阅读说明：本技术 模块化回收系统 (Modular recovery system ) 是由 钟金华 于 2021-11-03 设计创作，主要内容包括：本发明公开了一种模块化回收系统,包括：主控箱体模块和回收箱体模块。通过将传统的回收箱体变更为主控箱体和回收箱体的模块系统,在场地限制较小的区域,可以通过多设置回收箱体模块的方式,提供更多更大的回收空间,而在场地限制较大的区域,可以单独只设置一个回收箱体模块,从而适应场地需要,为回收系统的布置提供了更多的灵活性。(The invention discloses a modular recycling system, comprising: the main control box body module and the recovery box body module. Through the module system who changes traditional recovery box into master control box and recovery box, in the less region of place restriction, can provide more bigger recovery space through the mode of setting up more recovery box modules, and in the great region of place restriction, can only set up a recovery box module alone to adaptation place needs provide more flexibility for recovery system's arrangement.)

1. A modular recycling system, comprising:

the main control module comprises a main control box body provided with a main controller;

the recycling bin body module is detachably connected with the main control module repeatedly, a delivery door and a recycling door are sequentially arranged on the front side of the recycling bin body module from top to bottom, the recycling bin body module comprises a recycling bin arranged below the inside of the recycling bin body and a camera arranged in the recycling bin, the recycling bin is respectively communicated with the delivery door and the recycling door, a bottom plate is arranged at the bottom end of the recycling bin, a weight sensor is arranged at the bottom end of the bottom plate and used for weighing recycled articles on the bottom plate, and the camera is arranged above the bottom plate and is aligned with the recycled articles on the bottom plate and used for acquiring RGB data and IR data of the recycled articles;

the main controller comprises an article identifier and is used for acquiring the data of the recycled articles shot by the camera, and identifying and classifying the recycled articles.

2. The modular recycling system according to claim 1, wherein: the article identifier executes a recycled article classification identification method as follows:

normalizing the RGB data and the IR data respectively;

inputting the normalized RGB data into an RGB feature extraction network to obtain an RGB feature map, wherein the RGB feature map characterizes the outline and the texture of the recycled article;

inputting the normalized IR data into an IR feature extraction network to obtain an IR feature map, wherein the IR feature map characterizes the material of the recycled item, and the IR feature map and the RGB feature map are equal in length and width;

combining the RGB feature map and the IR feature map into a feature data set, and inputting the feature data set into a classification network to obtain the classification of the recycled articles, wherein the classification of the recycled articles comprises metal garbage, textile garbage and glass garbage, and the classification of the metal garbage at least comprises one sub-classification of iron, aluminum and copper.

3. The modular recycling system according to claim 1, wherein: and the charging system is used for generating payment amount according to the classification and the weight of the recovered articles and paying.

4. The modular recycling system according to claim 1, wherein: the device also comprises a notification system which notifies cleaning personnel to recover the recovered articles according to the condition that the total weight of the recovered articles exceeds a preset value.

5. The modular recycling system according to claim 1, wherein: the RGB feature extraction network and the IR feature network respectively comprise a first convolution network, the first convolution network shares parameters with a depth residual error network, and the training method of the RGB feature extraction network and the IR feature network specifically comprises the following steps:

extracting a first convolutional network: training the deep residual error network through a training set; a first convolutional network composed of parameters discarded by the deep residual network; scoring the first convolutional network through a test set;

and circularly executing extraction of a first convolutional network on the deep residual error network to obtain multiple groups of first convolutional networks and scores corresponding to the first convolutional networks.

And respectively appointing the most graded convolution layer in the multiple groups of first convolution networks as the convolution layer in the RGB characteristic extraction network or the IR characteristic network.

6. The modular recycling system according to claim 1, wherein: the main control box body module is fixedly connected with at least one recovery box body module.

7. The modular recycling system according to claim 1, wherein: the main control box module and the recovery box module are respectively provided with an interface, the interface at least comprises a data end interface and a power end interface, and the ports are electrically connected through data lines.

8. The modular recycling system according to claim 1, wherein: the master controller further comprises a wireless transmission unit, and the master controller is connected with the cloud end through the wireless transmission unit.

9. The modular recycling system according to claim 1, wherein: the top is equipped with L type support in the box L type support bottom is equipped with the camera, including IR camera and RGB camera, just the camera head is close recovery door sets up.

10. The modular recycling system according to claim 7, wherein: the recovery tank module comprises at least two of the interfaces.

Technical Field

The invention relates to the technical field of recovery, in particular to a modular recovery system.

Background

At present, with the increasing attention of people to the environmental protection problem, the recycling technology is greatly developed, and various modular recycling systems such as a clothes recycling box and a mobile phone recycling station are proposed for recycling. An automated modular recycling system for recycled items is required to identify different types of recycled items so as to recycle the recycled items according to different types of recycled items or provide the cost for the processing of the recycled items for customers.

However, most of the existing recycling boxes are single-body type, and cannot be installed individually according to the situation of the site, and therefore, a modular recycling system is needed to be provided to solve the above problems.

Disclosure of Invention

The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a modularized recycling system which is reasonable in layout and simple to manufacture.

The technical scheme is as follows: a modular recycling system, comprising:

the main control module comprises a main control box body provided with a main controller;

the recycling bin body module is detachably connected with the main control module repeatedly, a delivery door and a recycling door are sequentially arranged on the front side of the recycling bin body module from top to bottom, the recycling bin body module comprises a recycling bin arranged below the inside of the recycling bin body and a camera arranged in the recycling bin, the recycling bin is respectively communicated with the delivery door and the recycling door, a bottom plate is arranged at the bottom end of the recycling bin, a weight sensor is arranged at the bottom end of the bottom plate and used for weighing recycled articles on the bottom plate, and the camera is arranged above the bottom plate and is aligned with the recycled articles on the bottom plate and used for acquiring RGB data and IR data of the recycled articles;

the main controller comprises an article identifier and is used for acquiring the data of the recycled articles shot by the camera, and identifying and classifying the recycled articles.

Further, the article identifier performs a recycled article classification identification method as follows:

normalizing the RGB data and the IR data respectively;

inputting the normalized RGB data into an RGB feature extraction network to obtain an RGB feature map, wherein the RGB feature map characterizes the outline and the texture of the recycled article;

inputting the normalized IR data into an IR feature extraction network to obtain an IR feature map, wherein the IR feature map characterizes the material of the recycled item, and the IR feature map and the RGB feature map are equal in length and width;

merging the RGB feature map and the IR feature map into a feature data set, and inputting the feature data set into a classification network to obtain a classification of the recycled articles, wherein the classification of the recycled articles comprises metal waste, textile waste and glass waste, and the classification of the metal waste at least comprises one sub-classification of iron, aluminum and copper;

and further, the system also comprises a charging system which is used for generating and paying the payment amount according to the classification and the weight of the recycled articles.

Further, the device also comprises a notification system which notifies cleaning personnel to recycle the recycled articles according to the condition that the total weight of the recycled articles exceeds a preset value.

Further, the RGB feature extraction network and the IR feature network respectively include a first convolution network, the first convolution network shares a parameter with a deep residual error network, and the training method of the RGB feature extraction network and the IR feature network specifically includes:

extracting a first convolutional network: training the deep residual error network through a training set; a first convolutional network composed of parameters discarded by the deep residual network; scoring the first convolutional network through a test set;

and circularly executing extraction of a first convolutional network on the deep residual error network to obtain multiple groups of first convolutional networks and scores corresponding to the first convolutional networks.

And respectively appointing the most graded convolution layer in the multiple groups of first convolution networks as the convolution layer in the RGB characteristic extraction network or the IR characteristic network.

Furthermore, the main control box body module is fixedly connected with at least one recovery box body module.

Furthermore, the main control box module and the recovery box module are respectively provided with an interface, the interface at least comprises a data terminal interface and a power terminal interface, and the ports are electrically connected through data lines.

Furthermore, the master controller also comprises a wireless transmission unit, and the master controller is connected with the cloud end through the wireless transmission unit.

Further, the top is equipped with L type support in the box L type support bottom is equipped with the camera, including IR camera and RGB camera, just the camera head is close retrieve the door setting.

Further, the recovery tank module comprises at least two of the interfaces.

Has the advantages that: the modularized recovery system changes the traditional recovery box body into a module system of a main control box body and a recovery box body, is limited in a small area in a field, can provide more and larger recovery space in a mode of additionally arranging the recovery box body modules, is limited in a large area in the field, can be independently provided with only one recovery box body module, thereby being suitable for the field requirement, providing more flexibility for the arrangement of the recovery system, and simultaneously collecting IR (infrared) and RGB (red, green and blue) image data of recovered articles due to the arrangement of a camera, so that the identification and classification are more accurate.

Drawings

FIG. 1 is a schematic perspective view of one embodiment of a recycling system of the present application;

FIG. 2 is a schematic plan view, partially in section, of a recovery tank module of the recovery system of FIG. 1;

FIG. 3 is an article sorting flow diagram of the article sorter of the recycling system of FIG. 1;

FIG. 4 is a flow chart of a training method of the RGB feature extraction network and the IR feature network of the step of FIG. 3;

FIG. 5 is an additional flow diagram of the item classifier classification process of FIG. 3;

fig. 6 is a schematic diagram according to the embodiment shown in fig. 1.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the technical solutions in the embodiments may be combined with each other, but must be based on the realization of the technical solutions by those skilled in the art, and when the technical solutions are contradictory or cannot be realized, the combination of the technical solutions should be considered to be absent and not to be within the protection scope of the present invention.

Referring to fig. 1-6, one embodiment of the modular recycling system of the present invention comprises: the device comprises a main control module 1 and a recovery box body module 2 which is detachably connected with the main control module 1 repeatedly. The main control module 1 comprises a main control box 10 provided with a main controller 3.

Retrieve 2 front sides of box module from last to being equipped with in proper order down and deliver door 21 and throw recovery door 22, retrieve box module 2 including set up in retrieve 2 interior below of box module recovery storehouse 200 with locate retrieve camera 4 in the storehouse 200, retrieve storehouse 200 intercommunication respectively deliver door 21 with throw recovery door 22, the bottom end is provided with bottom plate 23 in retrieving storehouse 200, bottom plate 23 bottom is provided with weight inductor 24 for weigh the recovery article on the bottom plate 23, camera 4 is located bottom plate 23 top is aimed at recovery article on the bottom plate 23 and be used for acquireing the RGB data and the IR data of retrieving article.

The master controller 3 includes an article identifier for acquiring the data of the recycled articles shot by the camera 4, and identifying and classifying the recycled articles, and preferably, the master controller 3 is a hardware or software processor having the above functions, and in this embodiment, may specifically be an android device.

According to the modularized recycling system, the traditional recycling box body is changed into a module system of the main control module 1 and the recycling box body module 2, a smaller area is limited in a field, more and larger recycling space can be provided by arranging more recycling box body modules 2, a larger area is limited in the field, only one recycling box body module 2 can be independently arranged, so that the modularized recycling system is suitable for the field requirement, more flexibility is provided for arrangement of the recycling system, and meanwhile, due to the arrangement of the camera 4, IR and RGB image data of recycled articles are collected at the same time, and identification and classification are more accurate.

Further, the main control module 1 is fixedly connected with at least one recovery box module 2. In this embodiment, the main control module 1 may integrate one recycling box module 2 into a main control box 10 with one recycling box, as shown in fig. 1, or may be an independent main control box 10 to be applied to different field space requirements, for example, when the space is relatively simple, the main control module 1 with the recycling box module 2 may be applied, and when the field space is relatively complex, the independent main control module 1 may be applied to control a large number of recycling box modules 2 and be dispersedly arranged, such structural changes still fall within the protection scope of the present invention.

Specifically, the article identifier executes a recycled article classification identification method as follows:

step S100: acquiring RGB data and IR data of a recyclable item and normalizing the RGB data and IR data, respectively;

step S200: inputting the normalized RGB data into an RGB feature extraction network to obtain an RGB feature map, wherein the RGB feature map represents the outline and texture of the recyclable item;

step S300: inputting the normalized IR data into an IR feature extraction network to obtain an IR feature map, wherein the IR feature map characterizes material of the recyclable item, and the IR feature map and the RGB feature map are equal in length and width;

step S400: merging the RGB feature map and the IR feature map into a feature data set, and inputting the feature data set into a classification network to obtain a classification of the recyclable item, wherein the recyclable item classification includes metal waste, textile waste, and glass waste, and the metal waste classification includes at least one sub-classification of iron, aluminum, and copper. In one embodiment, three subclasses of iron, aluminum and copper can be provided, and two subclasses of iron, aluminum and iron and copper can be provided.

And the charging system is used for generating payment amount according to the classification and the weight of the recovered articles and paying.

In some other embodiments, the method can be divided into the following modules:

a preprocessing module 100, configured to normalize the RGB data and the IR data, respectively;

an RGB extraction module 200 for inputting the normalized RGB data into an RGB feature extraction network to obtain an RGB feature map, wherein the RGB feature map represents the outline and texture of the recyclable item;

an IR extraction module 300 for inputting the normalized IR data into an IR feature extraction network to obtain an IR feature map, wherein the IR feature map characterizes material of the recyclable item, and the IR feature map and the RGB feature map are equal in length and width;

a classification module 400 for merging the RGB feature map and the IR feature map into a feature data set and inputting the feature data set into a classification network to obtain a classification of the recyclable item, wherein the recyclable item classification includes metal waste, textile waste and glass waste, and the metal waste classification includes iron, aluminum and copper sub-classifications.

Specifically, the top of retrieving storehouse 200 in the recovery box is equipped with L type support 41 bottom is equipped with camera 4, just camera 4 is close to post door 2121 setting, two kinds of cameras 4 of IR and RGB are integrated simultaneously to camera 4. Meanwhile, because the camera 4 is arranged close to the delivery door 2121, the recovered articles can be identified at the moment of being thrown into the delivery door 2121, so that the identification errors caused by rolling and other reasons after falling into the recovery bin 2002 are prevented, and the identification efficiency and accuracy are improved.

Further, the device also comprises a notification system which notifies cleaning personnel to recycle the recycled articles according to the condition that the total weight of the recycled articles exceeds a preset value. Preferably, instant messaging tool software such as WeChat and Payment treasures can be used for notification, and short messages and telephone notifications can also be used.

Referring to fig. 3, the RGB feature extraction network and the IR feature network respectively include a first convolution network, the first convolution network shares a parameter with a deep residual error network, and the training method of the RGB feature extraction network and the IR feature network specifically includes:

step S11: extracting a first convolutional network: training the deep residual error network through a training set; a first convolutional network composed of parameters discarded by the deep residual network; scoring the first convolutional network through a test set;

step S12: and circularly executing extraction of a first convolutional network on the deep residual error network to obtain multiple groups of first convolutional networks and scores corresponding to the first convolutional networks.

Step S13: and respectively assigning the highest scoring layers in the multiple groups of first convolutional networks as convolutional layers in the rgb feature extraction network or the IR feature network.

The RGB feature extraction network and the IR feature extraction network are trained respectively, the steps required by training are the same, the required training set and parameters are adjusted, and the requirements of the RGB feature extraction network or the IR feature extraction network are set. Taking an RGB feature extraction network as an example, the method trains a deep residual error network through a training set, in the training process, in order to prevent gradient disappearance, the deep residual error network randomly discards a part of parameters, and a first convolution network with smaller parameter scale is formed by combining the discarded part of parameters with a feature data set.

Along with the training, the first convolutional network is extracted once in each round of training, the parameters discarded each time by the depth residual error network are random, the discarded parameters are also trained, the overall performance of the acquired first convolutional network is also improved, and a plurality of first convolutional networks generated after a plurality of times of training are scored.

For the extracted plurality of first convolution networks, the highest one of the extracted plurality of first convolution networks is extracted as the convolution part in the RGB feature extraction network, and similarly, the convolution part in the IR feature extraction network is also extracted and determined in the same manner. The scheme can ensure the feature extraction precision, reduce the size of the RGB feature extraction network and the IR feature extraction network in response on the basis of ensuring the classification accuracy, and reduce the parameter quantity of the RGB feature extraction network and the IR feature extraction network so as to reduce the calculation overhead in the feature extraction process. The efficiency of recoverable article discernment is promoted.

In one embodiment, the accuracy of feature extraction of the first network obtained after several training is 87.4%, 88.6%, 92.7%, 94.1%, 96.2%, 97.5%, respectively, and the accuracy of feature extraction of the residual network after final training is 98.7%, the accuracy of feature extraction of the first network with the highest accuracy is already very close to the level of the residual network, and the computational overhead is greatly reduced.

Further, in the depth residual network, the scaling parameter γ in the last BN layer in each residual block is set to 0 to output a full 0 vector.

The parameter gamma is set to be 0 to output the all 0 vectors, so that the residual block outputs the all 0 vectors before residual connection, the scale of the residual network output matrix is reduced, and the training efficiency of the first matrix is improved. The scheme can accelerate the speed of deploying the recyclable object identification method in any environment.

Further, the learning rate of the deep residual error network is set to be exponentially decayed, and the lowest learning rate of the deep residual error network is set to be five percent of the initial learning rate.

The scheme keeps the learning rate of the depth residual error network, prevents overfitting from occurring in the training process of the first network, and improves the richness of the recyclable feature extracted by the first network so as to improve the accuracy of recyclable classification.

Further, in the process of training the depth residual error network through the training set, when a parameter in a convolution kernel in the depth residual error network is smaller than a preset value, the parameter is set to be 0, so that each convolution kernel forms a sparse matrix.

Specifically, the number of parameters in the depth residual error network and the first convolution network is reduced, for some parameters with values close to 0, obvious features cannot be provided, computer resources are occupied, the calculated amount of the neural network can be greatly reduced by setting the parameters to zero, meanwhile, the features of the recyclable article cannot be lost, and the recognition efficiency of the recyclable article is improved by the scheme.

Further, non-zero parameters in the sparse matrix are concentrated to form blocks.

According to the scheme, the extraction of the features can be concentrated in the image, the recoverable object can provide the positions of the sufficient features, and the accuracy of feature extraction is improved.

Specifically, the main control module 1 and the recycling bin module 2 are respectively provided with an interface (not shown in the figure), the interface at least includes a data terminal interface and a power terminal interface, and each of the ports is electrically connected through a data line. So, then guaranteed to retrieve and can realize quick dismouting through the data line between box module 2 and the host system 1, formed the modularization production installation to improve on-the-spot installation/dismantlement speed, raise the efficiency. And the recovery box body module 2 comprises at least two interfaces, thereby ensuring that the recovery box body can be infinitely expanded through a data line, and meeting the volume requirement on the recovered articles in a larger field space.

In this embodiment, the master controller 3 further includes a wireless transmission unit (not shown), and the master controller 3 is connected to a cloud end through the wireless transmission unit, so as to facilitate payment and notification. Accordingly, the present embodiment can also be extended to a recycling bin or a main control bin 10, such as a mobile phone recycling bin 5 or a light box for advertising, and such structural changes still fall within the scope of the present invention.

The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.