KR20230100051A - Deep learning-based smart automatic logistics warehousing system linked to user terminals - Google Patents

Abstract

The deep learning-based smart automatic logistics warehouse system linked with the user terminal has a drawer at the bottom to take out or insert items, and on the front and back of the inside, there are rails at regular intervals to insert and store pallets. A storage unit formed in the storage unit, a stacker that is located in the center of the storage unit and transported in vertical and horizontal directions of the storage unit, moves the pallet to the storage unit or stores the pallet, and is provided and stored in the storage unit. A photographing unit for photographing and measuring the size of an item to be stored, a sensor provided in the drawer of the storage unit and measuring the weight of the stored item, and controlling the operation of the stacker, the photographing unit, and the sensor, Based on the information measured by the sensor, a deep learning algorithm is applied to automatically select the location where the item is stored, and the edge controller interactively processes necessary information in the form of a chatbot to control the operation of the assigned edge controller It includes a user terminal that

Description

Deep learning-based smart automatic logistics warehousing system linked to user terminals}

The present invention relates to an automated warehouse, and more particularly, to a deep learning-based smart automated logistics warehouse system that works with a user terminal.

In general, a warehouse for storing goods is configured in a form in which a plurality of divided spaces are arranged, and as the number of kinds of stored goods increases, it is difficult for a worker to quickly find a storage location.

With the development of technology, a system that allows workers to easily find the storage location through methods such as product name search or barcode recognition using a computer system is being applied, but there is still a risk of safety accidents due to the problem that workers have to bring items directly. There was a problem in that it was difficult to greatly expand the loading space.

Accordingly, there is a technology for setting the storage location of the load using a computer, searching for the load stored in a certain position in a short time, and automatically withdrawing and withdrawing the load, such as Korean Patent Publication No. 10-2016-0067639 "Automatic Warehouse System". has been developed

Conventional automatic warehouse systems are always managed to be located at designated locations by automatic warehouse computer programming, and the allocated space does not change even if the size of the load is changed, so there is a problem of inefficient use of storage space.

That is, in the conventional automated warehouse system, pallets located in the shortest distance are designated and loaded, or the order of each pallet is designated and sequentially loaded according to programming, or the moving speed is fixed regardless of the load.

KR 10-2016-0067639 A

The present invention has been proposed to solve the above technical problem, and is a deep learning-based smart automation interlocking with a user terminal that controls the operation of an assigned edge control unit by interactively processing necessary information in the form of a chatbot through a user terminal. Provides a logistics warehouse system.

According to one embodiment of the present invention for solving the above problem, a drawer is provided at one end of the lower end to take out or insert items, and on the inner front and rear surfaces there are mounting rails at regular intervals so that pallets can be inserted and stored. A storage unit formed in the storage unit, a stacker that is located in the center of the storage unit and transported in vertical and horizontal directions of the storage unit, moves the pallet to the storage unit or stores the pallet, and is provided and stored in the storage unit. A photographing unit for photographing and measuring the size of an item to be stored, a sensor provided in the drawer of the storage unit and measuring the weight of the stored item, and controlling the operation of the stacker, the photographing unit, and the sensor, Based on the information measured by the sensor, a deep learning algorithm is applied to automatically select the location where the item is stored, and the edge controller interactively processes necessary information in the form of a chatbot to control the operation of the assigned edge controller Provided is a deep learning-based smart automatic logistics warehouse system that is linked with a user terminal including a user terminal.

In addition, the deep learning algorithm in the present invention is characterized by applying a genetic algorithm (Genetic Algorithm, GA).

In addition, in the present invention, it is characterized in that it further comprises an alignment unit that is formed on both sides of the drawer of the storage unit and aligns the load mounted on the pallet by pressing the outer surface to be located in the center of the pallet.

Further, in the present invention, the sensing unit is formed in the drawer and is characterized in that it consists of a plurality of weight sensors for measuring the weight of the load loaded on the pallet by measuring the weight of the pallet mounted on the drawer.

The deep learning-based smart automated logistics warehouse system linked with the user terminal according to an embodiment of the present invention can interactively process necessary information in the form of a chatbot through the user terminal to control the operation of the assigned edge control unit.

In addition, the system can automatically estimate the fire risk of an object moving around the system based on the photographing information of a plurality of surrounding cameras and automatically adjust the storage position of the article in consideration of the fire risk.

In addition, the system applies a deep learning algorithm to perform work in the shortest time during loading and order picking. That is, based on the information received from the edge controller, a deep learning algorithm is applied to automatically select a storage location.

The deep learning algorithm optimizes the loading site through analysis and learning of loading data, and autonomous driving learning is performed by predicting the loading quantity, load, and required amount.

In other words, it is possible to predict the amount of warehousing/shipping by time and shorten the movement time, and to provide work convenience and solve the warehousing and warehousing bottleneck with the minimum work space.

In addition, the deep learning-based smart automatic logistics warehouse system linked with the user terminal identifies the optimal movement line based on deep learning, and identifies the overall flow of logistics rather than sequential and pre-programmed logic to determine the loading location. Efficiency can be maximized by specifying

In addition, since the speed can be adjusted according to the load, the travel time can be minimized, and unnecessary movement is minimized by selecting the optimal location for each period and frequency of use.

1 is a conceptual diagram of a smart automatic logistics warehouse system based on deep learning that works with a user terminal of the present invention
Figure 2 is a configuration diagram of a deep learning-based smart automatic logistics warehouse system linked with a user terminal according to an embodiment of the present invention
Figure 2a is a detailed configuration diagram of a deep learning-based smart automatic logistics warehouse system linked with the user terminal of FIG.
Figure 3 is a comparison of the operation of the deep learning-based smart automatic logistics warehouse system linked with the user terminal of Figure 2
Figure 4 is a perspective view of a stacker-type warehouse according to an embodiment in a deep learning-based smart automatic logistics warehouse system linked with the user terminal of FIG. 2
5 is a perspective view showing a stacker of a loading warehouse
6 is a perspective view showing a stacker fork of a loading warehouse
7 is a perspective view showing the operation of the stacker fork of the loading warehouse

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings in order to describe in detail enough for those skilled in the art to easily implement the technical idea of the present invention.

1 is a conceptual diagram of a deep learning-based smart automated logistics warehouse system interworking with a user terminal of the present invention, and FIG. 2 is a configuration of a deep learning-based smart automated logistics warehouse system interworking with a user terminal according to an embodiment of the present invention. FIG. 2A is a detailed configuration diagram of a deep learning-based smart automated logistics warehouse system linked with the user terminal of FIG. 2, and FIG. 3 is a deep learning-based smart automated logistics warehouse system linked with the user terminal of FIG. 2 It is an action comparison.

Referring to FIGS. 1 to 3 , the deep learning-based smart automatic logistics warehouse system that works in conjunction with a user terminal applies a deep learning algorithm to perform work in the shortest time during loading and order picking.

The deep learning algorithm optimizes the loading site through analysis and learning of loading data, and autonomous driving learning is performed by predicting the loading quantity, load, and required amount.

In other words, it is possible to reduce the travel time for prediction and learning of warehousing/shipping requirements by time zone, and to provide work convenience with a minimum work space and to solve the warehousing/shipping bottleneck.

In this embodiment, the deep learning-based smart automatic logistics warehouse system that works with the user terminal,

It is configured to include a first automatic logistics loading warehouse group (1), a second automatic logistics loading warehouse group (2), and a management server (3).

The management server 3 collects all data from the edge control unit of the first automatic logistics warehouse group 1 and the edge control unit of the second automatic logistics warehouse group 2, and then considers the amount of operation of each edge control unit. The data processing capacity of each edge controller can be set independently.

That is, when the first edge control unit receives data from the photographing unit and conducts deep learning, the management server 3 receives some data, performs distributed deep learning processing, and outputs an intermediate result to the first edge control unit when the amount of calculation exceeds the reference value. and the final result is processed to be derived by the first edge control unit.

Basically, since the edge control unit is directly installed at the site where the automatic logistics warehouse group is installed, processing speed can be improved by reducing the amount of operation of the management server 3 by performing pre-assigned functions at the site.

The user terminal 4 can interactively process necessary information in the form of a chatbot to control the operation of the assigned edge controller 600 .

Here, the user terminal 4 is a general term for a device that a user can carry and use, such as a mobile phone, a smart phone, a smart pad, and the like.

The status board 5 is a large screen that simultaneously displays the progress of logistics loading and processing items of the plurality of edge controllers 600, and all progress can be monitored through the large screen installed in the work space.

The augmented reality glasses 6 are defined as a general term for a device in which a screen itself is variably transformed according to a user's action. When the user wears the augmented reality glasses 6 and gazes at the dashboard 5, the logistics loading progress displayed on the dashboard is converted into virtual reality or augmented reality and displayed in the user's field of view.

The edge control unit 600 may be composed of a work computer equipped with a monitor. If the monitor is a multi-monitor - a plurality of flat monitors are arranged at a predetermined angle - or a curved monitor, the augmented reality glasses 6 may be applied to the application. It may not be possible to capture the screen of a single recording image. In this case, the augmented reality glasses 6 are set to select panoramic shooting.

When panoramic shooting is selected, the wearer of the augmented reality glasses 6 may rotate the monitor to capture a 360-degree circular image or rotate the body or head in the direction of the monitor to capture a 360-degree circular image (curved surface image).

The photographed 360-degree circular image is transmitted to the management server 3 and processed, and the management server 3 divides the circular image for partial distortion relief and partial image processing to generate a flat image from the circular image. That is, the unit of the divided circular image (circular division unit image) may be divided in consideration of resolution, arc, pixel, image distortion, and the like.

A method of generating a flat image based on a circular segmentation unit image is as follows.

First, a pixel structure of a circular division unit image may be grasped. Reference focus pixels for the circular segmentation unit image of which the pixel structure is determined may be extracted. A reference focus pixel may be a pixel that does not require expansion or contraction. The pixel structure may be divided into pixels to be reduced and pixels to be expanded based on the reference focus pixel.

A correction process per pixel may be performed to eliminate partial distortion for generating a plane division unit image. For example, reduction may be performed by determining a reduction area for an area expanded based on a reference focus pixel, and expansion may be performed through interpolation after determining an expansion area for the reduced area. A plane segmentation unit image may be extracted through additional image processing in consideration of resolution, arc, pixel, image distortion, and the like. Through this method, a plurality of circular division unit images may be generated as a plurality of planar division unit images.

That is, when a panoramic image is transmitted, the management server 3 generates a plurality of plane division unit images as described above, and then identifies an object or marker on the screen, so the automatic identification probability is further increased.

The management server 3 receives the image transmitted from the augmented reality glasses 6 and proceeds with an image processor that recognizes an object in real time. The management server 3 feeds back the recognized object information to the augmented reality glasses 6 . That is, the process of object recognition proceeds in the management server 3, and the management server 3 feeds back only the recognized result as object information to the augmented reality glasses 6, thereby reducing the computational load of the augmented reality glasses 6. a diminishing effect occurs. At this time, the augmented reality glasses 6 transmit the image and at the same time transmit the image processing command desired to be processed to the management server 3, and only the resultant value can be fed back.

In this way, the management server 3 feeds back the recognized object information to the augmented reality glasses 6, and each object information includes an identification code pre-allocated for each object type and location information for each object's central area by time. do.

For example, when an object called a forklift license plate is recognized from an image on the screen (a video taken by a surrounding camera), an identification code pre-assigned to the forklift license plate and location information by time about the location (coordinates) of the center area of the forklift license plate are transmitted do.

For reference, the identification code includes an object code and an additional code. The object code is a code given to the shape of a forklift license plate, and the additional code is defined as coded additional data information such as a forklift license plate number.

Therefore, from the point of view of the augmented reality glasses 6, the object information transmitted from the management server 3 can be received without directly processing the image processing work, and the recognized number of the object called the forklift license plate and the moving position can be grasped. The management server 300 simultaneously recognizes a plurality of objects at the request of the augmented reality glasses 6, and can track each location in real time. These requests can be input to the user terminal 4 in the form of a chatbot.

In addition, the management server 3 feeds back the part without change in the image among the captured images, and the augmented reality glasses 6 automatically removes the part without change in the image, thereby providing its own storage capacity (memory of the augmented reality glasses 6). capacity) may be reduced.

In addition, the management server 3 divides the image into a plurality of regions, and then feeds back information capable of varying the stored image frame for each region to the augmented reality glasses 6 in consideration of the movement speed of the object.

On the other hand, the augmented reality glasses 6 may divide a pre-set shooting area into a plurality of sub-areas, and then independently preset sensing time, sensing objects, and sensing events for each area. Accordingly, when the edge control unit 100 has a plurality of monitors, each monitor can be divided into sub-regions, and then an object event to be monitored can be independently set for each sub-region.

At this time, if the processing speed of the augmented reality glasses 6 is fast, the object recognition operation may be performed on its own and the result may be displayed on the touch panel, but the image and object recognition commands are transmitted to the management server 3 and managed. It is preferable that the server 3 performs an object recognition operation and then feeds back only the results to the augmented reality glasses 6 .

In addition, when a map is displayed in an application of the user terminal 4 and a marker on the map is a part to be emphasized to the user, it may be displayed as a virtual object. That is, a virtual object corresponding to real image information is displayed, and when the virtual object is touched, preset information of the virtual object may be displayed in detail.

For example, a notification icon is displayed as a virtual object, and detailed information to be provided to the user is displayed when a corresponding part is touched.

In addition, the edge control unit may automatically estimate a fire in an object moving around the system, for example, a forklift, based on photographing information of a plurality of peripheral cameras. At this time, the main control unit may warn of the risk of fire based on the temperature of the engine room of the forklift as well as the emission amount and emission accumulation time of the exhaust gas. At this time, when the edge control unit detects a fire risk of the forklift through a deep learning algorithm, it controls the load loaded around the forklift to be moved to the farthest position from the forklift.

At this time, since the edge control unit collects information on the type of load, it can move flammable and inflammable materials to a distance with priority in consideration of the fire risk of the load. In addition, the edge control unit may control the fire propagation to be delayed by moving the load containing the fire extinguishing liquid disposed at a predetermined location closest to the fire risk area.

The management server 3 transmits intermediate results to the first edge control unit after receiving and distributing some data when the first edge control unit receives photographic data from surrounding cameras and recognizes an object, when the amount of operation exceeds the reference value. Then, the final result is processed to be derived by the first edge control unit.

For reference, it is preferable that each peripheral camera of the plurality of peripheral cameras sequentially transfers captured images to neighboring peripheral cameras so that a peripheral camera closest to the edge controller wirelessly transmits captured images to the edge controller. That is, each peripheral camera transmits data to the edge control unit in a manner of wirelessly relaying images taken by neighboring peripheral cameras.

3 is a comparison diagram of operation of a deep learning-based smart automated logistics warehouse system linked with the user terminal of FIG. 2 .

Referring to FIG. 3 , a deep learning-based smart automated logistics warehouse system that works with a user terminal according to an embodiment of the present invention applies a deep learning algorithm to perform work in the shortest time during loading and order picking. That is, based on the information received from the edge controller, a deep learning algorithm is applied to automatically select a storage location.

The deep learning algorithm optimizes the loading site through analysis and learning of loading data, and autonomous driving learning is performed by predicting the loading quantity, load, and required amount.

In other words, it is possible to predict the amount of warehousing/shipping by time and shorten the movement time, and to provide work convenience and solve the warehousing and warehousing bottleneck with the minimum work space.

In addition, the deep learning-based smart automatic logistics warehouse system linked with the user terminal identifies the optimal movement line based on deep learning, and identifies the overall flow of logistics rather than sequential and pre-programmed logic to determine the loading location. Efficiency can be maximized by specifying

In addition, since it is possible to adjust the movement speed of items according to the load, it is possible to realize the minimization of movement time, and to minimize unnecessary movement lines by selecting the optimal location for each period and frequency of use.

In the present invention, the deep learning algorithm is a genetic algorithm (Genetic Algorithm, GA) can be applied together.

Figure 4 is a perspective view of a stacker-type warehouse according to an embodiment in the deep learning-based smart automatic logistics warehouse system linked with the user terminal of Figure 2, Figure 5 is a perspective view showing a stacker of the warehouse, Figure 6 is a warehouse It is a perspective view showing the stacker fork of, and FIG. 7 is a perspective view showing the operation of the stacker fork of the loading warehouse.

As shown in Figures 4 to 7, the loading warehouse according to the present invention has a drawer 120 provided at one end of the lower end to take out or insert goods, and the pallet 10 is inserted and stored on the inside front and rear surfaces. The storage unit 100 in which the mounting rails 110 are formed at regular intervals so as to be located in the center of the storage unit 100 and transferred to the vertical and horizontal directions of the storage unit 100, the storage unit 100 The stacker 200 for moving the pallet 10 stored in the drawer 120 or storing the pallet 10 in the storage unit 100, and the goods provided in the drawer 120 of the storage unit 100 to be stored A photographing unit 300 for photographing and measuring the size, a detector 400 provided in the drawer 120 of the storage unit 100 to measure the weight of items to be stored, a stacker 200, and a photographing unit 300 , Edge control unit that controls the operation of the sensing unit 400 and selects a position where an item is stored based on applying a deep learning algorithm based on the information measured by the photographing unit 300 and the sensing unit 400 ( 600). Here, the deep learning algorithm can apply a genetic algorithm (Genetic Algorithm, GA).

It is formed on both sides of the drawer 120 of the storage unit 100 and further includes an alignment unit 500 for aligning the loads mounted on the pallet 10 by pressing the outer surface to be located in the center of the pallet 10. to be characterized

The storage unit 100 is formed of a rectangular frame with a hollow inside, and at one end of the lower end is a drawer made of a conveyor belt for taking out or receiving pallets 10 stacked with loads formed inside the storage unit 100. (120) is formed.

At this time, a plurality of mounting rails 110 are formed on the inner front and rear surfaces of the storage unit 100 so that the pallets 10 can be accommodated and stored, respectively, and a plurality of pallets 10 inside the storage unit 100 are stored, so that the loads to be stored are exposed to the drawer 120 one by one by the stacker 200, and then stored on top of the pallet 10, and then stored into the storage unit 100 by the stacker 200.

Since the mounting rails 110 are formed to support the lower surfaces of both sides of the pallet 10, the space in which the loads are accommodated can be adjusted by changing the insertion position of the pallet 10 according to the size of the loads.

The stacker 200 is located in the center of the storage unit 100 and is formed to be movable in a horizontal or vertical direction, and is slid toward a mounting rail 110 formed on the front or rear surface of the storage unit 100 to palletize. (10) is used to withdraw or insert.

To this end, the stacker 200 is formed inside the storage unit 100 and formed in the first transfer unit 210 formed to be vertically elevated by the edge control unit 600, and formed on the first transfer unit 210 to store the ( 100) is formed on the second transfer unit 220 that moves horizontally toward one end or the other end, and the first transfer unit 210 slides toward the front or back of the storage unit 100 and then grips the pallet 10 It consists of a fork 230 positioned above the first transfer unit 210.

At this time, it is preferable that the first transfer unit 210 and the second transfer unit 220 determine the lifting position by numerical control so that the pallet 10 stored in the set position can be withdrawn or inserted, and the fork 230 is formed so that the length can be stretched so that it can be inserted into the groove formed in the lower front surface of the pallet 10.

The first transfer unit 210 is connected to the take-up motor 212 through a wire 211, and the take-up motor 212 rotates the take-up reel 213 connected to the wire 212 to wind the wire 212 in the winding direction. It is formed so that it can be raised and lowered according to, and the second transfer unit 220 is formed so that it can be transferred in a horizontal direction along the rail by the transfer motor 234.

The fork 230 has a base stage 231 fixed to the top of the first transfer unit 210, and a first stage 232 and a second stage 233 are sequentially formed on the top of the base stage 231. are combined

The first stage 232 is formed so that only one end and the other end of the base stage 231 can be constrained and moved, and the second stage 233 protrudes toward one end or the other end of the first stage 232 to palletize. (10) It can be inserted inside.

When the second stage 233 is inserted into the pallet 10, the first transfer unit 210 is raised to separate the pallet 10 from the mounting rail 110, and the second stage 233 and the first stage 232 As is positioned at the center of the base stage 231, the pallet 10 can be positioned above the first transfer unit 210.

The photographing unit 300 is for measuring the size of the load stored in the storage unit 100, and a space stored by photographing the load using a camera and measuring the size of the load in three dimensions based on the photographed information. is used to assign

To this end, the photographing unit 300 includes a first camera 310 located above the drawer 120 and measuring the size of the upper surface of the load loaded on the pallet 10, and the stacker 200 formed on the pallet. A second camera 320 for measuring the size of the front of the load loaded on the top of the 10 is photographed, and the first camera 310 and the second camera 320 are formed on the pallet 10 to detect the load. It is characterized in that it is made of a reference marker 330 that is photographed together when photographing and determines the reference size.

Since the first camera 310 of the photographing unit 300 is located above the drawer 120, it can always measure the horizontal and vertical sizes of loads loaded on the top of the pallet 10 located in the drawer 120. The second camera 320 is formed on the side of the drawer 120 and takes pictures in the height direction of the load, thereby measuring the three-dimensional size of the load.

At this time, a plurality of reference markers 330 for setting reference dimensions in the photographed image are formed on the pallet 10, and the first camera 310 and the second camera 320 are photographed together when photographing the load. Based on the size of the marker 330, it is possible to calculate the size of the load.

At this time, it is preferable that the size of the reference marker 330 is formed in units of 1 cm or 10 cm, and this size may vary depending on the size of the loaded object.

The edge controller 600 allocates a location to be stored in the storage unit 100 according to the measured size of the load. At this time, the size of the load is stored in the edge controller 600 and stored when the next load is stored. It becomes possible to allocate space from the area excluding the space of .

In other words, since the exact location can be determined through the dimensions, the space can be utilized to the maximum.

The sensing unit 400 measures the weight of loads loaded on the pallet 10 so that the high weight is located at the lower part of the storage unit 100 and the lower weight is located at the upper part of the storage unit 100 so that the storage unit 100 It is used to minimize damage when the storage unit 100 collapses or when the stacker 200 is driven by positioning the center of gravity of the lower part.

To this end, the detector 400 is formed on the drawer 120 to measure the weight of the pallet 10 mounted on the drawer 120 and to measure the weight of loads loaded on the pallet 10 based on the empty pallet. It is characterized in that it consists of a dog weight sensor (410).

The drawer 120 is composed of a conveyor belt formed so that a plurality of rods are rotated. A weight sensor 410 is formed on each rod so that when the pallet 10 is placed and pressed downward by the weight of the load, the weight sensor 410 measures the weight so that the weight of the load can be measured.

At this time, if the weight exceeds the allowable range to be loaded on the pallet 10 or the allowable range for the stacker 200 to drive, the stacker 200 is stopped so that the stacker 200 is not accommodated by the edge control unit 600 and a warning alarm is generated. can be controlled.

The aligning unit 500 is located on both sides of the drawer 120 to center the position of the load loaded on the upper part of the pallet 10, and on both sides, a support stand 510 that slides and moves toward the center is formed, and a pressure rail 520 for transferring the pressure table 510 is formed inside the drawer 120.

That is, when the position of the load is out of the shooting range or the reference marker 330 is covered when the load is photographed by the photographing unit 300, the alignment unit 500 locates the load at the center of the pallet 10 so that the photograph can be performed. It can be easily aligned.

In addition, the aligning unit 500 can selectively align the pallets 10, and through this, it is possible to control the forks 230 of the stacker 200 to be accurately inserted into the grooves formed in the pallets 10.

For reference, the edge control unit 600 may be configured with a personal computer or business computer installed in the field, and a control application installed in the computer may operate in conjunction with a control application installed in a pre-registered smartphone.

Therefore, the administrator can check the operating status of the system in real time through the control application installed on his/her smartphone.

In addition, the control application can provide optimized information by managing the type of load of each pallet, storage status, accumulated storage time, warehousing time, expected shipping time, etc. in the form of a database.

The control application can display vibration conditions during loading of pallets. That is, the magnitude of vibration according to the moving path of the load is displayed, and if the magnitude of vibration exceeds the preset level, the state of the corresponding section is designated as a significant section, and then automatically controlled to automatically reduce the movement speed during the loading process of other loads. can This determination may be determined by a learned deep learning algorithm.

In addition, a plurality of peripheral cameras may be provided, and photographed information is transmitted to the edge controller.

The peripheral camera captures the surroundings of the system. It not only captures the moving state of the load, but also operates to record the actions of nearby workers and the moving state of the forklift.

The image transmitted to the edge control unit may be transmitted to the manager's portable terminal (smart phone), and a virtual object corresponding to the actual image information of the load displayed on the screen of the portable terminal may be displayed. When a virtual object is touched, a motion of a preset virtual object corresponding to the touched part may be reproduced.

For example, an icon of a (packaged) load is displayed as a virtual object, and when a corresponding part is touched, a 3D modeling image of the load may be rotated and displayed.

As such, when a virtual object corresponding to real image information is displayed on the screen of the portable terminal, the coordinates of the selected virtual object are changed from the spatial coordinate system for the real image information to the mobile coordinate system of the portable terminal, and the coordinates of the deselected virtual object Coordinates can be changed from the mobile coordinate system of the portable terminal to the spatial coordinate system for real image information.

Meanwhile, the plurality of peripheral cameras include a visible image capture unit, a thermal image capture unit, and a radar sensor unit, and each camera adjusts rotation (PAN), directional tilt (TILT), and zoom (ZOOM) using an internal motor. Preferably, the visible image capturing unit, the thermal image capturing unit, and the radar sensor unit are all configured to detect the same capturing area.

If the visible image capture unit, the thermal image capture unit, and the radar sensor unit detect different shooting areas, the edge control unit automatically captures only the common area among the areas detected by the visible image capture unit, the thermal image capture unit, and the radar sensor unit. Set it as an area and proceed with the operation of processing the image.

For reference, the edge control unit may automatically estimate a fire in an object moving around the system, for example, a forklift, based on photographing information of a plurality of peripheral cameras. At this time, the main control unit may warn of the risk of fire based on the temperature of the engine room of the forklift as well as the emission amount and emission accumulation time of the exhaust gas. At this time, when the edge control unit detects a fire risk of the forklift through a deep learning algorithm, it controls the load loaded around the forklift to be moved to the farthest position from the forklift.

At this time, since the edge control unit collects information on the type of load, it can move flammable and inflammable materials to a distance with priority in consideration of the fire risk of the load. In addition, the edge control unit may control the fire propagation to be delayed by moving the load containing the fire extinguishing liquid disposed at a predetermined location closest to the fire risk area.

In addition, when estimating the speed of the forklift, the edge control unit additionally considers the pixel change amount of the visible image and the measured value of the radar sensor unit as well as the amount of exhaust gas emission and temperature change of the exhaust gas detected by the thermal imaging unit to determine the change in speed of the forklift can also be estimated.

As described above, according to the loading warehouse according to the present invention, the storage space can be efficiently used by automatically measuring the size of the loads loaded on the pallet, and the weight of the loads is measured so that the heavy loads are located at the bottom. By doing so, it is possible to reduce the distortion or load of the equipment due to the load, and there is an effect of automatically aligning the center of the pallet so that the position of the load mounted on the upper part of the pallet is not distorted.

In addition, the deep learning-based smart automatic logistics warehouse system linked with the user terminal according to an embodiment of the present invention automatically estimates the fire risk of objects moving around the system based on the shooting information of a plurality of surrounding cameras and calculates the fire risk. It is possible to automatically adjust the storage position of the item by taking into account

In addition, the system applies a deep learning algorithm to perform work in the shortest time during loading and order picking. That is, based on the information received from the edge controller, a deep learning algorithm is applied to automatically select a storage location.

The deep learning algorithm optimizes the loading site through analysis and learning of loading data, and autonomous driving learning is performed by predicting the loading quantity, load, and required amount.

In other words, it is possible to predict the amount of warehousing/shipping by time and shorten the movement time, and to provide work convenience and solve the warehousing and warehousing bottleneck with the minimum work space.

In addition, the deep learning-based smart automatic logistics warehouse system linked with the user terminal identifies the optimal movement line based on deep learning, and identifies the overall flow of logistics rather than sequential and pre-programmed logic to determine the loading location. Efficiency can be maximized by specifying

In addition, since the speed can be adjusted according to the load, the travel time can be minimized, and unnecessary movement is minimized by selecting the optimal location for each period and frequency of use.

As such, those skilled in the art to which the present invention pertains will be able to understand that the present invention may be embodied in other specific forms without changing its technical spirit or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. The scope of the present invention is indicated by the following claims rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention. do.

10 : Palette
100: storage unit
110: rail
120: cashier
200: Stacker
210: first transfer unit
211: wire
212: winding motor
213: winding reel
220: second transfer unit
221: transfer motor
230: fork
231: base stage
232: first stage
233: second stage
234: drive motor
300: shooting unit
310: first camera
320: second camera
330: reference marker
400: sensing unit
410: weight sensor
500: alignment unit
510: Pressure stand
520: pressurized rail
600: edge control

Claims (4)

A drawer is provided at one end of the lower part to take out or insert goods, and a storage unit with rails formed at regular intervals so that pallets can be inserted and stored on the inside front and back sides;
a stacker located at the inner center of the storage unit, transported in vertical and horizontal directions of the storage unit, and moving the pallet to the storage unit or receiving the pallet;
a photographing unit provided in the drawer of the storage unit and measuring the size of the stored goods by photographing them;
a sensing unit provided in the drawer of the storage unit to measure the weight of stored items;
an edge control unit that controls operations of the stacker, the photographing unit, and the sensing unit, and automatically selects a location where the product is accommodated by applying a deep learning algorithm based on information measured by the photographing unit and the sensing unit; and
A user terminal that interactively processes necessary information in the form of a chatbot and controls the operation of the assigned edge control unit;
Deep learning-based smart automatic logistics warehouse system that works with user terminals including
According to claim 1,
The deep learning algorithm is a deep learning-based smart automated logistics warehouse system that works with a user terminal, characterized in that by applying a genetic algorithm (Genetic Algorithm, GA).
According to claim 1,
An alignment unit formed on both sides of the drawer of the storage unit and aligning the loads mounted on the pallet by pressing the outer surface to be located in the center of the pallet; warehouse system.
According to claim 1,
the sensor,
A plurality of weight sensors formed in the drawer and measuring the weight of the pallet placed in the drawer and measuring the weight of the load loaded on the pallet; Logistics warehouse system.

Images