JP7041049B2 - Object identification device and object identification system - Google Patents

Description

The present invention relates to an object identification device and an object identification system, and more particularly to an object identification device and an object identification system suitable for identifying an object put in a trash can.

Patent Documents 1-3 disclose a technique for constructing a communication path such as a sensor network between trash cans installed in public facilities such as stations and utilizing it for anti-terrorism measures and environmental beautification.

Patent Document 1 describes a technology for measuring the capacity of a trash can with a sealed trash can on a weight basis, and when a weight increase is detected, a sensor network is used to notify that a foreign substance has been added. It has been disclosed.

Patent Document 2 states that a security label composed of an RFID member having an IC chip and an antenna element and a communication shielding layer capable of shielding RFID communication is capable of communication by destroying only the communication shielding layer at the time of unauthorized opening. The anti-terrorism technology for monitoring the input of suspicious objects is disclosed by using this security label to seal the garbage input port of the trash can installed in a public place.

Patent Document 3 discloses a trash can monitoring system in which a trash detection sensor that transmits a detection wave in the horizontal direction is provided on the inner surface of the trash can, and the dust accumulation state is determined based on the signal detected by the dust detection sensor. ing.

Japanese Unexamined Patent Publication No. 2009-249118 Japanese Unexamined Patent Publication No. 2018-013891 JP-A-2017-206333

In Patent Document 1, it is possible to detect that some object has been thrown into the sealed trash can, but it is not possible to identify the thrown object. In addition, sensor networks have narrow wireless coverage, which limits the location of trash cans.

In Patent Document 2, since the input port of the trash can must be sealed with a security label, it is not suitable for daily use. Further, in Patent Document 2, it is possible to detect that the security label is broken and an object is thrown in, but it is not possible to identify the thrown-in object.

In Patent Document 3, although the amount of garbage put into the trash can can be detected, it is not possible to identify what kind of garbage is put into the trash can, so that it is not particularly suitable for anti-terrorism measures for identifying the put of dangerous substances.

An object of the present invention is to solve the above-mentioned technical problems and to provide an object identification device and an object identification system capable of accurately identifying a thrown object without interfering with its normal use even when applied to a trash can. be.

In order to achieve the above object, the present invention is characterized in that it has the following configurations.

(1) The object identification device of the present invention is known to the housing unit as a loading detecting means for detecting that an object has been loaded into the housing unit, a weight detecting means for detecting the weight of the housing unit and outputting a weight signal, and a weight detecting means for outputting a weight signal. The time series of the weight signal output by the weight detecting means when the object is thrown is learned in advance for each object, and the unknown is based on the time series of the weight signal detected when the unknown object is thrown. It is characterized by having a prediction model for predicting an object and an identification means for identifying the object by applying a time series of weight signals detected each time the object is thrown to the prediction model .

(2) The object identification device of the present invention is characterized in that it includes a prediction model that predicts the object based on the time series of the weight signal when the object is thrown.

(3) The object identification device of the present invention is characterized in that the weight detecting means includes a plurality of weight sensors and identifies an object based on the time series of the weight signal output by each weight sensor.

(4) In the object identification device of the present invention, the time series of weight signals output by a plurality of weight sensors are arranged in series to construct weight data for one record, and the identification means is based on the weight data for one record. It is characterized in that it identifies the object.

(5) The object identification device of the present invention is characterized in that it is further provided with an alarm means for issuing a predetermined alarm for the identification result.

(6) The object identification system of the present invention is configured by connecting at least one object identification device that identifies an object thrown into the accommodation unit and a monitoring server so as to be able to communicate with each other, and the object identification device is the accommodation unit. A throwing detection means that detects that an object has been thrown into the housing, a weight detecting means that detects the weight of the housing and outputs a weight signal, and the weight detecting means that outputs when a known object is thrown into the housing. A prediction model that learns the time series of the weight signal to be performed in advance for each object and predicts the unknown object based on the time series of the weight signal detected when the unknown object is thrown, and the object is thrown . It is provided with an identification means for identifying the object by applying the time series of the weight signal detected for each to the prediction model and a means for transmitting the identification result to the monitoring server, and the monitoring server identifies each object. It is characterized in that it is provided with a means for determining that the object to be notified has been thrown in based on the identification result transmitted from the device.

(7) The object identification system of the present invention is characterized in that the monitoring server is further provided with a means for notifying a predetermined destination of an alarm when it is determined that the object to be notified has been thrown in.

(8) The object identification system of the present invention is characterized in that the weight detecting means includes a plurality of weight sensors and identifies an object based on the time series of the weight signal output by each weight sensor.

(9) In the object identification system of the present invention, the time series of weight signals output by a plurality of weight sensors are arranged in series to construct weight data for one record, and the object is generated based on the weight data for one record. It is characterized by the point of identification.

(10) The object identification system of the present invention is characterized in that the monitoring server further includes a means for generating the prediction model and a means for distributing the generated prediction model to each object identification device.

(11) The object identification system of the present invention further includes means for transmitting the time series of the weight signal detected by the object identification device when the object is thrown to the monitoring server, and the monitoring server provides each object. It is provided with a weight data collecting means for accumulating a time series of weight signals transmitted from the identification device and a means for generating and / or updating the prediction model using the time series of the accumulated weight signals as a white list. There is a feature.

(12) In the object identification system of the present invention, the object identification device is provided with a storage unit for each attribute of the object to be loaded, and at least a charging detection means and a weight detection means are provided for each storage unit, and detection is performed for each storage unit. The time series of the weight signal is transmitted to the monitoring server in association with the attribute of the corresponding object, and the monitoring server sends the time series of the weight signal associated with the attribute of each object to the object of the attribute. It is characterized in that a prediction model is generated and / or updated as teacher data for identifying.

According to the present invention, the following effects are achieved.

(1) According to the object identification device of the present invention, the weight signal of the accommodation unit detected when the object is input to the accommodation unit is input to the accommodation unit based on the fact that the time series of the weight signal is unique to the object. Since the object is identified, the input object can be accurately identified with a simple configuration.

(2) According to the object identification device of the present invention, the time series of the weight signal of the accommodating portion detected when the object is thrown into the accommodating portion is applied to the prediction model to identify the object, so that each object is identified. Accurate identification based on unique and potential features is possible.

(3) According to the object identification device of the present invention, since the object is identified based on the time series of the weight signals output by the plurality of weight sensors, the characteristics of each object are effectively changed in the time series of the detected weight signals . It will be possible to reflect.

(4) According to the object identification device of the present invention, the weight data for one record is constructed by arranging the time series of the weight signals output by the plurality of weight sensors in series, so that the prediction model based on the plurality of weight signals is constructed. Will be easier.

(5) According to the object identification device of the present invention, an alarm can be issued as a predetermined alarm for the identification result, for example, by at least one of an alarm sound, a flash light, and a message output. When is identified as a dangerous substance, an alarm is issued to the surroundings to ensure safety.

(6) According to the object identification system of the present invention, not only the prediction model learned by the monitoring server can be easily provided to each object identification device, but also the monitoring server can obtain the object identification result by each object identification device. You will be able to centrally manage.

(7) According to the object identification system of the present invention, when any object identification device identifies a dangerous substance or the like, the monitoring server can notify a predetermined destination of an alarm, so that an administrator located at a remote location can notify the alarm. It becomes easy to report to such as.

(8) According to the object identification system of the present invention, the monitoring server collects a time series of weight signals detected by each object identification device in normal times, and uses this as a whitelist to generate and / or update a prediction model. , It will be possible to easily and a large amount of teacher data for machine learning.

(9) According to the object identification system of the present invention, for example, when an object identification device is applied to each storage unit of a sorting trash box, the time series of the weight signal detected for each storage unit into which the sorted object is input is determined. Since it is used for training the prediction model as teacher data unique to each separated object, it becomes possible to easily and a large amount of teacher data for machine learning for each attribute of the object.

It is a functional block diagram of the object identification apparatus which concerns on 1st Embodiment of this invention. It is a side view [the figure (a)] of the trash can and the plan view [the figure (b)] of the bottom of the accommodating part. It is a figure which showed an example of the bottom plate of the accommodating part. It is a figure which showed the example which integrated the output of a plurality of weight sensors and generates the weight data for one record. It is a block diagram of the object identification system which concerns on 2nd Embodiment of this invention. It is a functional block diagram of the object identification apparatus in 2nd Embodiment. It is a functional block diagram of the trash can monitoring server in the 2nd Embodiment. It is a functional block diagram of the trash can monitoring server in the 3rd Embodiment. It is a block diagram of the object identification system which concerns on 4th Embodiment of this invention. It is a functional block diagram of the object identification apparatus in 4th Embodiment. It is a functional block diagram of the trash can monitoring server in 4th Embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a functional block diagram showing a configuration of a main part of an object identification device 10 mounted on a trash can 1 according to a first embodiment of the present invention and identifying an object thrown into the storage part 12. FIG. 2A is a cross-sectional view of the trash can 1, and FIG. 2B is a plan view of the accommodating portion 12 of the trash can 1 from above.

As shown in FIG. 2A, the throw-in detection unit 101 is, for example, a motion sensor provided inside the upper surface of the trash can 1 or an acceleration sensor that detects the vibration of the trash can 1, and an object can be seen from the trash can 11. It is detected that the container 12 has been charged. The weight detection unit 102 measures the weight of the contained object for a predetermined measurement period (for example, for 4 seconds in a 0.1 second cycle) in a predetermined sampling cycle when the loading detection unit 101 detects the loading of the object. Then, the time series of the weight signal is output.

In the present embodiment, the weight detection unit 102 includes a bottom plate 13 and a plurality of weight sensors 1021. As shown in FIGS. 2A and 2B, the plurality of weight sensors 1021 are arranged apart in a plane so as to support the bottom plate 13. As shown in FIG. 3, for example, the bottom plate 13 may be divided into substantially the same shape for each weight sensor 1021, or may be divided so that the areas of each weight sensor 1021 are equal.

The prediction model storage unit 103 stores in advance a prediction model M that identifies an object based on a time series of weight signals output by each weight sensor 1021 of the weight detection unit 102 when an object is thrown into the trash box 1. Has been done. The identification unit 104 identifies the object by applying the time series of the weight signal detected when the object is thrown into the prediction model M.

The alarm unit 105 can issue a warning to a person near the trash can or the manager of the trash can. For example, if the identification result of the thrown object is a dangerous substance such as an explosive or a flammable object, there is a possibility that the dangerous substance is thrown in by generating an alarm sound or an alarm message, or emitting a flash light. Announce that there is something. Further, if the trash can 1 is provided with a display, a message indicating that a dangerous substance may be thrown in may be displayed.

In the prediction model M, various objects are stored in the accommodating portion 12 in various amounts in a time series of weight signals detected when an object having a known attribute is thrown in for each specification or standard of the trash box. It can be generated by repeatedly measuring while simulating the state and performing a large amount of machine learning as teacher data on the relationship between the input known object and the time series of the weight signal.

Further, in the present embodiment, since the trash box 1 includes a plurality of weight sensors 1021, as shown in FIG. 4, the time series of the weight signals output by each weight sensor 1021 are arranged in series and integrated into one record. By handling the time series of the weight signal for one record as one weight data, it is easy to apply the prediction model when a plurality of weight sensors are provided.

In such a configuration, an object is thrown into the accommodating section 12 from the throwing port 11 of the trash can 1, and when this is detected by the throwing detection section 101, a measurement command is output to the weight detecting section 102. The weight detection unit 102 acquires the time series of the weight signal by sampling the output of each weight sensor 1021 at a predetermined cycle for a predetermined measurement period in response to the measurement command, and integrates these to obtain the one record. Generate weight data for minutes.

The identification unit 104 identifies the input object by applying this weight data to the prediction model M. The alarm unit 105 issues an alarm to warn the surroundings when the identification result is an alarm target such as a dangerous substance registered in advance.

According to the present embodiment, the object thrown into the accommodating portion 12 is based on the fact that the time series of the weight signal output by the weight sensor 1021 each time the object is charged into the accommodating portion 12 is unique to the object. Since it is identified, it becomes possible to accurately identify the input object with a simple configuration.

Further, according to the present embodiment, since the time series of the weight signal of the accommodating portion detected when the object is thrown into the accommodating portion 12 is applied to the prediction model M to identify the object, it is unique to each object. Moreover, accurate identification based on potential features becomes possible.

Further, according to the present embodiment, since the object is identified based on the time series of the weight signals output by the plurality of weight sensors 1021, the characteristics of each object can be effectively reflected in the time series of the detected weight signals . become.

Further, according to the present embodiment, since the weight data for one record is constructed by arranging the time series of the weight signals output by the plurality of weight sensors 1021 in series, it is easy to construct the prediction model M based on the plurality of weight signals. become.

Further, according to the present embodiment, an acoustic and / or optical alarm can be issued for the identification result, so that an alarm is issued to the surroundings when the thrown object is identified as a dangerous substance. And you will be able to ensure safety.

FIG. 5 is a block diagram showing a configuration of an object identification system according to a second embodiment of the present invention, and here, a system for identifying an object discarded in the trash can 1 will be described as an example. The identification system of the present embodiment is configured by connecting a plurality of trash cans 1 and a trash can monitoring server 2 via a network, but here, for convenience, only one trash can 1 is shown.

FIG. 6 is a block diagram showing the configuration of the main part of the object identification device 10 mounted on each trash can 1, and the same reference numerals as those above represent the same or equivalent parts, so the description thereof will be omitted.

In the present embodiment, instead of the alarm unit 105, the communication unit 106 exhibits a transmission / reception function such as establishing a communication session with the trash can monitoring server 2 and transmitting an object identification result or the like to the trash can monitoring server 2. It is characterized by the fact that it is provided.

FIG. 7 is a functional block diagram showing the configuration of the main part of the trash can monitoring server 2, and the communication unit 201 cooperates with the communication unit 106 of the object identification device 10 mounted on each trash can 1 for a communication session. To establish. The prediction model generation unit 203 applies machine learning to a large amount of teacher data separately input from the input unit 202, and the object is based on the time series of the weight signal detected when the object is thrown into the trash box. Generate a predictive model M that identifies.

The prediction model distribution unit 204 distributes the generated prediction model to each trash can 1. The reporting unit 205 determines the identification result of the input object notified from each trash can container 1 and reports to a predetermined destination.

In such a configuration, the trash can monitoring server 2 is input with a large amount of teacher data obtained by learning the relationship between a known object thrown into each trash can 1 and the time series of weight signals under various environments. It is input from the unit 202. The prediction model generation unit 203 generates a prediction model M that identifies an input object from a time series of weight signals by machine learning a large number of input teacher data. The generated prediction model M is distributed to each trash can 1 by the prediction model distribution unit 204 and stored in each prediction model storage unit 103.

In each trash can 1, when the loading of an object is detected by the loading detecting unit 101, the weight measuring unit 102 performs weight measurement and generates weight data for one record. The identification unit 104 identifies the input object by applying the data string for one record to the prediction model M delivered in advance, and notifies the identification result from the communication unit 106 to the trash can monitoring server 2.

In the trash can monitoring server 2, the notification unit 205 determines the notified identification result, and if the thrown object is a notification target object such as a dangerous substance, the notification unit 205 accesses a predetermined destination and reports the identification result.

According to this embodiment, not only the prediction model M learned by the monitoring server 2 can be easily provided to the object identification device 10 of each trash can 1, but also the object identification result by each object identification device 10 can be obtained by the monitoring server. 2 can be centrally managed.

Further, according to the present embodiment, when the object identification device 10 mounted on any of the trash cans 1 identifies a dangerous substance or the like, the monitoring server 2 can notify a predetermined destination of an alarm, so that it is a remote location. It will be easier to report to the managers, etc. located in.

FIG. 8 is a diagram showing the configuration of the main part of the trash can monitoring server 2 adopted by the object identification system according to the third embodiment of the present invention, and the same reference numerals as those described above represent the same or equivalent parts. , The description is omitted.

In the second embodiment described above, it has been described as specifically identifying what the thrown object is, but if the purpose is to identify dangerous substances used for terrorist acts such as explosives and ignited objects. For example, it is sufficient to make a binary judgment as to whether or not the thrown object is one of these dangerous goods.

On the other hand, considering that there is a high possibility that dangerous goods will be thrown into the trash can 1 during the limited period ahead of major events, all objects thrown in during other periods should be regarded as non-dangerous goods. Can be done.

Therefore, in the present embodiment, all objects thrown in within the period in which the possibility of throwing dangerous goods is generally considered to be non-dangerous goods, and the time series of the weight signals accumulated during that period is registered in the whitelist. do. Then, using this whitelist, we try to build a predictive model that identifies dangerous goods.

If it is confirmed that the dangerous goods have been thrown within the period when the dangerous goods are unlikely to be thrown in, the time series of the weight signal estimated to correspond to the dangerous goods may be deleted from the whitelist. desirable. Therefore, it is desirable to easily eliminate the trash can ID of the transmission source by registering the trash can ID of the transmission source in the time series of the weight signal to be registered in the white list.

The present embodiment is characterized in that the trash can monitoring server 2 adopted by the two embodiments is provided with a weight data collecting unit 206 that collects and stores a time series of weight signals from each trash can 1. Further, the prediction model generation unit 203 is characterized in that the prediction model generation unit 203 is provided with an update unit 2031 that updates the prediction model M by using the time series of a large amount of weight signals stored in the weight data collection unit 206 as a whitelist. ..

In such a configuration, in each trash can 1, when the throwing detection unit 101 detects the throwing of an object, the weight detecting unit 102 performs weight measurement and constructs weight data for one record. The identification unit 104 discriminates whether or not the input object is a dangerous substance by applying the weight data for one record to the prediction model M, and the identification result is combined with the weight data for one record in the communication unit 106. Notifies the trash box monitoring server 2 from.

In the trash can monitoring server 2, the weight data transmitted from each trash can 1 is stored in the weight data collecting unit 206. In the prediction model generation unit 203, the update unit 2031 updates the prediction model M by performing machine learning using a large amount of weight data accumulated in the weight data collection unit 206 as a whitelist. The updated prediction model M is distributed to each trash can 1 by the prediction model distribution unit 204 and is updated and stored in each prediction model storage unit 103.

According to the present embodiment, the monitoring server 2 collects a time series of weight signals detected by the object identification device 10 of each trash box 1 in normal times, and uses this as a whitelist to generate and / or update the prediction model M. , It will be possible to easily and a large amount of teacher data for machine learning.

FIG. 9 is a block diagram showing a configuration of an object identification system according to a fourth embodiment of the present invention, in which a plurality of trash cans including at least one separated trash can 4 for separating and collecting garbage and a non-separable general trash can 1. It is configured by connecting to the trash can monitoring server 2 via a network. In the present embodiment, it is assumed that the sorting trash can 4 is provided with independent garbage storage portions 12a, 12b, 12c so that the "can", "PET bottle", and "magazine" can be separated.

When the sorting trash can 4 is installed, there is a high probability that an empty can will be loaded into the empty can storage unit 12a, a PET bottle storage unit 12b will have a high probability of a PET bottle being loaded, and the magazine storage unit 12c will have a magazine. It is empirically recognized that the probability of being put in is high.

Therefore, in the present embodiment, when an object is thrown into the empty can accommodating portion 12a, the object is regarded as an empty can, and the time series of the detected weight signal is collected as unique teacher data when the object is an empty can. do. Similarly, the time series of the weight signal detected by the PET bottle storage unit 12b is collected as teacher data for PET bottle identification, and the time series of the weight signal detected by the magazine storage unit 12c is used as the teacher data for magazine identification. collect.

Then, by machine learning the large amount of teacher data collected in this way, a predictive model that identifies whether the input object is an empty can, a PET bottle, or a magazine based only on the time series of the weight signal. Is generated and updated, and this is applied to the non-sorted general trash can 1 to improve the identification in the general trash can 1.

FIG. 10 is a functional block diagram showing the configuration of the main part of the sorted trash can 4 in the fourth embodiment, and the same reference numerals as those described above represent the same or equivalent parts, and thus the description thereof will be omitted.

In the present embodiment, the input detection units 101a, 101b, 101c and the weight detection units 102a, 102b, 102c are separately provided for each of the storage units 12a, 12b, 12c of the "empty can", "PET bottle", and "magazine". The communication unit 106, the identification unit 104, and the prediction model M are shared by each accommodation unit 12.

FIG. 11 is a block diagram showing the configuration of the main part of the trash can monitoring server 2 according to the fourth embodiment, and the same reference numerals as those described above represent the same or equivalent parts, and thus the description thereof will be omitted.

In the present embodiment, the weight data collecting unit 206 is divided into an empty can area 206a, a PET bottle area 206b, and a magazine area 206c, and the empty can area 206a is the weight provided in the storage unit 12a for separately collecting the empty cans. The weight data detected by the detection unit 102a is collected. Similarly, the weight data detected by the weight detection unit 102b provided in the storage unit 12b for separately collecting the PET bottles is collected in the PET bottle area 206b, and the magazines are separately collected in the magazine area 206c. The weight data detected by the weight detecting unit 102c provided in the accommodating unit 12c is collected.

In the update unit 203a of the prediction model generation unit 203, the time series of the weight signal stored in the empty can area 206a is the teacher data for identifying the empty can, and the time series of the weight signal stored in the PET bottle area 206b is the PET bottle. The prediction model M is updated by periodically re-learning the identification teacher data and the time series of the weight signals stored in the magazine area 206c as the magazine identification teacher data. The prediction model distribution unit 204 distributes the updated latest prediction model M to the unsorted general trash can 1.

According to the present embodiment, the time series of the weight signal detected for each accommodating portion 12 into which the separated objects are input is used for learning the prediction model M as teacher data peculiar to each separated object. It becomes possible to easily and a large amount of teacher data for machine learning for each attribute of an object.

1 ... Recycle bin, 2 ... Garbage bin monitoring server, 4 ... Sorted waste bin, 11 ... Garbage input port, 12 ... Storage unit, 13 ... Bottom plate, 101 ... Input detection unit, 102 ... Weight detection unit, 103 ... Prediction model storage unit, 104 ... Identification unit, 105 ... Alarm unit, 106 ... Communication unit, 201 ... Communication unit, 202 ... Input unit, 203 ... Prediction model generation unit, 204 ... Prediction model distribution unit, 205 ... Notification unit, 206 ... Weight data collection unit, 1021 ... Weight sensor

Claims (12)

In the object identification device that identifies the object thrown into the housing
A throwing detection means that detects that an object has been thrown into the containment unit,
A weight detecting means that detects the weight of the housing and outputs a weight signal ,
The time series of the weight signal output by the weight detecting means when a known object is thrown into the accommodating portion is learned in advance for each object, and based on the time series of the weight signal detected when an unknown object is thrown. And a prediction model that predicts the unknown object,
An object identification device comprising an identification means for identifying an object by applying a time series of a weight signal detected each time an object is thrown to the prediction model . The weight detecting means includes a plurality of weight sensors, and the weight detecting means includes a plurality of weight sensors.
The object identification device according to claim 1, wherein the identification means identifies an object based on a time series of weight signals output by each weight sensor. The weight detecting means constructs weight data for one record by arranging the time series of weight signals output by each weight sensor in series.
The object identification device according to claim 2 , wherein the identification means identifies the object based on the weight data for one record. The object identification device according to any one of claims 1 to 3 , further comprising an alarm means for issuing a predetermined alarm to the identification result. The object identification device according to claim 4 , wherein the alarm means issues an alarm by at least one of an alarm sound, a flash light, and a message output. In an object identification system configured by connecting at least one object identification device that identifies an object thrown into an accommodation unit and a monitoring server so that they can communicate with each other.
The object identification device
A throwing detection means that detects that an object has been thrown into the containment unit,
A weight detecting means that detects the weight of the housing and outputs a weight signal ,
The time series of the weight signal output by the weight detecting means when a known object is thrown into the accommodating portion is learned in advance for each object, and based on the time series of the weight signal detected when an unknown object is thrown. And a prediction model that predicts the unknown object ,
An identification means for identifying the object by applying the time series of the weight signal detected each time the object is thrown into the prediction model.
A means for transmitting the identification result to the monitoring server is provided.
The monitoring server
An object identification system characterized in that it is provided with a means for determining that an object to be notified has been thrown in based on the identification result transmitted from each object identification device. The monitoring server
The object identification system according to claim 6 , further comprising means for notifying a predetermined destination of an alarm when it is determined that the object to be notified has been thrown in. The weight detecting means includes a plurality of weight sensors, and the weight detecting means includes a plurality of weight sensors.
The object identification system according to claim 6 or 7 , wherein the identification means identifies an object based on a time series of weight signals output by each weight sensor. The weight detecting means constructs weight data for one record by arranging the time series of weight signals output by each weight sensor in series.
The object identification system according to claim 8 , wherein the identification means identifies the object based on the weight data for one record. The monitoring server
The means for generating the prediction model and
The object identification system according to any one of claims 6 to 9 , further comprising means for delivering the generated prediction model to each object identification device. The object identification device
Further provided with means for transmitting the time series of the weight signal detected when the object is thrown to the monitoring server.
The monitoring server
A weight data collection means that accumulates a time series of weight signals transmitted from each object identification device,
The object identification system according to any one of claims 6 to 10 , further comprising means for generating and / or updating the prediction model using the time series of the accumulated weight signals as a whitelist. At least one of the object identification devices
Equipped with a housing for each attribute of the object to be thrown
At least a loading detection means and a weight detection means are provided for each accommodating unit.
The time series of the weight signal detected for each accommodating unit is transmitted to the monitoring server in association with the attribute of the corresponding object.
The monitoring server is characterized in that the time series of the weight signal associated with the attribute of each object is generated and / or the prediction model is updated as the teacher data for identifying the object of the attribute . The object identification system according to any one of 11 .

Images