CN111626681B

CN111626681B - Image recognition system for inventory management

Info

Publication number: CN111626681B
Application number: CN202010492368.XA
Authority: CN
Inventors: 贺桂和
Original assignee: Hunan University of Humanities Science and Technology
Current assignee: Hunan University of Humanities Science and Technology
Priority date: 2018-02-01
Filing date: 2018-08-23
Publication date: 2023-09-12
Anticipated expiration: 2038-08-23
Also published as: CN109089082A; CN114548882A; CN109089082B; CN109214751A; CN109146380B; CN109146380A; CN111950967A; CN109214751B; CN108846621A; CN114819797A; CN108846621B; CN111626681A

Abstract

The invention relates to an image recognition system for inventory management, comprising an image acquisition unit and an image recognition unit, wherein the image acquisition unit acquires a first image and acquires a second image and a third image of a hand of at least one access mechanism, the image recognition unit determines the absolute position of the at least one access mechanism in the area of a goods shelf and the relative position of the access mechanism relative to at least one goods shelf carrying at least one article based on the first image, determines the state of the corresponding access mechanism holding the article based on the comparison of the second image and the third image, and responds to the state of the access mechanism holding the article to recognize the change of at least one inventory position of the goods shelf in the relative position around the absolute position of the access mechanism and judges the corresponding relation between the article and the change of the inventory position.

Description

Image recognition system for inventory management

The invention relates to a filing number 201810967152.7, filing date 2018, 8-month and 23-day filing type invention, and a filing name of the filing of a separate filing of an inventory management system based on a strategy module.

Technical Field

The invention relates to the technical field of inventory management, in particular to an inventory management system based on a policy module.

Background

Inventory management is the management and control of various articles, products and other resources throughout the manufacturing or service industry production, and operations, keeping their reserves at an economically reasonable level. In inventory, the flow rate, volume and period of the goods are greatly different. Thus, different goods should take inventory management modes with their corresponding features, which facilitates resource sharing, saves inventory management costs, and inventory optimization. According to the present research, the existing inventory management has a plurality of technical problems of resource waste, high management cost and the like.

CN107833363a discloses an image recognition-based vending machine automatic inventory management system and method. The system comprises a self-service vending machine provided with an automatic lifting platform, wherein the automatic lifting platform is provided with a camera device which can shoot goods in a goods supplementing position or a goods in a harvesting position to the goods in the outermost goods position. Each vending position self-service vending machine, each camera device and each automatic lifting platform are provided with a local control system, and each local control system can be connected with a cloud service system through the Internet; when the image pick-up device captures an image of the goods supplementing position of the self-service vending machine or the goods at the outermost end of the vending machine, the image is identified and then the goods on sale database of the self-service vending machine is updated, namely the latest inventory of the self-service vending machine is obtained.

CN106295165a discloses a system and method for preventing drug delivery errors based on image recognition. The system comprises an image acquisition unit, a processing unit and a processing unit, wherein the image acquisition unit is used for acquiring instant image information of medicines; the database unit is used for storing medicine image information and comprises an instant image database and a pre-stored image data storage unit; the image recognition and comparison unit is used for recognizing and comparing the instant image information with the pre-stored image information in the pre-stored database; the HIS interface is used for being connected with a hospital HIS system, acquiring medicine prescription information, and the alarm and confirmation unit is used for sending an alarm signal or a verification passing signal according to the comparison result of the instant image and the pre-stored image; the control unit is connected with the image recognition comparison unit, the HIS interface and the alarm and confirmation unit. The system can effectively reduce the medicine dispensing error rate, is objective and accurate, and does not influence the medicine dispensing work efficiency.

Chinese patent document CN105531715a discloses detecting item interactions and movements, wherein updating of inventory and shopping list are determined by comparing images before and after taking an item. The problem with this approach is that once the user has removed the item, but returned to an improper location, the shopping list and inventory updates are inaccurate. In addition, when users repeatedly compare in the areas with similar types and similar placement positions, challenges are brought to photographing and calculation, and if users take articles with both hands and put back one article to the wrong position, errors often occur in shopping list and inventory update. It omits the settlement link, but other problems are caused thereby. Therefore, there is room for improvement in such automatic shopping.

CN107077659a provides an intelligent inventory management system, server, method, terminal and program product for reducing complexity of inventory management; it comprises the following steps: the intelligent machine terminal shoots the object and sends the shot image to the intelligent inventory management server; the intelligent inventory management server identifies and determines articles and stock in the image according to the image shot by the intelligent machine terminal; and warehousing the articles and stock information identified by the object identification module. The object identification technology is utilized, so that the workload of manual input is reduced, the accuracy of article information storage is improved, and the complexity of inventory management is reduced; inventory changes cannot be accurately assessed in real time. Unless there is manual intervention in the statistics and inventory, the inventory remains confusing.

CN104361039a provides an auxiliary method of embedded intelligent refrigerator based on Qt development, which comprises the following steps: step 1, displaying overdue reminding preview and message information; step 2, food stock management, based on data and user operation, checking all food detailed information in the current stock, and performing unified management of editing and deleting the food information; sorting and managing all food materials according to the production date, the residual quality guarantee period and the expiration date so as to check the information of each food material; step 3, expiration reminding, namely counting the remaining shelf life of the food material information in stock, and giving visual information preview to the food material with the remaining shelf life of less than 5 days, wherein the expiration date and the remaining shelf life of the food material are specifically displayed and updated in real time; step 4, inquiring recipes and recommending recipes, providing massive recipe data, providing a manufacturing method of various delicious dishes, and searching according to the existing food materials or custom food materials to obtain specific recipes; recommending recipes with different tastes and effects according to food materials or current season and health nutrition information; step 5, a shopping list is added with existing food materials or custom food material information to the shopping list, statistics and management are carried out on the food materials in the shopping list, a shortage reminding list is provided on the basis of data mining, and shopping selection is provided; and 6, identifying and accessing food materials, and identifying and storing the food materials. The technical scheme cannot count inventory consumption in real time, and cannot effectively manage inventory in a kitchen and a kitchen warehouse where food is frequently taken.

In addition, in order to accurately manage kitchen inventory, large hotels and restaurants typically have dedicated personnel to manage inventory (i.e., inventory administrators), who only provide relevant food items to the relevant personnel based on the order (the ingestion information) when the restaurant provides the order (equivalent to the ingestion information described below in the present invention). Thereafter, correspondence between the order and the food item is manually checked by the inventory manager. Not only is this approach inefficient, but it is often subject to substantial loss due to the extremely bulky variety of food products.

The Chinese patent document with publication number CN104637198A discloses an intelligent settlement shopping cart system with an autonomous tracking function, which comprises an MCU microcontroller, a controllable direct current motor, a pressure weighing sensor, an RFID reader, a camera, a wireless transceiver, a touch display screen, a bar code scanner and a silver-linked card reader, and is characterized in that: the MCU microcontroller 1 controls the driving force of the controllable direct current motor through the weight of the vehicle body sensed by the pressure weighing sensor, and moves the shopping cart at a proper speed; the bar code scanner automatically scans the articles, the touch display screen displays the article information and the total price of the currently purchased articles in real time, the wireless transceiver realizes the real-time interaction of the MCU microcontroller and the supermarket background server about the article information, and the consumer automatically swipes the card through the Unionpay card reader to settle accounts. In the system, an RFID radio frequency card is used as a signal source for shopping cart tracking, and establishes point-to-point single-line connection with a shopping cart; shooting the surrounding environment of the shopping cart by the camera at any time, automatically avoiding obstacles through a path recognition algorithm, and selecting a tracking path; the driving wheel driven by the driving motor can deflect left and right, and the steering of the driving wheel is controlled by the MCU microcontroller according to the path information provided by the path recognition algorithm.

The Chinese patent literature with the bulletin number of CN202904765U provides a self-help intelligent shopping cart for a mall, which comprises a frame, wheels arranged below the frame, a basket arranged above the frame, a push rod, and an included angle between the lower end of the push rod and the outer side surface of the bottom of the basket is 40-60 degrees, a touch display screen is arranged between the upper end of the basket and the push rod, a scanner is arranged on the left side of the touch display screen, a card swiping groove is arranged on the right side of the touch display screen, and an automatic opening and closing sensing door is arranged on the upper end of the basket. The automatic opening and closing sensing door is arranged at the upper end of the basket, and the automatic opening and closing sensing door is provided with the sensing instrument which can sense the placed articles, so that trouble of buying or buying more by mistake in the shopping process is avoided.

By studying the above technical solutions, those skilled in the art cannot obtain a satisfactory inventory management solution for how to realize resource sharing and save inventory management costs even if combining them.

Disclosure of Invention

In view of the foregoing, the present invention is directed to providing an inventory management system based on image recognition. Because most warehouse supermarkets are equipped with video monitoring management systems at present, the accuracy of the monitoring systems is better and better, and the image accuracy of millimeter level can be achieved, and the low-cost image recognition solution is still developing at a high speed.

In view of this, the present invention provides an inventory management system based on image recognition, comprising: the image acquisition unit is used for acquiring dynamic images and/or static images of the memory access mechanism in the stock area; an image recognition unit that determines a status of an article held by the access mechanism based on the dynamic and/or static image comparison; a computing device that generates inventory information based on the characteristic parameters; a policy unit that generates an inventory management policy based on the inventory information; the image recognition unit is used for responding to the state that the goods are held by the access mechanism and recognizing the change of at least one inventory position of a goods shelf around the absolute position of the access mechanism and/or at a preset relative position, and the computing device is used for judging the corresponding relation between the goods and the change of the inventory position and generating the inventory information; the computing equipment records the inventory information of each goods in the historical information of each goods in a certain period, and obtains the inventory state information of each goods after analyzing based on the historical information and the inventory information, and the strategy unit triggers and adjusts the acquisition scheme of the image acquisition unit and/or adjusts the identification scheme of the image identification unit based on the inventory state information.

According to a preferred embodiment, the policy unit is able to categorize the change of the relative and/or absolute position to the stock position with the access mechanism and to associate the access mechanism with a shelf interacting with the access mechanism, such that the image acquisition scheme of the image acquisition unit to the access mechanism and/or the identification scheme of the image identification unit is dynamically adjusted at any time and any place depending on the change of the goods and the stock position.

According to a preferred embodiment, the system further comprises an RFID unit for marking and/or scanning the item; the computing device is capable of storing image attributes of an RFID-tagged item in association with underlying information of the RFID-tagged item in a storage unit; the image acquisition unit can record image information of the goods before the RFID scans the goods; the computing device is capable of performing information matching with image attributes of the RFID-tagged items based on the characteristic parameters of the image information identified by the image identification unit; when the matching is successful, the storage unit automatically registers the related information of the goods and/or the RFID automatically scans the label of the goods; and when the matching fails, the alarm unit prompts the user to register the relevant information of the goods.

According to a preferred embodiment, the image acquisition unit comprises first image acquisition means for acquiring a first image and determining, by the image recognition unit, based on the first image, the absolute position of at least one access mechanism within the area of the pallet and the relative position of the access mechanism with respect to at least one pallet carrying at least one item; a second image acquisition device for detecting a second image of the hand of the at least one access mechanism, the second image being taken before the hand of the access mechanism enters at least one inventory location of the shelf; a third image acquisition device for detecting a third image of the hand of the at least one access mechanism, the third image being taken after the hand of the access mechanism has been moved out of the at least one inventory location of the shelf; the image recognition unit determines a state of holding an article by a corresponding access mechanism based on a comparison of the second image and the third image, recognizes a change in at least one stock position of a shelf at a predetermined relative position around the absolute position of the access mechanism in response to the state of holding an article by the access mechanism, and thereby judges a correspondence between the article and the change in the stock position.

The absolute position is a position in a coordinate system established by a pointer to a supermarket or a warehouse of the inventory management system, and the origin of the coordinate can be, for example, a corner of the occupation area of the supermarket or the warehouse, such as southeast corner, or the position of a settlement counter. The relative position refers to the relative position between shelves in the supermarket, and may also be the relative position between the access mechanism or person and the shelves.

It is contemplated that personnel or access mechanisms will in most cases only be able to operate shelves in close proximity (typically shelves in the range of about 2 meters) to which they can access. In the invention, the relative position and the absolute position are adopted to carry out classification analysis on the stock position change and the personnel/access mechanism, so that the supermarket dynamically divides the area needing analysis and calculation according to the change generation place at any time and any place, the goods shelf and the personnel or the access mechanism which are interacted with the goods shelf at present can be accurately related to each other, and the difficulty in shopping or stock arrangement of a large number of dynamic personnel and the dynamic access mechanism is solved. In other words, the invention firstly corresponds the stock change to personnel or an access mechanism according to the position, which has decisive effect on the real-time performance of the stock management of the supermarket. For supermarkets with thousands of inventory positions, the inventory change is firstly related with the absolute position and the relative position, so that the operation amount of image recognition is greatly reduced, and the real-time inventory analysis can be completed without adding a large amount of computing equipment. Chaotic ordering of inventory data is thereby largely avoided.

In this way, not only can the increase or decrease of the items at the inventory location be analyzed, but also whether the corresponding item is placed back in the correct inventory location on the shelf can be analyzed. Here, the access mechanism may be a person or a machine. For a person, the object of image recognition is the person's arm; for a machine device, the object of image recognition is a robot arm or the like; therefore, the invention can flexibly cope with the supermarket environment where the robot and the robot coexist.

According to a preferred embodiment, the method further comprises: at least one fourth image acquisition device is used, which is arranged on the goods temporary storage unit and is used for detecting the action of the goods entering the goods temporary storage unit, wherein the corresponding relation between the goods entering the goods temporary storage unit and the change of the stock position is confirmed in response to the action of the goods entering the goods temporary storage unit. By the method, after the goods are put into the goods temporary storage unit, the corresponding relation between the goods entering the goods temporary storage unit and the change of the stock position can be confirmed, so that the statistics of the change of the stock position is more accurate, and the practicability of the system is improved. The goods temporary storage unit is a shopping frame for personnel and a goods storage frame for machines.

According to a preferred embodiment, the fourth image acquisition device is further configured for: a fourth image and a fifth image of the hand are taken in response to an action of the hand of the access mechanism entering the item escrow unit, wherein the fourth image is taken before the hand of the access mechanism enters the item escrow unit, and the fifth image is taken after the hand of the access mechanism has been removed from the item escrow unit, wherein a status of removal of the item by the respective access mechanism is determined based on a comparison of the fourth image and the fifth image, a change in at least one inventory position of the pallet at a predetermined relative position around an absolute position of the access mechanism is identified in response to the status of the item held by the access mechanism, and a correspondence between removal of the item and the change in the inventory position is determined therefrom. The removal of the article here corresponds to the removal of the respective article which has been put into the article holding unit from the article holding unit, i.e. the action of the article extraction is altered when the access mechanism has received an instruction from the outside to withdraw an article; alternatively, the access mechanism alters the action of the goods extraction in response to rejection of the erroneous goods extraction by its own optical or RFID identification mechanism; of course, in the case of personnel, it may be the person's act of returning the goods after evaluating the goods. In this way, the invention can analyze the change of the stock position possibly caused by the process again when the goods are removed from the goods temporary storage unit, so that the stock position can be updated more comprehensively and accurately.

According to a preferred embodiment, the method further comprises: at least one fifth image acquisition device is used, which is arranged in the access way of the access mechanism, for acquiring identification marks for payment and/or registration. Through the mode, the fifth image acquisition device is arranged in the access mechanism access passage to acquire the identification for payment and/or registration, the price can be settled rapidly, the goods turnover efficiency, the settlement efficiency and the user experience are greatly improved, and the method has obvious advantages especially under the conditions of large passenger flow and large inventory quantity. The image recognition mode is adopted, so that the existing supermarket is not required to change the infrastructure on a large scale when coping with the coming situation of the intelligent age, and the machine and the person are regarded as clients in a certain sense at the same time, and the goods temporary storage unit is uniformly borrowed to finish the stock arrangement and settlement.

According to a preferred embodiment, the fifth image acquisition device is activated only in response to an event that the fourth image acquisition device provides an image. In this way, the fifth image acquisition device can be started up only when needed, and can be in a dormant or standby state when idle, so that energy is saved.

According to a preferred embodiment, the fifth image capturing device adjusts its optical capturing direction and focal length in response to an event that the fourth image capturing device provides an image and in dependence on the position of the cargo holding unit provided with the fourth image capturing device relative to the access way. In this way, the fifth image acquisition device can clearly and conveniently acquire the identification mark for payment and/or registration in a lens following and focal length adjusting mode according to the position of the goods temporary storage unit relative to the access passage, so that the range interval and flexibility of price settlement are increased, and the user experience is improved.

According to a preferred embodiment, the first image is captured by the first image capturing device after the second image capturing device captures a second image of the hand of the access mechanism, wherein the second image is captured before the hand of the access mechanism enters at least one inventory location of the shelf. In this way, the supermarket can deal with the scene of multi-person and/or multi-equipment simultaneous store purchase at the same time, and the system can start a series of responsible image analysis operations only when the access mechanism with the first image performs the object access or replacement again. In other words, the tracking of the access mechanism by the system begins when it attempts to reach the items on the shelves. Therefore, the calculation cost can be reduced, the storage cost can be reduced, the construction cost of the system can be reduced, and the use efficiency of each component in the system and the smoothness of the system can be improved.

According to a preferred embodiment, the item registering unit sets an introduction timing period in response to the state of the item held by the access means, starting from an initial point in time when the absolute position is determined based on the first image, and after the certain introduction timing period has elapsed, if the fourth image capturing device fails to capture an action of the item entering the item registering unit, a prompt concerning the property of the item is issued to the access means holding the item by the item registering unit, wherein the prompt may be in the form of an electronic signal, a voice, an image or a video. Since the characters on the package of the present articles are very small and dense, even if the access mechanism with the high-definition camera cannot quickly find the required information in a short time, the access mechanism is more difficult for people with vision impairment, myopia or hyperopia. After the introduction timing period, the corresponding access mechanism does not put the article into the cargo temporary storage unit, and the access mechanism does not find the required information. In this way, the invention can give a prompt to the person or the access mechanism holding the article in an auxiliary way when the corresponding person or the access mechanism does not decide whether to purchase the article after a certain time, so that the person or the access mechanism makes a decision according to the prompt.

The invention provides an inventory management system and method in warehouse style supermarkets aiming at coexistence of personnel and intelligent carrying equipment, which has at least the following advantages:

(1) The invention dynamically adjusts the image acquisition unit and the image recognition unit according to the inventory information so as to achieve the resource sharing of equipment, avoid arranging more devices and save the cost.

(2) The invention can analyze not only the increase and decrease of the articles at the stock position, but also whether the corresponding articles are put back to the correct stock position on the goods shelf;

(3) According to the invention, after the goods are put into the goods temporary storage unit, the corresponding relation between the goods entering the goods temporary storage unit and the change of the stock position can be confirmed, so that the statistics of the change of the stock position is more accurate, and the practicability of the system is improved;

(4) The invention can analyze the change of the stock position possibly caused by the process again when the goods are removed from the goods temporary storage unit, so that the stock position can be updated more comprehensively and accurately.

Drawings

FIG. 1 is a schematic block diagram of a system of the present invention; and

fig. 2 is a schematic block diagram of a preferred embodiment of the present invention.

List of reference numerals

1: image acquisition unit 2: image recognition unit

5: policy unit 6: memory cell

7: alarm unit 11: first image acquisition device

12: the second image acquisition device 13: third image acquisition device

14: fourth image acquisition means 15: fifth image acquisition device

16: sixth image acquisition device 17: seventh image acquisition device

18: eighth image acquisition device 21: indicating mark

22: position error flag 30: computing device

40: goods temporary storage unit

Detailed Description

The following detailed description refers to the accompanying drawings.

To facilitate understanding, identical reference numerals have been used, where possible, to designate similar elements that are common to the figures.

Example 1

The invention discloses an inventory management system based on image recognition, which can be used as a supplement to other embodiments without causing conflict or contradiction.

As shown in fig. 1, the system includes: an image acquisition unit 1, an image recognition unit 2, a computing device 30 and a policy unit 5. The image acquisition unit 1 is used for acquiring dynamic images and/or static images of an access mechanism in an inventory area. The image recognition unit 2 determines the status of the access mechanism holding the item based on dynamic and/or static image comparison. The computing device 30 generates inventory information based on the characteristic parameters. The policy unit 5 generates an inventory management policy based on the inventory information. The image recognition unit 2 recognizes a change in at least one stock position of the shelves around the absolute position of the access mechanism and/or at a predetermined relative position in response to the state of the access mechanism holding the goods. The computing device 30 determines a correspondence between items and changes in inventory locations, generating inventory information. The computing device 30 records inventory information for each item in the history information for each item over a period and derives inventory status information for each item based on analysis of the history information and the inventory information. The policy unit 5 triggers an adjustment of the acquisition scheme of the image acquisition unit 1 and/or an adjustment of the identification scheme of the image identification unit 2 based on the stock state information. For example, the expected focusing position of the image acquisition unit, the monitoring time length and the monitoring quantity of the image acquisition unit for an article, and the identification times and the identification accuracy of the image identification unit are determined according to the inventory state information. The invention dynamically adjusts the image acquisition unit and the image recognition unit according to the inventory information so as to achieve the resource sharing of equipment, avoid arranging more devices and save the cost.

Preferably, the policy unit 5 is capable of classifying and analyzing the change of the relative position and/or the absolute position to the inventory position with the access mechanism and associating the access mechanism with the shelf interacting with the access mechanism, so as to dynamically adjust the image acquisition scheme of the image acquisition unit 1 to the access mechanism and/or the identification scheme of the image identification unit 2 according to the change of the inventory position and the goods at any time and any place. Preferably, the policy unit 5 comprises a pre-module, a processing module and a selection module. The pre-module stores policy conditions, for example, when the flow rate of commodity A in 1 hour is 100, the image acquisition unit is updated from 2 image acquisition devices to 4 image acquisition devices. The selection module accesses at least one policy scheme, for example, the image acquisition unit will update from 2 image acquisition devices to 4 image acquisition devices. For example, the computing device 30 analyzes the flow state of the commodity a according to the inventory state information and guides the flow state to the front-end module, and the processing module analyzes and processes the flow state of the commodity a, and feeds back to the selection module to select the policy scheme.

The system further comprises an RFID unit 6 for marking and/or scanning items; the computing device 30 is capable of storing in the storage unit 6 the image attribute of the RFID-tagged item in association with the underlying information of the RFID-tagged item. The image acquisition unit 1 is capable of recording image information of an article before the RFID scans the article. The computing device 30 is capable of information matching with image attributes of the RFID-tagged item based on characteristic parameters of the image information identified by the image identification unit. When the matching is successful, the storage unit automatically registers relevant information of the goods and/or the RFID automatically scans the labels of the goods. When the match fails, the alarm unit 7 prompts the user to register the relevant information of the article.

According to a preferred embodiment, an image acquisition and recognition unit includes an Support Vector Machine (SVM) operation unit for performing SVM operation on an input image, and a data storage unit for temporarily storing data generated during image recognition processing, the SVM operation unit including: a feature value calculation unit that calculates a feature value indicating a degree to which an identification object, which is an object photographed in an inputted image, is similar to a comparison object, which is an object to be identified; and an accumulation and addition unit that performs accumulation and addition of feature values corresponding to the teaching data classified as the same type of comparison object in the teaching data group, wherein in the SVM calculation process, the feature value calculation unit calculates all feature values corresponding to the teaching data and stores the feature values in the data storage unit, and the accumulation and addition unit performs accumulation and addition of feature values of the same type of comparison object and outputs the result as a result of recognition of the recognition object in the image recognition process.

According to a preferred embodiment, referring to fig. 2, the system may comprise: a first image acquisition means 11 for acquiring a first image and determining, by the image recognition unit 2, based on said first image, the absolute position of at least one access mechanism within the area of the pallet and the relative position of the access mechanism with respect to at least one pallet carrying at least one item. Preferably, the ground may be provided with corresponding reference points for identification and/or a reference line assist system to determine absolute and relative positions.

According to a preferred embodiment, the system may further comprise: a second image acquisition device 12 for detecting a second image of the hand of the at least one access mechanism, the second image being taken before the hand of the access mechanism enters at least one inventory location of the shelf. Preferably, for a shelf, it may have at least one inventory location. One shelf may be divided into multiple layers. Each tier may be an inventory location. Alternatively, a floor may be divided into a plurality of areas in the horizontal direction, each of which may be an inventory location. It should be noted that the separation lines between these areas may be image-identifiable separation means and/or non-image-identifiable virtual separation lines. The separating means are, for example, dividing walls and/or reference lines for spraying. The virtual dividing line not capable of identifying the image may be that after the shelf, the second image capturing device 12 and/or the third image capturing device 13 corresponding to the shelf are fixed, the corresponding articles are placed, and then the system captures the image information of the shelf to divide the shelf, so as to obtain at least one inventory position containing at least one type of articles.

According to a preferred embodiment, the system may further comprise: a third image acquisition device 13 for detecting a third image of the hand of at least one access mechanism, said third image being taken after the hand of the access mechanism has been moved out of at least one stock position of the shelf.

According to a preferred embodiment, the image recognition unit 2 determines the status of the respective access mechanism holding the item based on the comparison of the second image and the third image, recognizes a change in at least one stock position of the shelf at a predetermined relative position around the absolute position of the access mechanism in response to the status of the access mechanism holding the item, and thereby judges the correspondence between the item and the change in the stock position. And then, classifying and analyzing the change of the inventory position and the personnel/access mechanism by adopting the relative position and the absolute position, and accurately associating the goods shelf and the personnel or the access mechanism which are interacted with the goods shelf with each other, so that the areas needing to be analyzed and calculated can be dynamically divided according to the change of the goods and the inventory position at any time and any place. Because the stock change is firstly related with the absolute position and the relative position, the operation amount of image identification is greatly reduced, and thus, the real-time stock analysis can be completed without adding a large amount of computing equipment.

For example, there are a plurality of shelves around the absolute position, but the actual stock position that changes cannot be accurately determined by the absolute position, while the relative position may relate to the distance between the two, and only when the distance between the corresponding access mechanism and one shelf is less than or equal to the preset distance, the situation that the access mechanism has an article transfer relationship with the shelf in the relative position may occur. In the case where the access mechanism is located close to the corresponding inventory location where the item is stored, when the access mechanism returns the item to the corresponding inventory location, the change in inventory location includes only an increase in quantity and no increase in items. In the case where the access mechanism is remote from the corresponding inventory location where the item is stored, the access mechanism may not return the item to the corresponding inventory location for the item due to its own operational error. Thus, in the event that an item is placed back on a shelf and not back to its corresponding inventory location, then a change in at least one inventory location of the shelf in a predetermined relative position may include: the inventory location has an increased number of items and an increased number of items stored. Preferably, the status of holding the item may include the type and/or number of items. Preferably, before determining the correspondence between the item and the change in inventory location, the identity of the item removed from and/or placed into the inventory location of the shelf may be identified based on data received from various input devices. For example, a series of images of the hand of the respective access mechanism before the hand passes through the plane into the stock position and also after the hand leaves the stock position can be acquired by the image acquisition device. Based on the comparison of the images, it may be determined whether the respective access mechanism removes or places the object from or into the inventory location. For example, based on the analysis of the images, it may be determined that the respective access mechanism is not holding the object when the hand enters the inventory location, but is holding the object when the hand is removed from the inventory location (the respective access mechanism removes the object from the inventory location). Likewise, image analysis may be used to identify whether the respective access mechanisms hold objects on their hands before the hands enter the inventory location, but do not hold objects when the hands are removed from the inventory location (the respective access mechanisms place the objects in the inventory location). That is, the second image and the third image for comparison of the present invention are not only the case including a single image, but may be a series of the second image and a series of the third image for comprehensive comparison.

Preferably, the management system of the present invention further discloses a fourth image acquisition device 14, which is disposed at the cargo scratch unit 40, for detecting an action of an article entering the cargo scratch unit 40, wherein the correspondence between the article entering the cargo scratch unit 40 and the change of the stock position is confirmed in response to the action of the article entering the cargo scratch unit 40. Preferably, if and only after the step of validating, an item identifier representing the item is added to a bill of lading associated with at least one access mechanism. By adding the item identifier to the bill of lading, the bill of lading is checked in real time according to the items placed in the item temporary storage unit 40 during the process of shopping in the supermarket, and finally, the fee settlement can be directly performed according to the bill of lading, thereby avoiding the problem of long time consumption caused by sequentially removing the items from the scanning bar code for settlement. Preferably, the cargo temporary storage unit 40 is, for example, at least one of a storage frame, a shopping cart, a shopping basket, and a shopping bag dedicated to an access mechanism.

Preferably, the fourth image acquisition device 14 may be further configured to: a fourth image and a fifth image of the hand are taken in response to an action of the hand of the access mechanism entering the item registering unit 40, wherein the fourth image is taken before the hand of the access mechanism enters the item registering unit 40, and the fifth image is taken after the hand of the access mechanism is removed from the item registering unit 40, wherein a state in which the respective access mechanism moves out of the item is determined based on a comparison of the fourth image and the fifth image, a change in at least one inventory position of the shelf at a predetermined relative position around an absolute position of the access mechanism is identified in response to the state in which the access mechanism holds the item, and a correspondence between the removal of the item and the change in the inventory position is thereby judged. Here, the return of an item from the item escrow unit to the inventory location is determined by both the removal of the item and the change in inventory location. The changed stock location may thus be first confirmed based on the absolute and relative locations determined by the first image, while the returned item is analytically determined based on the second, third, fourth and fifth images. At this point the inventory location may be updated accordingly or a cargo cleaning procedure initiated, such as a summoning manual intervention to clear items that do not fit the shelf cargo, or to send the wrong item into the recycling channel in an automatic feed. For example, items that do not fit the pallet load or wrong items are returnable or placed in the recycling tunnel by a movable robot arm or an automated cleaning vehicle with a robot arm mounted on the ceiling. Preferably, after determining the returned item based on the second, third, fourth and fifth images, the item identifier representing the item is removed from a pick list associated with at least one access mechanism. Preferably, the fourth image is taken by the hand of the access mechanism before the hand enters the load lock 40 without holding an item. The fifth image is taken after the hand of the access mechanism removes the item from the item holding unit 40.

According to a preferred embodiment, the fourth image acquisition device 14 may be further configured for: in response to the access mechanism reaching the settlement area being activated, the space size of the cargo scratch pad unit 40 occupied by the items within the cargo scratch pad unit 40 is collected and a matched size bag is provided to the access mechanism by the system providing equipment indicating the settlement area based on the space size. By this means, a bag of a proper size can be provided to the access mechanism, and excessive time waste due to the judgment of the required size of the bag in the settlement area is avoided, thereby improving the settlement efficiency.

According to a preferred embodiment, the invention also discloses a fifth image acquisition device 15 arranged in the access way of the access mechanism for acquiring identification marks for payment and/or registration. Preferably, the identification mark may be at least one of RFID, a two-dimensional code, a fingerprint, a human body contour, a human face contour, an iris image, and a typical thermal characteristic image. For example, after an access mechanism is detected as holding a scratch pad unit 40, the system associates the scratch pad unit 40 with the access mechanism. In the case where the access mechanism is an unregistered user at the time of payment, the owner of the access mechanism may be alerted to register or remotely complete payment directly at or before the time of payment. When the owner of the access organization or the user as a natural person registers, the owner or the user can be required to select an identification mark as a certificate for later fee settlement, and the user is required to input corresponding information according to the identification mark selected by the user. Assuming that the user selects to pay through iris image, graphic two-dimensional code or RFID, the access mechanism posts the graphic two-dimensional code (posted in the access mechanism) or corresponding eyes to the acquisition area of the fifth image acquisition device 15 during the payment, and after the verification of the system is passed, the payment is completed by using the bound payment account; or the RFID of the access mechanism or a graphic two-dimensional code input system thereof is used for settlement.

According to a preferred embodiment, the fifth image acquisition device 15 is activated only in response to an event that the fourth image acquisition device 14 provides an image. Preferably, the fifth image capturing device 15 adjusts its optical capturing direction and focal length in response to an event that the fourth image capturing device 14 provides an image and according to the position of the cargo temporary storage unit 40 equipped with the fourth image capturing device 14 with respect to the access way.

According to a preferred embodiment, the first image is captured by the first image capturing device 11 after the second image capturing device 12 captures a second image of the hand of the access mechanism, wherein the second image is captured before the hand of the access mechanism enters at least one inventory location of the shelf. Therefore, the unmanned supermarket can simultaneously deal with the scene of purchasing by a plurality of people, and the system can start a series of responsible image analysis operations only when the access mechanism with the first image performs object access or replacement again. In other words, the tracking of the access mechanism by the system begins when it attempts to reach the items on the shelves.

Preferably, the item registering unit 40 sets an introduction timing period (for example, 5 seconds) in response to the state of the item held by the access mechanism from the initial point of time when the absolute position is determined based on the first image, and after the specific timing period elapses, if the fourth image capturing device 14 fails to capture the action of the item entering the item registering unit 40, a prompt regarding the attribute of the item is issued to the access mechanism holding the item through the item registering unit 40, wherein the prompt may be in the form of an electronic signal, voice, image, or video. When the product temporary storage unit determines that the person is not a person, but is not an access mechanism in the form of a handling device, the product temporary storage unit 40 may prompt the access mechanism to sell the product that is removed from the shelf but is not loaded into the product temporary storage unit 40 for 20 yuan, and ten months are also required according to the expiration date. Preferably, the goods registration unit 40 further sets a reminding timing period (for example, 30 seconds) in response to the setting of the introduction timing period, and if the fourth image acquisition device 14 fails to acquire the action of entering the goods registration unit 40 after the reminding timing period, a prompt for correctly loading the current goods into the goods registration unit 40 is sent to an access mechanism holding the goods by the goods registration unit 40, wherein the prompt may take a graphic form such as an arrow pointing to the receiving port. Preferably, after the reminder timing period has elapsed, in the event that the fourth image acquisition device 14 fails to acquire an action of the item entering the item staging unit 40, the system also indicates, via the indicator 21, that the current item is to be placed back in the correct position on the shelf. The indication mark 21 may be in the form of a graphic pattern such as an arrow pointing to the area where the like article of the present article is located. More specifically, for example, the indication mark 21 is a green fluorescent arrow and can blink at a preset frequency. When the goods temporary storage unit judges that the holding party is an access mechanism in the form of carrying equipment rather than a person serving as a natural person, the goods temporary storage unit 40 transmits relevant goods introduction information to the access mechanism in an electronic signal mode, and the goods temporary storage unit sends a prompt to a remote user so as to finish purchasing decision as soon as possible.

Preferably, the system further comprises a sixth image acquisition device 16, the sixth image acquisition device 16 being activated after the lapse of the alert timing period in case the fourth image acquisition device 14 fails to acquire an action of the item entering the item holding unit 40 for acquiring a sixth image of a shelf associated with the item and a seventh image of the shelf, wherein the sixth image is taken before a hand holding the item by the access mechanism enters the shelf, and the seventh image is taken after the hand holding the item by the access mechanism is removed from the shelf, and the correspondence between the item and the change in the stock position is secondarily determined based on comparison of the sixth image and the seventh image. The sixth image capture device 16 is only activated after the reminder timing period has elapsed in the event that the fourth image capture device 14 fails to capture an item's action into the item staging unit 40 because the item requires more environmental information to playback onto the shelf and the difficulty of analysis is greater than removing an item from the shelf. Preferably, in the case of performing the first judgment by the second and third images, the second judgment is performed by the sixth and seventh images, so that the obtained correspondence between the article and the change of the stock position is more accurate and reliable. Preferably, in the event that the results of the two determinations are inconsistent, a manual check may be arranged or a field check by an intelligent robot. Preferably, the sixth image and the seventh image differ from the second image and the third image mainly in that the objects of the sixth image and the seventh image are shelves. While the objects of the second and third images are hands. The sixth image acquisition device 16 is provided in the sense that it is also possible to track the articles from the shelves correspondingly when they are taken out by the corresponding access means and then placed on other shelves. For example, after a storage mechanism takes a corresponding item at one inventory location, another item of the same type is seen at another inventory location, which is more attractive, and then the user wishes to remove the corresponding item and replace it, but at this point the storage mechanism is likely not to be in the vicinity of the corresponding shelf that takes the corresponding item, resulting in a low willingness to return the item to the corresponding inventory location, and the corresponding item may be placed directly on the shelf at the current location, which may result in the corresponding item being placed at a non-corresponding inventory location on a non-corresponding shelf. Thus, preferably, the system audibly plays an advisory reminder to the access mechanism holding the item when it is not being returned to its correct position on the shelf as indicated by the indicator mark 21. For example, a respectful customer, you do not put a good back in his correct position, in order to maintain a good shopping environment, avoid misleading others, ask you to put a good back in the correct position according to the indicator 21, and thank you for your cooperation. Preferably, the position error marker 22 is displayed in the wrong position of the item after the advisory reminder is played and the corresponding item has not been put back in the correct position, wherein the position error marker 22 may take the form of at least one of an exclamation mark, a cross mark or a letter. Preferably, after an item that has not been placed back in the correct position is removed from the shelf displaying the position error marker 22, the correct position and/or the correct selling price of the item is indicated by voice and/or graphical means. For example, by flashing the indicator mark 21 and voice: the respectful customer, you good, the article on your hand is placed in the wrong position, the correct position is the position of the flashing indication mark 21, please buy with reference to the corresponding price at the correct position, thank you for-! By the mode, workers can quickly find misplaced articles and quickly put the misplaced articles back to the correct positions, and working efficiency is improved. Meanwhile, wrong shopping guidance caused by the fact that price tags and actual articles are not corresponding to shoppers can be avoided, and user experience is improved. Moreover, in this way, since the correct position of the item can be quickly known, in the case where the distance required for the user to walk to the correct position is not far, even if the user chooses not to purchase the item after seeing the actual price tag and/or the correct selling price, some users can assist in putting the item back to its correct position, reducing the workload of staff. Preferably, the system may further comprise an indication display which may be used to play the correct selling price of the misplaced item by indicating to the access mechanism picking up the misplaced item when its correct location is not on its misplaced shelf.

According to a preferred embodiment, the position error markers 22 can distinguish the degree of deviation of the error in a hierarchical display by at least two different visual states. The degree of deviation is, for example, at least one of a distance deviation, a price deviation, and a utility deviation. Thus intuitively prompting staff or shoppers. For example, when the position error mark 22 can distinguish the distance deviation by a hierarchical display, the degree of deviation of the distance deviation is distinguished in three steps by a first color, a second color different from the first color, and a third color different from the first and second colors. The position error flag 22 is configured to a first color when the required path for the article to be placed back into its correct position from the current error position is less than or equal to a first path threshold, the position error flag 22 is configured to a second color when the required path for the article to be placed back into its correct position from the current error position is greater than the first path threshold and less than or equal to a second path threshold, and the position error flag 22 is configured to a third color when the required path for the article to be placed back into its correct position from the current error position is greater than a third path threshold. Preferably, the first color, the second color and the third color may be blue, yellow and red, respectively. The first path threshold is for example 5 meters. The second distance threshold is for example 20 meters. The third path threshold is, for example, 40 meters. Preferably, the utility deviation is, for example, a deviation of weight gain and weight loss, laxation and antidiarrheal. It should be noted that the invention may also be implemented in other ways or in combinations of ways to achieve different visual states, not just in different colors. For example, the invention may also be implemented by a combination of numbers and colors. Specifically, again taking the distance deviation as an example, the position error flag 22 is configured to be a first color and display a first number when the distance required to place the article back to its correct position from the current error position is less than or equal to a first path threshold, the position error flag 22 is configured to be a second color and display a second number when the distance required to place the article back to its correct position from the current error position is greater than the first path threshold and less than or equal to a second path threshold, and the position error flag 22 is configured to be a third color and display a third number when the distance required to place the article back to its correct position from the current error position is greater than a third path threshold. The first number is for example 1, the second number is for example 2 and the third number is for example 3.

According to a preferred embodiment, the system further comprises seventh image acquisition means 17, the seventh image acquisition means 17 being arranged to acquire an eighth image before the system indicates, via the indicator 21 on the shelf, that the current item is to be put back in the correct position on the shelf, the indicator 21 being displayed only if the indicator 21 is within the visual range of the access mechanism picking up the current item, by the system analysing the visibility of the indicator 21 corresponding to the correct position from the eighth image. Preferably, the correct position and/or the correct selling price of the item is indicated by means of voice in case the indicator 21 is not located in the visible range of the access mechanism picking up the current item, or in case there is a fixed obstacle between the access mechanism picking up the current item and the corresponding indicator 21, obstructing the line of sight. In this way, it is possible to avoid misleading others in the case that the correct position of the item is not within the visual range. Preferably, the system also analyzes the identity of the access mechanism picking up the current item according to the eighth image, and displays the indication mark 21 only if the indication mark 21 is located within the visual range of the access mechanism picking up the current item and the identity of the access mechanism is a sort of access mechanism. For example, the identity of the analyst may be determined by at least one of face recognition, wearing a specific workwear, and carrying an electronic tag.

According to a preferred embodiment, before displaying the indicator 21, the system further analyzes, based on the eighth image, a first display number of the indicator 21 to be displayed simultaneously in a visual range of the access mechanism picking up the current article, and when the first display number is two or more, associates the respective position error markers 22 and the indicator 21 in a grouping manner, and sets visually associated between the same group of position error markers 22 and the indicator 21 and visually discriminated between different groups. For example, in the same group, the position error mark 22 is at least partially colored the same as the indicator mark 21, and the color marks are different between different groups. For another example, in the same group, the position error mark 22 and the indication mark 21 display the same numerical mark, and the numerical marks are different between different groups. For another example, in the same group, the position error mark 22 and the indication mark 21 display the same graphic mark, and the graphic marks are different between different groups. The graphic indicia are, for example, circles, triangles or quadrilaterals.

According to a preferred embodiment, the system may further comprise: and eighth image acquisition means 18 for acquiring a ninth image of at least a part of the area of the shelf set position error marker 22, and analyzing, by the system, a match between the display of the actual position error marker 22 and the setting of the position error marker 22 in the system based on the ninth image, wherein, when the both do not match, the shelf where the non-matching position error marker 22 is located and a maintenance person is notified. In this way, a malfunctioning shelf of the location error marker 22 can be quickly identified and located, and maintenance can be quickly performed, reducing the occurrence of missing or mislabeling.

Preferably, the second image acquisition device 12, upon detecting a second image of the hand of one of the access mechanisms, also analyzes both: "the condition that the access mechanism holds an item determined based on the comparison of the second image and the third image" and "the action of the fourth image pickup device 14 of the item holding unit 40 carried by the access mechanism entering the item holding unit 40", the system grants the access mechanism access to an item in at least one inventory position of a rack in a predetermined relative position located near the absolute position of the access mechanism when the analysis result indicates that the access mechanism does not hold an item or holds only one item.

According to a preferred embodiment, the image acquisition devices mentioned in the present invention may use any type of camera and other alternative imaging devices. For example, one or more of the cameras may be RGB cameras. One or more of the cameras may be depth sensing cameras.

Preferably, in order to acquire a desired image, each image acquisition device may be one or more, and when a certain image acquisition device is plural, it may be provided in plural positions in a dispersed manner. For example, some image capture devices may be located overhead, such as on a ceiling, to capture images of the corresponding locations of the access mechanism and/or venue. As another example, some image capture devices may be located at or within an inventory location. For example, it may be located on an exterior portion of the inventory location and positioned to capture images of locations surrounding the access mechanism and/or inventory location. For another example, some image capture devices may be positioned in the area of the price tag toward the opposite side shelf to capture image information on the opposite side shelf.

According to a preferred embodiment, the system of the present invention further comprises a computing device 30. Computing device 30 may be local, remote, or any combination thereof. Likewise, the computing device 30 may be configured to communicate with at least one of the first, second, third, fourth, fifth, sixth, seventh, eighth image acquisition apparatuses and the cargo scratch pad unit 40 via a network, with an input component, an output component, and/or directly. It will be appreciated by those skilled in the art that the communication may be by wired and/or wireless means. Preferably, the computing device 30 is, for example, one or more servers. The server may include a processor and memory. The system may include or provide image processing (e.g., for user identification, expression identification, and/or item identification), inventory tracking, and/or location determination.

Example 2

This embodiment is a further improvement of embodiment 1, and the repeated contents are not repeated. Other embodiments may be used in addition to the preferred implementation of the present embodiment without conflict or contradiction.

The invention also discloses an inventory management method which can be realized by the system and other alternatives. The method of the present invention is implemented, for example, by using various components in the system.

According to a preferred embodiment, the method may comprise: using at least one first image acquisition means 11 for acquiring a first image and determining by the management system, based on said first image, the absolute position of at least one access mechanism within the area of the pallet and the relative position of the access mechanism with respect to at least one pallet carrying at least one item; using at least one second image acquisition device 12 for detecting a second image of the hand of the at least one access mechanism, said second image being taken before the hand of the access mechanism enters at least one inventory location of the shelf; using at least one third image acquisition device 13 for detecting a third image of the hand of at least one access mechanism, said third image being taken after the hand of the access mechanism has been moved out of at least one stock position of the shelf; wherein the status of the respective access mechanism holding the item is determined based on a comparison of the second image and the third image, a change in at least one inventory position of the rack at a predetermined relative position around the absolute position of the access mechanism is identified in response to the status of the access mechanism holding the item, and a correspondence between the item and the change in inventory position is determined therefrom.

According to a preferred embodiment, the method may further comprise: at least one fourth image acquisition device 14 is used, which is arranged at the item registering unit 40 for detecting an action of an item entering the item registering unit 40, wherein the correspondence between the item entering the item registering unit 40 and the change of the stock position is confirmed in response to the action of the item entering the item registering unit 40.

Example 3

This embodiment is a further modification of embodiments 1, 2 or a combination thereof, and the repeated description is omitted. Other embodiments may be used in addition to the preferred implementation of the present embodiment without conflict or contradiction.

According to a preferred embodiment, a shielding door is installed in front of the bar code scanner, the shielding door takes various payable two-dimensional codes of individuals used by personnel as switches, only one person can enter at a time, and the door is closed immediately after the codes are scanned. The person moves the shopping cart at a proper speed, the article is scanned by the bar code scanner, the information of the article to be paid and the total price of the currently purchased article are displayed on the touch display screen, and after the person confirms that the information is correct, the system automatically settles on the previously scanned two-dimensional code account. After confirming that settlement is correct, the shielding door behind the bar code scanner is opened, and meanwhile, the shielding door in front of the bar code scanner starts to scan the two-dimensional code of the next person. By the method, the problem that personnel steal commodities or settlement is wrong is avoided, and loss of supermarkets can be reduced.

Example 4

This embodiment is a further modification of embodiments 1, 2, 3 or a combination thereof, and the repeated descriptions are omitted. Other embodiments may be used in addition to the preferred implementation of the present embodiment without conflict or contradiction.

According to a preferred embodiment, the shelf is provided with at least one weight acquisition device for acquiring a weight difference between a hand of a respective person before entering at least one stock position of the shelf and after removal of the hand from the at least one stock position of the shelf, and after the system has determined the status of the respective person holding the item based on a comparison of the second image and the third image, the status of the item is checked by the determined weight of the item and the weight difference recorded in the system.

In addition, the invention can also be used for preliminarily judging the state of the article through directly installing the weight acquisition device on the goods shelf and through the weight difference acquired by the acquisition device. It will be appreciated by those skilled in the art that in such a case, the weight of the plurality of items on one shelf should be differentiated to achieve the purpose of identifying the status of the items.

Example 5

This embodiment is a further modification of embodiments 1, 2, 3, 4 or a combination thereof, and the repeated descriptions are omitted. Other embodiments may be used in addition to the preferred implementation of the present embodiment without conflict or contradiction.

The embodiment also discloses an unmanned supermarket system, which has the technical characteristics that the system is the same as the inventory management system in whole or in part so as to achieve the corresponding technical effects.

The embodiment also discloses an inventory management system of the unmanned supermarket, and the technical characteristics of the system are the same as those of the inventory management system in whole or in part so as to achieve corresponding technical effects.

The embodiment also discloses an inventory management method of the unmanned supermarket, and the technical characteristics of the method are the same as those of the inventory management method in whole or in part so as to achieve corresponding technical effects.

Although the present invention has been described in detail, modifications within the spirit and scope of the invention will be apparent to those of skill in the art. Such modifications are also considered to be part of this disclosure. In view of the foregoing discussion, related knowledge in the art, and references or information discussed above in connection with the background (all incorporated by reference herein), further description is deemed unnecessary. Furthermore, it should be understood that various aspects of the invention and portions of various embodiments may be combined or interchanged both in whole or in part. Moreover, those of ordinary skill in the art will appreciate that the foregoing description and drawings are by way of example only and are not intended to limit the invention.

The foregoing discussion of the present disclosure has been presented for purposes of illustration and description. This is not intended to limit the disclosure to the form disclosed herein. In the foregoing detailed description, for example, various features of the disclosure are grouped together in one or more embodiments, configurations, or aspects for the purpose of streamlining the disclosure. Features of embodiments, configurations, or aspects may be combined in alternative embodiments, configurations, or aspects than those discussed above. This method of disclosure is not to be interpreted as reflecting an intention that the disclosure requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment, configuration, or aspect. Thus, the following claims are hereby incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this disclosure.

Claims

1. An image recognition system for inventory management comprises an image acquisition unit (1) and an image recognition unit (2),

it is characterized in that the method comprises the steps of,

the image acquisition unit (1) acquires a first image and a second image and a third image of the hand of at least one access mechanism, the second image being taken before the hand of the access mechanism enters at least one inventory location of a shelf, the third image being taken after the hand of the access mechanism has been moved out of the at least one inventory location of the shelf, wherein,

The image recognition unit (2) determines the absolute position of at least one access means in the area of the pallet and the relative position of the access means with respect to at least one pallet carrying at least one item based on the first image and the state of the corresponding access means holding the item based on a comparison of the second image and the third image, wherein,

in response to the state of the storage mechanism holding the articles, the change of at least one inventory position of the goods shelves at the relative position around the absolute position of the storage mechanism is identified, and the correspondence between the articles and the change of the inventory position is judged, then the relative position and the absolute position are adopted to conduct classification analysis on the inventory position change and the personnel/storage mechanism, the goods shelves and the personnel or the storage mechanism which are interacted with the storage mechanism are accurately related to each other, so that the areas needing analysis and calculation can be dynamically divided according to the change of the articles and the inventory position at any time and any place.

2. The image recognition system as defined in claim 1, wherein the image acquisition unit (1) comprises at least:

a first image acquisition device (11) for acquiring a first image;

A second image acquisition device (12) for detecting a second image of the hand of the at least one access mechanism;

and a third image acquisition device (13) for detecting a third image of the hand of the at least one access mechanism.

3. The image recognition system of claim 2, wherein the system further comprises a computing device (30), wherein,

in case the image recognition unit (2) recognizes a change in at least one inventory position of shelves around the absolute position of the access mechanism and/or in a predetermined relative position in response to the state of the access mechanism holding an item,

the computing device (30) determines a correspondence between the item and a change in the inventory location, generates the inventory information, wherein,

the computing device (30) records inventory information of each item in history information of each item in a certain period, and obtains inventory status information of each item after analysis based on the history information and the inventory information.

4. An image recognition system as claimed in claim 3, characterized in that the system further comprises a policy unit (5), the policy unit (5) triggering an adjustment of the acquisition scheme of the image acquisition unit (1) and/or an adjustment of the recognition scheme of the image recognition unit (2) based on the stock state information, wherein,

The strategy unit (5) can carry out classification analysis on the change of the relative position and/or the absolute position to the stock position and the access mechanism, and associate the access mechanism with a goods shelf interacted with the access mechanism, so that the image acquisition scheme of the image acquisition unit (1) to the access mechanism and/or the identification scheme of the image identification unit (2) can be dynamically adjusted at any time and any place according to the change of the goods and the stock position.

5. The image recognition system according to claim 4, characterized in that the system further comprises an RFID unit (6) for marking and/or scanning the item;

the computing device (30) is capable of storing image attributes of an RFID-tagged item in association with underlying information of the RFID-tagged item in a storage unit (6);

the image acquisition unit (1) can record image information of the goods before the RFID scans the goods; the computing device (30) is capable of performing information matching with image attributes of the RFID-tagged item based on the characteristic parameters of the image information identified by the image identification unit;

When the matching is successful, the storage unit automatically registers the historical information of the goods and/or the RFID automatically scans the label of the goods;

when the matching fails, the alarm unit (7) prompts the user to register the historical information of the goods.

6. The image recognition system according to claim 5, wherein the system further comprises fourth image acquisition means (14) arranged at the item holding unit (40) for detecting an entry of an item into the item holding unit (40),

wherein the computing device (30) is responsive to an action of an item entering the item staging unit (40), thereby confirming a correspondence between the item entering the item staging unit (40) and a change in the inventory location.

7. The image recognition system as defined in claim 6, wherein the fourth image acquisition device (14) is further configured to: capturing a fourth image of the hand and a fifth image of the hand in response to the access of the hand to the load lock (40), wherein the fourth image is captured before the access of the hand to the load lock (40), and the fifth image is captured after the access of the hand to the load lock (40), wherein,

Determining a status of the respective access mechanism moving out of the item based on a comparison of the fourth image and the fifth image, identifying a change in at least one inventory location of the racks in a predetermined relative position around an absolute position of the access mechanism in response to the status of the access mechanism holding the item, and thereby determining a correspondence between the movement out of the item and the change in the inventory location.

8. The image recognition system according to claim 7, wherein the fourth image acquisition device (14) is further configured to: in response to the access mechanism reaching the settlement area, the method includes collecting the space size of the goods temporary storage unit (40) occupied by the goods in the goods temporary storage unit (40), and providing bags with matched sizes for the access mechanism by providing equipment according to the space size knowledge settlement area.

9. The image recognition system of claim 8, wherein the system further comprises: a fifth image acquisition device (15) arranged in the access way of the access means for acquiring identification marks for payment and/or registration, wherein,

the fifth image acquisition device (15) is activated only in response to an event that the fourth image acquisition device (14) provides an image, wherein,

In response to an event that the fourth image capturing device (14) provides an image and in accordance with the position of the cargo scratch unit (40) equipped with the fourth image capturing device (14) relative to the access way, the fifth image capturing device (15) adjusts its optical capturing direction and focal length.

10. The image recognition system as claimed in claim 9, wherein the first image is captured by the first image capturing device (11) after the second image capturing device (12) captures a second image of the hand of the access mechanism,

the second image is taken before the hand of the access mechanism enters at least one inventory location of the shelf.