence and spectrum detection

This creation is a detector and detection method, and especially a fruit sweetness detection device and detection method that combines artificial intelligence and spectrum detection.

Generally, fruit farmers cannot afford to purchase high-cost sweetness detectors, so they generally only screen the appearance of the fruit and record the fruit weight. If a large number of fruits are harvested, it will take more time to screen, only when the fruit is sent to the farmers’ association or the fruit market After that, it is possible to measure the sweetness of the fruit, and then grade the fruit, but this will take too much time, and the measurement method will destroy the appearance of the fruit and affect the appearance. At present, Taiwan's agriculture mostly uses manpower or destructive technology to detect the sweetness of fruits. This not only consumes a lot of manpower, but the accuracy is also unstable.

According to the current production traceability system of agricultural products, the QR Code of each farm is used to query which fruit grower the producer is and what product it produces. So as long as the traceability number is marked in the production traceability system, you can know what the consumers are buying. Where did you come from, but the retro number is QR The Code method is pasted on the fruit carton, which is very likely to be forged, and it is possible that the box of fruit obtained by the consumer is not produced by the real farmer.

In view of this, the purpose of this creation is to provide a fruit sweetness detection device and its detection method combining artificial intelligence and spectral detection.

In order to achieve one of the foregoing objectives, this creation provides a fruit sweetness detection device combining artificial intelligence and spectral detection, which at least includes: a detection platform for placing a fruit; a near infrared detection device for generating at least one infrared beam Irradiating fruit and receiving at least one reflected light beam; a processing device compares a sweetness database corresponding to a variety of fruits according to a set of wavelength data in the reflected light beam to determine the sweetness data of a fruit, wherein The processing device uses an artificial intelligence framework to train and establish a sweetness database corresponding to the variety of the fruit; and an identification code printer for printing an identification code on a sticker according to a printing command of the processing device, wherein the identification code It corresponds to a piece of information, so that the user can at least obtain the sweetness data of the fruit by scanning the information of the identification code.

The processing device uploads the sweetness data of the fruit and/or the production history of the fruit to a cloud database in the cloud, and the information of the identification code corresponds to an access link of the cloud database.

Among them, the access link has a timeliness, so that the identification code has a valid period.

Among them, the fruit has an expiration date, and the cloud database simultaneously records the sweetness data and expiration date of the fruit.

Among them, the validity period of the identification code corresponds to the expiration date of the fruit.

Among them, the group wavelength of the reflected beam is composed of 450nm, 500nm, 550nm, 570nm, 600 and 650nm.

Among them, the artificial intelligence framework is a random forest classification framework, and the processing device uses the random forest classification framework to train according to the group wavelength of a plurality of sample fruits and the sweetness value of the sample fruits to obtain a sweetness database.

Among them, the sweetness value of the sample fruit is measured with a brix meter.

Wherein, the detection platform has a rotating storage tray for placing fruits so that the fruits rotate relative to the near-infrared detection device, so that the near-infrared detection device receives the reflected light beam from the ring surface of the fruit.

In order to achieve the other objective mentioned above, this creation also provides a fruit sweetness detection method combining artificial intelligence and spectral detection, which at least includes: using an artificial intelligence framework to train to establish a sweetness database corresponding to at least one variety; The near-infrared detection device generates at least one infrared light beam to irradiate at least one reflected light beam corresponding to a fruit of a variety and receives at least one reflected light beam; according to the data of a set of wavelengths in the reflected light beam, the sweetness database of a variety corresponding to the fruit is compared, thereby Determine the sweetness data of one of the fruits; and form an identification code on the fruit. The identification code corresponds to a piece of information, so that the user can obtain the sweetness data of the fruit by scanning the information of the identification code.

According to this creation, the fruit sweetness detection device and its detection method combined with artificial intelligence and spectral detection have the following advantages:

(1) Non-destructive measurement of the sweetness of fruits, so there is no need to worry about damaging the appearance of the fruits, thereby ensuring farmers' rights to produce high-quality fruits.

(2) The near-infrared detection device of this creation only uses data of six wavelengths, so it is different from the traditional expensive high-end spectrometers that sell for hundreds of thousands of Taiwan dollars, and this creation has a certain degree of accuracy and low price. Generally used by small farmers.

(3) The identification code of this creation is time-sensitive and can prevent the identification code from being forged.

10: Detection platform

11: storage tray

20: Infrared detection device

30: Processing device

40: Identification code printer

42: Sticker

44: identification code

100: Fruit

200: Cloud

202: Cloud database

Figure 1 shows a schematic diagram of the fruit sweetness detection device combining artificial intelligence and spectral detection in this creation.

Figure 2 is a graph showing the wavelength data curve of fruits measured by the near-infrared detection device of this creation.

Figure 3 is a schematic diagram showing the random forest classification structure used in this creation.

Figure 4 shows a schematic diagram of the inverted neural network architecture used in this creation.

In order to help understand the technical features, content and advantages of this creation and the effects it can achieve, the matching schemes of this creation are explained in detail in the form of embodiments as follows, and the schemes used therein have their main points It is only for illustrative and auxiliary manual purposes, and may not be the true proportions and precise configuration after the implementation of this creation. Therefore, the proportion and configuration relationship of the attached drawings should not be interpreted or limited to the scope of rights of the creation in actual implementation. In addition, in order to facilitate understanding, the same elements in the following embodiments are indicated by the same symbols. In addition, the dimensional ratios of the components shown in the drawings are only to facilitate the explanation of the components and their structures, and are not intended to limit.

In addition, the terms used in the entire specification and the scope of the patent application, unless otherwise specified, usually have the usual meaning of each term used in the field, in the content disclosed here, and in the special content. Some terms used to describe this creation will be discussed below or elsewhere in this specification to provide those skilled in the art with additional guidance on the description of this creation.

Secondly, if the terms "including", "including", "having", "containing", etc. are used in this article, they are all open terms, which means including but not limited to.

Since most of the methods currently available on the market to detect the sweetness of fruits use destructive analysis methods, this creation is an analysis technique that uses near-infrared spectroscopy and only six wavelengths of data for compound analysis, which is a type of non-destructive analysis. method. The principle is to use the characteristics of reflectivity between molecules and atoms in the substance due to the difference in energy level, which varies according to the wavelength change, to identify the characteristic substance in the test object, and then cooperate with the characteristic spectrum database of the known substance to deduce the composition of each point And structure. Near-infrared spectroscopy is the most studied non-invasive analysis technology. The wavelength range of the near-infrared light region is 780~2500nm. The molecular functional groups of most substances such as CH, OH, NH, etc., have fixed absorption vibration spectra in this range. Use this as the basis for measuring sweetness to achieve the purpose of measuring fruit sweetness.

This creation is a fruit sweetness detection device and its detection method combining artificial intelligence and spectral detection. Please refer to FIG. 1, the fruit sweetness detection device combining artificial intelligence and spectral detection of the present invention at least includes a detection platform 10, a near infrared detection device 20, a processing device 30 and an identification code printer 40. Wherein, the detection platform 10 has a storage tray 11 for placing fruits 100. The near-infrared detection device 20 is installed on the detection platform 20, for example, next to the tray 11 to generate at least one infrared light beam to irradiate the surface of the fruit 100 on the tray 11 and to receive at least one of the surfaces reflected from the fruit 100 Reflected beam. In this embodiment, the fruit variety is guava. However, this creation is not limited to this, and the variety applicable to this creation can also be watermelon, pineapple or cantaloupe. The tray 11 can be a stationary type or a rotating type. Wherein, if it is a static storage tray 11, the near-infrared detection device 20 can measure the reflected light beam data on a single surface of the fruit 100. If it is a rotating tray 11, the fruit 100 can be rotated relative to the near-infrared detection device 20, so the near-infrared detection device 20 can measure the data of the reflected light beam on the torus of the fruit 100.

In the fruit sweetness detection device of the present invention, the processing device 30 is provided on the detection platform 10, and is electrically connected to the infrared detection device 20 via a wire, for example. The processing device 30 receives the data signal of the reflected light beam measured by the near-infrared detection device 20 for comparison and analysis. In detail, the processing device 30 compares the sweetness database based on a set of specific wavelength data in the reflected beam to determine the sweetness data of the fruit, wherein the sweetness database corresponds to the variety of the fruit. One of the features of this creation is the use of non-destructive detection methods, and the aforementioned set of wavelengths consists of only six wavelengths, of which the six wavelengths are 450nm, 500nm, 550nm, 570nm, 600 and 650nm respectively. In addition, another feature of this creation is to use artificial intelligence framework to train and build a sweetness database based on fruit varieties. For example, this creation can first use an artificial intelligence framework to target a specific variety, such as pearl guava, to train and build its sweetness database. After the user places the pearl guava on the tray 11 and uses the near-infrared detection device 20 to measure the reflected light beam of the fruit, the processing device 30 can compare the sweetness database established in advance with the six wavelengths in the reflected light beam of the fruit The data (shown in Figure 2) to determine the sweetness data of this fruit. In the same way, this creation can also first use the artificial intelligence framework to train separately for multiple fruit varieties and establish their sweetness database, so it can detect and judge the sweetness data for multiple fruit varieties. Taking guava as an example, the fruit variety can be, for example, pearl guava, imperial guava, red heart guava, or tu guava. The artificial intelligence architecture used in this creation is, for example, a Back Propagation Neural Network (BPNN) or a random decision forests (random decision forests).

For example, as shown in FIG. 3, the processing device 30 is trained based on the group wavelengths of a plurality of sample fruits and the sweetness values of the sample fruits by using a random forest classification framework to obtain the respective It should be in the sweetness database of the sample fruits. Among them, taking guava as an example, the sample fruit can be, for example, pearl guava, imperial guava, red heart guava, or tu guava. Random forest is classified as a machine learning integrated learning algorithm. The process is to randomly generate a forest, and the forest is composed of multiple decision trees. The node of each tree is the characteristic data of the creative training. When there is a new When the input sample of enters, let each decision tree in the forest judge separately. This creation first uses the data of six different wavelengths of light obtained by the spectrum detection device, and then cuts the fruit (sample fruit) to measure the sweetness of the commercially available sugar content to establish a database. After the random forest method, it predicts which category the spectrum of the fruit may belong to. For the pearl guava tested in this creation, the corresponding sugar content is between 10-15, and the test results will be divided into one of six categories. Since those with general knowledge in this creation should be able to learn how to use the random forest method to train based on the data of six spectral wavelengths to build a database, this creation will not be repeated here.

In addition, as shown in FIG. 4, the processing device 30 of the present creation may, for example, use a Back Propagation Neural Network (BPNN) based on the wavelengths of the plurality of sample fruits and the sweetness of the sample fruits. The degree value is trained to obtain the sweetness database corresponding to the types of the sample fruits. Backpropagation neural network architecture is a supervised learning algorithm combined with Back Propagation algorithm to calculate the gradient of the neural network and modify the weights to achieve better model accuracy, and train with a supervised learning method The data in the spectral database is trained to analyze the fruit sweetness detection model in this creation, and artificial intelligence (AI) detection methods are used to reduce labor costs and error rates caused by human factors, and improve fruit quality. Based on the same reason as the foregoing, since those with ordinary knowledge in this creation should be able to learn how to train based on the data of six spectral wavelengths with the inverted neural network architecture to build a database based on the teaching of this creation. The creation will not be repeated here.

The present creation may also have, for example, an identification code printer 40, commonly known as a labeling machine, which is electrically connected to the processing device 30 with a wire, and prints the identification code 44 on the sticker 42 according to the printing instructions of the processing device 30, so that A fruit supplier (for example, a farmer) can stick the sticker 42 on the fruit 100. In other words, after the processing device 30 determines the sweetness data of the fruit, the processing device 30 can generate a printing command to print the identification code 44 on the sticker 42. The identification code 44 corresponds to a piece of information, so that a user (such as a consumer) can recognize the information corresponding to the identification code 44 by scanning. The identification code 44 is, for example, a two-dimensional code, and the user can scan and recognize the information corresponding to the identification code 44 with a QR code scanner (not shown) such as a mobile phone, so as to obtain at least the sweetness data of the fruit 100 . The type and model of the identification code printer 40 are not particularly limited. As long as the sticker 42 with the identification code 44 can be printed, it is suitable for this creation. In addition, the identification code printer 40 can also be a laser printer, for example, a sticker can be omitted and a two-dimensional code can be directly printed on the surface of the fruit.

In an implementation aspect of this creation, the processing device 30 further uploads the sweetness data of the fruit 100 and/or the production history of the fruit to the cloud database 202 located in the cloud 200, and the information of the identification code 44 is, for example, the corresponding Access link to this cloud database 202. Therefore, when the user scans with a QR code scanner such as a mobile phone and recognizes the access link corresponding to the identification code 44, the user can read the cloud database 202 with the mobile phone to obtain at least the sweetness data of the fruit 100 . In other words, the identification code 44 of a certain fruit or batch of fruit is only applicable to this fruit or batch of fruit. However, in order to prevent the identification code 44 from being stolen by others and copied in large quantities, another feature of this creation is that the aforementioned access link is time-sensitive, so that the identification code 44 has a valid period. For example, the validity period of the aforementioned access link may be, for example, two weeks, such as September 1, 2019 to September 14, 2019. When the user scans the identification code at a time outside this time interval, even if the access link is obtained, the cloud database cannot be read due to the timeliness of the access link. In this way, this creation can improve food safety and prevent others from counterfeiting (theft) Print) label sticker. In another embodiment, the cloud database 202 can also record the sweetness data and expiration date of the fruit 100 at the same time. Therefore, when the user (consumer) reads the cloud database 202, he can not only know the sweetness data of the fruit, but also know whether the fruit has expired. In another embodiment, the expiration date of the identification code 44 may also correspond to the expiration date of the fruit 100, for example. In other words, if the expiration date of the fruit 100 is one month, the validity period of the identification code 44 can also be set to one month, for example. Therefore, even if the identification code is pirated by others, it will be easily detected by the user because the identification code has a valid period.

According to this creation, the fruit sweetness detection device and its detection method combined with artificial intelligence and spectral detection have the following advantages:

(1) Non-destructive measurement of the sweetness of fruits, so there is no need to worry about damaging the appearance of the fruits, thereby ensuring farmers' rights to produce high-quality fruits.

(2) The near-infrared detection device of this creation only uses data of six wavelengths, so it is different from the traditional expensive high-end spectrometers that sell for hundreds of thousands of Taiwan dollars, and this creation has a certain degree of accuracy and low price. Generally used by small farmers.

(3) The identification code of this creation is time-sensitive and can prevent the identification code from being forged.

The above description is only illustrative, and not restrictive. Any equivalent modifications or alterations that do not depart from the spirit and scope of the present invention should be included in the scope of the attached patent application.

10: Detection platform

11: storage tray

20: Infrared detection device

30: Processing device

40: Identification code printer

42: Sticker

44: identification code

100: Fruit

200: Cloud

202: Cloud database

Claims (5)

A fruit sweetness detection device combining artificial intelligence and spectrum detection, at least comprising: a detection platform for placing a fruit; a near-infrared detection device for generating at least one infrared beam to illuminate the fruit and receiving at least one reflection of reflection Light beam; a processing device, according to the data of a set of wavelengths in the reflected light beam is compared to a sweetness database corresponding to a variety of the fruit, wherein the set of wavelengths of the reflected light beam are from 450nm, 500nm, 550nm, 570nm , 600nm and 650nm composition, to determine the sweetness data of the fruit, wherein the processing device uses an artificial intelligence framework to train according to the set of wavelengths of a plurality of sample fruits and the sweetness value of the sample fruits to establish the corresponding fruit The sweetness database of the variety, the processing device uploads the sweetness data of the fruit and/or the production history of the fruit to a cloud database in the cloud; and an identification code printer for An identification code is printed on a sticker according to a print command of the processing device, wherein the identification code corresponds to an access link of the cloud database, so that the user can identify the access of the identification code by scanning Link to obtain at least the sweetness data of the fruit, wherein the access link has a timeliness, so that the identification code has an expiration date, wherein the fruit has an expiration date, and the cloud database records at the same time The sweetness data and the expiration date of the fruit, wherein the expiration date of the identification code corresponds to the expiration date of the fruit. Such as the fruit sweetness detection device combining artificial intelligence and spectral detection as described in item 1 of the scope of patent application, wherein the artificial intelligence framework is a random forest classification framework. The fruit sweetness detection device combining artificial intelligence and spectral detection as described in the first item of the scope of patent application, wherein the sweetness values of the sample fruits are measured by a brix meter. The fruit sweetness detection device combining artificial intelligence and spectral detection as described in the first item of the patent application, wherein the detection platform has a rotating tray for placing the fruit so that the fruit is relative to the near-infrared detection device Rotating, so that the near-infrared detection device receives the reflected light beam on the ring surface of the fruit. A fruit sweetness detection method combining artificial intelligence and spectral detection at least includes: using an artificial intelligence framework to train to establish a sweetness database corresponding to at least one variety; and generating at least one infrared beam corresponding to at least one infrared light beam by a near-infrared detection device A fruit of the variety receives at least one reflected light beam; the data of a set of wavelengths in the reflected light beam is compared with a sweetness database of the variety corresponding to the fruit to determine the sweetness data of the fruit , Wherein the set of wavelengths of the reflected light beam is composed of 450nm, 500nm, 550nm, 570nm, 600nm and 650nm; and an identification code is formed on the fruit, and the identification code corresponds to an access link of a cloud database, In this way, the user can identify the access link of the identification code by scanning, and obtain the sweetness data of the fruit, wherein the access link has a timeliness, so that the identification code has a valid period, wherein the The fruit has an expiration date, and the expiration date of the identification code corresponds to the expiration date of the fruit.

Images