WO2019218303A1 - 智能售货柜、物品识别方法、装置、服务器和存储介质 - Google Patents

智能售货柜、物品识别方法、装置、服务器和存储介质 Download PDF

Info

Publication number
WO2019218303A1
WO2019218303A1 PCT/CN2018/087304 CN2018087304W WO2019218303A1 WO 2019218303 A1 WO2019218303 A1 WO 2019218303A1 CN 2018087304 W CN2018087304 W CN 2018087304W WO 2019218303 A1 WO2019218303 A1 WO 2019218303A1
Authority
WO
WIPO (PCT)
Prior art keywords
camera
image
mirror
smart vending
placement space
Prior art date
Application number
PCT/CN2018/087304
Other languages
English (en)
French (fr)
Inventor
张站朝
Original Assignee
深圳前海达闼云端智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳前海达闼云端智能科技有限公司 filed Critical 深圳前海达闼云端智能科技有限公司
Priority to PCT/CN2018/087304 priority Critical patent/WO2019218303A1/zh
Priority to CN201880001294.6A priority patent/CN108780505B/zh
Publication of WO2019218303A1 publication Critical patent/WO2019218303A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition
    • G—PHYSICS
    • G07—CHECKING-DEVICES
    • G07F—COIN-FREED OR LIKE APPARATUS
    • G07F11/00—Coin-freed apparatus for dispensing, or the like, discrete articles
    • G07F11/02—Coin-freed apparatus for dispensing, or the like, discrete articles from non-movable magazines
    • G—PHYSICS
    • G07—CHECKING-DEVICES
    • G07F—COIN-FREED OR LIKE APPARATUS
    • G07F9/00—Details other than those peculiar to special kinds or types of apparatus
    • G07F9/02—Devices for alarm or indication, e.g. when empty; Advertising arrangements in coin-freed apparatus
    • G07F9/026—Devices for alarm or indication, e.g. when empty; Advertising arrangements in coin-freed apparatus for alarm, monitoring and auditing in vending machines or means for indication, e.g. when empty

Definitions

  • the present application relates to the field of smart retailing, and in particular, to a smart vending cabinet, an item identification method, an apparatus, a server, and a storage medium.
  • the operation mode of the intelligent vending cabinet is usually three processes of scanning the code to open the door, the user selects the item, and closes the door automatically, which is convenient and quick and does not need to be on duty.
  • the inventors found that the current smart vending cabinet usually installs a camera in the container, and the image is captured by the camera and visually identifies the type and quantity of the items in the container through artificial intelligence (AI). Since the identification of the type of goods and the amount of data in the container is dependent on the image captured by the camera, if the product in the captured image is distorted or the distance between the camera and the product is too close, the image of the product is not comprehensive, which will greatly increase the difficulty of AI recognition. , to reduce the accuracy of the identification of the type and quantity of goods, at the same time, because the camera and the product need a certain distance, the space in the cabinet can be placed, resulting in low space utilization. Therefore, how to improve the image quality of the products in the smart vending cabinet by the camera is an urgent problem to be solved.
  • AI artificial intelligence
  • the technical problem to be solved in some embodiments of the present application is to provide an intelligent vending cabinet, an item identification method, a device, a server and a storage medium, which improve the effective pixels of the goods in the smart vending cabinet in the collected image, and improve the accuracy of the product identification. Sex, while increasing the space utilization of smart vending.
  • An embodiment of the present application provides a smart vending cabinet comprising: a cabinet provided with at least one placement space, a camera disposed at the bottom of each placement space, a mirror covering the top of the placement space and having a mirror surface facing the bottom of the placement space, and
  • the main control module is connected with the camera; the lens of the camera is facing the mirror in the space where the camera is placed, the range of the camera's angle of view covers the mirror of the mirror in which the camera is placed; the main control module controls the camera to shoot the mirror in the space. And transmitting the captured image to the item identification device for processing.
  • An embodiment of the present application further provides an item identification method for identifying an item in the smart vending cabinet, comprising: receiving an image of each placement space transmitted by the smart vending cabinet; respectively performing an image of each placement space Anti-interference processing; identification of items in the image after anti-interference processing.
  • the embodiment of the present application further provides an item identification device, wherein the item identification device is communicatively coupled to the smart vending container, the item identification device includes: a communication module, an image processing module, and an image recognition module; and the communication module is configured to receive the smart vending container transmission An image of each of the placement spaces; the image processing module is configured to perform anti-interference processing on each of the images of the placement space; and the image recognition module is configured to identify the objects in the image after the anti-interference processing.
  • the embodiment of the present application further provides a server, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instruction being at least one The processor executes to enable the at least one processor to perform the item identification method described above.
  • the embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and when the computer program is executed by the processor, implements the item identification method described above.
  • the camera captures the image in the mirror and then captures the image.
  • the items in the entire placement space increase the object distance between the object and the camera due to the image obtained from the mirror, reduce the possibility of distortion of the object in the captured image, and improve the ability to recognize the image;
  • the object distance between the camera and the article thereby increasing the available space in the placement space, and improving the utilization rate of the placed space.
  • FIG. 1 is a schematic front view of a smart vending cabinet in a first embodiment of the present application
  • FIG. 2 is a schematic front structural view of a smart vending cabinet in a second embodiment of the present application.
  • FIG. 3 is a schematic front structural view of a smart vending cabinet in a third embodiment of the present application.
  • FIG. 4 is a schematic front structural view of a smart vending cabinet in a fourth embodiment of the present application.
  • FIG. 5 is a schematic flow chart of an item identification method in a fifth embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of an article identification device in a sixth embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of a server in a seventh embodiment of the present application.
  • the first embodiment of the present application relates to a smart vending cabinet that can be placed in a shopping mall, a subway station, a theater, etc., so that the user can purchase items at any time.
  • the smart vending cabinet 10 includes: a cabinet 101 provided with at least one placement space 102, a camera 103 disposed at the bottom of each placement space, a mirror 105 covering the top of the placement space and having a mirror surface facing the bottom of the placement space, and communicating with the camera 103
  • the connected main control module 106 has a specific structure as shown in FIG.
  • the lens of the camera 103 faces the mirror 105 in the space 102 where the camera 103 is located, and the range of the camera 103 covers the mirror surface of the mirror 105 in which the camera 103 is placed; the main control module 106 controls the camera to the space.
  • the mirror in 102 is photographed, and the image obtained by photographing is transmitted to the article identification device for processing.
  • the placement space is obtained by separating the cabinet 101 by the layer partition 104.
  • the layer partition 104 is bridged on opposite sides of the cabinet 101, thereby dividing the cabinet 101 into a plurality of placement spaces. 102.
  • the number of the placement spaces 102 is determined by the number of the layer partitions 104. For example, assuming that there are three layer partitions, and the layer partitions separate the cabinets by crossing the cabinets with respect to the two sides, the obtained space is obtained. The number is four.
  • the manner in which the layer partition 104 separates the cabinet 101 is not limited, and may be specifically set according to actual needs.
  • the bottom device camera 103 of the space 102 is placed, and a mirror 105 is mounted on the top of the placement space 102.
  • the mirror 105 covers the top of the entire placement space 102, and the mirror 105 can be attached to the placement space in a conforming manner.
  • the top of 102 is the top of 102.
  • the main control module 106 controls the camera 103 to be in different viewing angle ranges and captures the mirror in the placement space 102, and obtains an image taken by the camera 103 in different viewing angle ranges; from the image taken by the camera 103 in different viewing angle ranges. Selecting the best image, determining the range of the angle of view corresponding to the best image as the optimal range of angle of view, and adjusting the angle of view of the camera 103 to the optimal range of angles of view, wherein the best image includes all items in the space 102 in which the camera 103 is located And the item has the largest area in the image.
  • the angle of view of the camera 103 can be changed by adjusting the focal length of the camera or by adjusting the distance between the camera and the mirror.
  • the main control module 106 controls the camera 103 to capture the mirror in the placement space 102 at different focal lengths, and obtain images captured by the camera 103 at different focal lengths; select the most images from the camera at different focal lengths. Good image, the focal length corresponding to the best image is determined as the best focal length, and the focal length of the camera is adjusted to the optimal focal length. Of course, every image taken should be clear.
  • the main control module 106 can adjust the viewing angle of the camera at a fixed focal length interval until the maximum focal length or minimum focal length of the camera is reached, wherein when the camera 103 is in a range of viewing angles and photographed, the captured image and the focal length corresponding to the image are acquired.
  • the main control module 106 compares each image and selects the best image by comparing the image information of the items in the image. The best image is determined to know the focal length corresponding to the best image, and the main control module controls the camera 103 to adjust the focal length to the optimal focal length.
  • the main control module controls the focal length of the camera to be 3 mm, and takes a picture of the mirror in which the camera is placed, obtains image 1 and records the focal length corresponding to image 1, and then controls the camera.
  • the focal length is 4.0mm, the mirror of the space is photographed, and the image 2 is obtained and the focal length corresponding to the image 2 is recorded.
  • the main control module adjusts the focal length of the camera at a focal length of 1 mm, and shoots the mirror in the current space.
  • Corresponding 10 clear images and recording 10 corresponding focal length information compare the obtained 10 images (such as judging whether the image contains all the items in the current placement space, and comparing each item in the current image Pixel area size), if the image 2 includes all the items in the current placement space, and the image has the largest area in the image 2, then the image 2 is selected as the best image, the optimal focal length is 4.0 mm, and the main control module will be the camera.
  • the focal length is adjusted to 4.0mm.
  • the main control module 106 controls the camera 103 to take a picture of the mirror in the placement space 102 when the camera 103 is in different viewing angle ranges.
  • the staff places items into the smart vending cabinet. After the placement is completed, the staff sends a signal to the vending cabinet through the terminal to place the contents of the cabinet. After receiving the signal, the main control module can control the camera to be in different viewing angles. The mirror of the mirror placed in the space 102 is photographed.
  • the main control module 106 can periodically control the camera 103 to shoot the mirror in the placement space 102 at different viewing angle ranges, and can also control the placement space 102 when the camera 103 is in different viewing angle ranges when detecting other trigger signals.
  • the inside mirror is taken for shooting, and this embodiment is no longer an example.
  • the mirror surface of the mirror in the current placement space may be photographed periodically, or the current placement space 102 may be received after receiving the photographing signal.
  • the mirror of the inner mirror is photographed, and the captured image is transmitted to the article identification device for image recognition, and the type of the item in the current cabinet and the corresponding number are determined.
  • the camera captures the image in the mirror and then captures the image.
  • the items in the entire placement space increase the object distance between the object and the camera due to the image obtained from the mirror, reduce the possibility of distortion of the object in the captured image, and improve the ability to recognize the image;
  • the object distance between the camera and the article thereby increasing the available space of the placement space, and improving the utilization of the placed space.
  • the second embodiment of the present application relates to a smart vending cabinet.
  • the embodiment is further improved on the basis of the first embodiment.
  • the specific improvement is: in this embodiment, the smart vending cabinet further includes A height-adjustable bracket 107 is disposed at a bottom center of the space 102, and the camera 103 is fixed to the bracket 107.
  • the specific structure of the smart vending cabinet 10 is as shown in FIG. 2.
  • the main control module 106 is configured to adjust the height of the bracket 107, and control the camera 103 to capture the mirror in the placement space 102 when the camera 103 is in different viewing angle ranges, and obtain an image taken by the camera 103 in different viewing angle ranges. And adjusting the viewing angle of the camera 103 to an optimal viewing angle range by adjusting the height of the bracket 107.
  • the bracket 10 includes a motor 1071, and the main control module 106 is specifically configured to adjust the height of the bracket by controlling the motor 1071.
  • the motor 1071 is mounted in the bracket 107, and the motor 1071 is connected to the main control module 106 (not shown in FIG. 2).
  • the main control module 106 rotates by controlling the motor 1071 to drive the height-adjustable bracket, thereby driving the bracket 107.
  • the camera 103 moves up and down. By adjusting the height of the camera 103, the distance between the camera 103 and the mirror 105 is changed, thereby changing the viewing angle range of the camera 103.
  • the motor 1071 can be mounted on other parts of the bracket 107, and is not limited to the positions listed in the embodiment.
  • the main control module 106 can adjust the bracket 107 at a fixed height interval until the height of the bracket 107 reaches the maximum height or minimum height of the bracket 107.
  • the main control module 106 acquires the image taken after adjusting the bracket 107 each time and the height of the bracket corresponding to the recorded image, and the main control module compares and analyzes each image.
  • the depth learning method can be used to compare the graphic features in each image. And color features, thereby determining an image including all items in the current placement space 102, and then selecting an image having the largest area of the item from the images including all items in the current placement space 102.
  • the selection process may be deep learning.
  • the method is to learn the graphic features and color features in the image, so as to select the best image that meets the conditions.
  • the method of deep learning is not described here.
  • the height of the bracket corresponding to the best image can be known, and the main control module 106 adjusts the height of the bracket 107 to the height of the bracket corresponding to the best image.
  • each captured image should be a clear image.
  • the height of the bracket can be adjusted from 10 cm to 15 cm, and the main control module adjusts the bracket at a height of 1 cm. Then, when the bracket is 10 cm, the image is taken for the current space. , obtain image 1 and record the height of the bracket when the current image is taken; when the bracket is 11 cm high, take an image of the current placement space, obtain image 2 and record the height of the bracket when the current image is taken; adjust the height in turn until the height of the bracket reaches 15 cm. At this time, 6 images and corresponding bracket heights are acquired. If the best image is determined as image 3 after deep learning, the main control module adjusts the bracket height to 12 cm.
  • the camera 103 may be a zoom camera, or the camera is a fixed focus camera. If the camera is a zoom camera and is mounted on the bracket, the main control module can adjust the focal length of the zoom camera first, and then adjust the height of the bracket; or adjust the height of the bracket first, and then adjust the focal length of the camera.
  • the intelligent vending cabinet provided by the embodiment adjusts the angle of view of the camera to the optimal viewing angle range by adjusting the adjustable height bracket, improves the quality of the image captured by the intelligent vending cabinet, and improves the image through the image.
  • the accuracy of item identification is a simple measure of the angle of view of the camera.
  • the third embodiment of the present application relates to a smart vending cabinet.
  • the embodiment is further improved on the basis of the second embodiment.
  • the specific improvement is: in the embodiment, the smart vending cabinet further includes a cabinet door 108.
  • the cabinet door 108 is hinged to the cabinet 101.
  • the specific structure of the smart vending cabinet 10 is shown in FIG.
  • the main control module 106 controls the camera 103 to photograph the mirror in each placement space in an optimal viewing angle range, and transmits the captured image to the article identification device for image. Identification.
  • a wired connection may be used between the main control module 106 and the camera 103, such as a network cable, a USB data cable, or the like, or a wireless connection, such as a wireless fidelity (wifi) or a wireless communication network (3G/4G/5G). Either way; or both wired and wireless, to ensure the data is not transmitted.
  • a wireless connection such as a wireless fidelity (wifi) or a wireless communication network (3G/4G/5G). Either way; or both wired and wireless, to ensure the data is not transmitted.
  • the main control module 106 can detect whether the door 108 is closed by the sensor installed on the door 108 or the sensor mounted on either side of the cabinet, and the main control module 106 is connected with the sensor (FIG. 3) The connection between the main control module and the sensor is not shown. After the main control module 106 detects that the cabinet door 108 is closed by the sensor, the main control module 106 controls the camera 103 to be in the optimal viewing angle range and images the mirror in the current placement space 102. The item is photographed, and the photographed image is transmitted to the item recognizing device for image recognition, and the kind of the item in the photographed image and the corresponding number are determined.
  • the detection method of the door 108 by the main control module 106 is not limited to the method listed in this embodiment, and may be detected by other methods, which are not enumerated here.
  • the smart vending cabinet provided by the embodiment can ensure the safety of the items placed in the smart vending cabinet by setting the cabinet door, and at the same time detecting the mirror in the space when the door is closed, the smart vending cabinet can be ensured in time. Check the items and quantity in the cabinet.
  • the fourth embodiment of the present application relates to a smart vending cabinet.
  • the embodiment is further improved on the basis of the third embodiment.
  • the specific improvement is: in this embodiment, any side of the placement space is set to a limited height.
  • Line 109, height limit line 109 is used to identify the highest height of the placement that is allowed to be placed, as shown in Figure 4 (only one height limit line is shown).
  • a height limit line may be disposed on any side of the cabinet 101, or a height limit line 109 may be disposed on the cabinet door 108.
  • the number of the height limit line 109 is not limited, and a limit height may be set on each side.
  • line 109 only one height limit line 109 may be provided.
  • the height limit line 109 can be drawn on the side of the cabinet 101, or disposed on the side of the cabinet 101 in a lined manner, and the color of the height limit line 109 can be in a distinct color, such as red. For example, suppose the height of the placement space is 20 cm.
  • the height limit line can be drawn at four-fifths of the height of the placement space, ie the height-limit line is drawn at the bottom of the placement space. At the centimeter. It can be understood that the height limit line 109 in each placement space may be different from the height of the bottom of the current placement space, or may be the same.
  • the above mirror can be a plane mirror, a convex mirror or a concave mirror, and the corresponding mirror can be selected according to actual needs.
  • the article placed in the space is prevented from being too high and affecting the shooting effect of the camera.
  • a fifth embodiment of the present application relates to an item identification method for identifying an item in a smart vending cabinet, and a specific flow of the item identification method is as shown in FIG. 5.
  • Step 501 Receive an image of each placement space transmitted by the smart vending cabinet.
  • the smart vending cabinet and the item identification device are communicably connected, and the main control module of the intelligent vending cabinet controls the camera to shoot the mirror in each of the placement spaces in the cabinet, and sends the captured image to the item identification device, and the item identification device receives For each image of the placement space, of course, in order to facilitate recognition of the image, at least one image is taken for each placement space and transmitted to the item identification device.
  • Step 502 Perform anti-interference processing on the images of each placement space.
  • anti-interference processing includes at least three methods, and three anti-interference processing methods will be described in detail below.
  • Anti-jamming treatment method 1
  • the interference image is removed from each image, and the interference image includes the image of the camera and the image outside the space.
  • the mirror surface of the mirror faces the bottom of the space and the camera is provided at the bottom of the space, there is a camera in the image in the mirror.
  • the shape of the camera is similar to that of the black cap. Therefore, in order to avoid misidentification of the camera as a cap or other items during image recognition, to ensure the accuracy of image recognition, it is necessary to remove the camera image from the image.
  • the image of each angle of the camera may be pre-stored, the image containing the camera is compared with the pre-stored camera image, and the camera image to be removed is determined according to the position of the camera.
  • the area outside the space can be removed from the image.
  • the four side walls of the space (which may include the cabinet door) are used as the boundary of the target area.
  • the target area boundary is removed from the image outside the boundary.
  • the effective features of the height-limiting line are stored in advance, so that it is convenient to determine whether the image includes a height-limiting line through the feature comparison, if the limit line is not detected, or the limit is detected.
  • the image is cropped directly using the edge of the mirror in the image as the cropping boundary. If the height limit line is detected and the height of the item in the image is below the height limit line, the image is cropped at the boundary of the height limit line.
  • the four sides of the placement space are not all finite high lines, they can be calculated by other height limit lines. For example, if there is only one height limit line, the height of the height limit line in the image can be obtained by calculation. Then, you can draw the other height limit lines in the image.
  • the actual image and the image in the mirror can be prevented from appearing at the same time, which reduces the difficulty of image recognition and avoids the probability of identifying objects in the non-placement space.
  • Anti-jamming treatment method three is the following:
  • the anti-interference processing method 1 and the anti-interference processing method 2 are arbitrarily combined, that is, the anti-interference processing method 1 can be performed first, and then the anti-interference processing method 2 is performed, or the anti-interference processing method 2 is performed first, and then the anti-interference processing method is performed. Interference processing method one.
  • Step 503 Identify the items in the image after the anti-interference processing.
  • the type of each item in the image after the anti-interference processing and the position in the image are identified.
  • the depth-learning method can be used to identify the image after the interference processing, and the type of the item in the figure and the position of the item in the image are determined. It is also possible to identify the items in the figure by means of comparison.
  • the method needs to store images of various angles of the items in advance, match the anti-interference processed image with the pre-stored item image, and determine the item type in the image according to the similarity.
  • the anti-interference processing of the received image eliminates the interference image in the imaging in the mirror, and the image can be prevented from appearing simultaneously with the image in the mirror by cutting the image.
  • the difficulty of image recognition is reduced, and the probability of identifying objects in the non-placement space is avoided.
  • a sixth embodiment of the present application relates to an article identification device that is communicatively coupled to a smart vending container, the article identification device 60 including: a communication module 601, an image processing module 602, and an image recognition module 603, the article identification device The specific structure is shown in Figure 6.
  • the communication module 601 is configured to receive an image of each placement space transmitted by the smart vending cabinet; the image processing module 602 is configured to perform anti-interference processing on each of the images of the placement space; and the image recognition module 603 is configured to identify the anti-interference The items in the image after processing.
  • This embodiment is a virtual device embodiment corresponding to the foregoing item identification method.
  • the technical details in the foregoing method embodiments are still applicable in this embodiment, and details are not described herein again.
  • a seventh embodiment of the present application relates to a server 70, the structure of which is shown in FIG.
  • the method includes: at least one processor 701; and a memory 702 communicatively coupled to the at least one processor 701.
  • the memory 702 stores instructions that are executable by at least one processor 701.
  • the instructions are executed by at least one processor 701 to enable the at least one processor 701 to perform the item identification method described above.
  • Memory 702 and processor 701 are connected in a bus manner, and the bus can include any number of interconnected buses and bridges that link together one or more processors 701 and various circuits of memory 702.
  • the bus can also link various other circuits, such as peripherals, voltage regulators, and power management circuits, as is well known in the art and, therefore, will not be further described herein.
  • the bus interface provides an interface between the bus and the transceiver.
  • the transceiver can be an element or a plurality of elements, such as multiple receivers and transmitters, providing means for communicating with various other devices on a transmission medium.
  • the data processed by the processor 701 is transmitted over the wireless medium via an antenna. Further, the antenna also receives the data and transmits the data to the processor 701.
  • the processor 701 is responsible for managing the bus and normal processing, and can also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions.
  • the memory 702 can be used to store data used by the processor when performing operations.
  • processor in this embodiment can perform the steps in the foregoing method embodiments, and the specific implementation functions are not described in detail. For details, refer to the technical details in the method embodiments, and details are not described herein.
  • An eighth embodiment of the present application is directed to a computer readable storage medium, which is a computer readable storage medium having stored therein computer instructions that enable a computer to perform the present application The item identification method involved in the fifth embodiment.
  • the display method in the above embodiment is completed by a program instructing related hardware, and the program is stored in a storage medium, and includes a plurality of instructions for making a device (may be It is a single chip, a chip, etc. or a processor that performs all or part of the steps of the methods described in various embodiments of the present application.
  • the foregoing storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), and a random access memory (RAM, Random-Access).
  • RAM random access memory

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Control Of Vending Devices And Auxiliary Devices For Vending Devices (AREA)

Abstract

一种智能售货柜、物品识别方法、装置、服务器和存储介质,包括：设置有至少一个放置空间(102)的柜体(103)，设置在每个放置空间(102)底部的摄像头(103)，覆盖放置空间(102)顶部且镜面朝向放置空间(102)底部的镜子(105)，以及与摄像头(103)通信连接的主控模块(106)；摄像头(103)的镜头正对摄像头(103)所在放置空间(102)内的镜子(105)，摄像头(103)的视角范围覆盖摄像头(103)所在放置空间(102)的镜子(105)的镜面；主控模块(106)控制摄像头(103)对放置空间(102)内的镜子(105)进行拍摄，并将拍摄获得的图像传输至物品识别装置进行处理。该智能售货柜，提高了采集到的图像中智能售货柜内商品的有效像素，提高商品识别的准确性，同时提高智能售货柜的空间利用率。

Description

智能售货柜、物品识别方法、装置、服务器和存储介质 技术领域
本申请涉及智能零售领域，尤其涉及一种智能售货柜、物品识别方法、装置、服务器和存储介质。
 背景技术
随着科技的不断发展，为了便于人们随时购买物品，出现了智能售货柜。智能售货柜的运行模式通常是扫码打开柜门、用户选用物品、关门自动结算三个过程，方便快捷且无需人值守。
 技术问题
发明人在研究现有技术过程中发现，目前的智能售货柜通常在货柜中安装摄像头，由摄像头采集图像并通过人工智能（AI）视觉识别货柜内的物品种类和数量。由于对货柜内商品种类和数据量的识别是依赖于摄像头采集的图像，若采集的图像中商品发生畸变或者因摄像头与商品距离过近，而造成商品图像不全面，将大大增加AI识别的困难，降低对商品种类和数量识别的准确性，同时，由于摄像头与商品之间需要一定的距离，缩小了柜体内可放置商品的空间，造成空间利用率低。因而，如何提高摄像头采集智能售货柜内商品的图像质量是亟待解决的问题。
 技术解决方案
本申请部分实施例所要解决的技术问题在于提供一种智能售货柜、物品识别方法、装置、服务器和存储介质，提高了采集到的图像中智能售货柜内商品的有效像素，提高商品识别的准确性，同时提高智能售货柜的空间利用率。
本申请的一个实施例提供了一种智能售货柜，包括：设置有至少一个放置空间的柜体，设置在每个放置空间底部的摄像头，覆盖放置空间顶部且镜面朝向放置空间底部的镜子，以及与摄像头通信连接的主控模块；摄像头的镜头正对摄像头所在放置空间内的镜子，摄像头的视角范围覆盖摄像头所在放置空间的镜子的镜面；主控模块控制摄像头对该放置空间内的镜子进行拍摄，并将拍摄获得的图像传输至物品识别装置进行处理。
本申请的一个实施例还提供了一种物品识别方法，用于识别上述智能售货柜中的物品，包括：接收智能售货柜传输的每个放置空间的图像；分别对每个放置空间的图像进行抗干扰处理；识别抗干扰处理后的图像中的物品。
本申请实施例还提供了一种物品识别装置，物品识别装置与上述智能售货柜通信连接，物品识别装置包括：通信模块、图像处理模块和图像识别模块；通信模块用于接收智能售货柜传输的每个放置空间的图像；图像处理模块用于分别对每个所述放置空间的图像进行抗干扰处理；图像识别模块用于识别所述抗干扰处理后图像中的物品。
本申请实施例还提供了一种服务器，包括：至少一个处理器；以及，与至少一个处理器通信连接的存储器；其中，存储器存储有可被至少一个处理器执行的指令，该指令被至少一个处理器执行，以使至少一个处理器能够执行上述的物品识别方法。
本申请实施例还提供了一种计算机可读存储介质，存储有计算机程序，该计算机程序被处理器执行时实现上述的物品识别方法。
 有益效果
相对于现有技术而言，本申请部分实施例中通过在每个放置空间顶部设置镜子，并将摄像头设置在放置空间底部且镜头朝向镜子的镜面，摄像头通过拍摄镜子中的成像，进而拍摄到整个放置空间内的物品，由于从镜子中获取物品的图像，增加了物品与摄像头之间的物距，减小拍摄图像中物品发生畸变的可能性，提高对图像的识别能力；同时由于减小了摄像头与物品之间的物距，进而增大了放置空间内的可利用空间，提高放置空间放置物品的利用率。
 附图说明
一个或多个实施例通过与之对应的附图中的图片进行示例性说明，这些示例性说明并不构成对实施例的限定，附图中具有相同参考数字标号的元件表示为类似的元件，除非有特别申明，附图中的图不构成比例限制。
图1是本申请第一实施例中智能售货柜的正面结构示意图；
图2是本申请第二实施例中智能售货柜的正面结构示意图；
图3是本申请第三实施例中智能售货柜的正面结构示意图；
图4是本申请第四实施例中智能售货柜的正面结构示意图；
图5是本申请第五实施例中物品识别方法的流程示意图；
图6是本申请第六实施例中物品识别装置的结构示意图；
图7是本申请第七实施例中服务器的结构示意图。
 本发明的实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请部分实施例进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。然而，本领域的普通技术人员可以理解，在本申请的各实施例中，为了使读者更好地理解本申请而提出了许多技术细节。但是，即使没有这些技术细节和基于以下各实施例的种种变化和修改，也可以实现本申请所要求保护的技术方案。
本申请的第一实施例涉及一种智能售货柜，该智能售货柜可放置于商场、地铁站、影院等地方，便于用户随时购买物品。该智能售货柜10，包括：设置有至少一个放置空间102的柜体101，设置在每个放置空间底部的摄像头103，覆盖放置空间顶部且镜面朝向放置空间底部的镜子105，以及与摄像头103通信连接的主控模块106，具体的结构如图1所示。
一个具体的实现中，摄像头103的镜头正对摄像头103所在放置空间102内的镜子105，摄像头103的视角范围覆盖摄像头103所在放置空间102的镜子105的镜面；主控模块106控制摄像头对放置空间102内的镜子进行拍摄，并将拍摄获得的图像传输至物品识别装置进行处理。
具体的说，放置空间由层隔板104分隔柜体101获得，如图1所示，层隔板104跨接在柜体101相对的两个侧面，从而将柜体101分隔成多个放置空间102。放置空间102的个数由层隔板104的数量决定，例如，假设由3个层隔板，且层隔板采用跨接柜体相对两个侧面的方式分隔该柜体，那么获得的放置空间的个数为4个。当然，本实施例中，并不限制层隔板104分隔柜体101的方式，具体可以根据实际需要设置。针对每个放置空间，都在放置空间102的底部设备摄像头103，在放置空间102的顶部安装镜子105，镜子105覆盖整个放置空间102的顶部，镜子105可以采用贴合的方式贴合在放置空间102的顶部。
一个具体的实现中，主控模块106控制摄像头103处于不同视角范围并对放置空间102内的镜子进行拍摄，获得摄像头103在不同视角范围拍摄的图像；从摄像头103在不同视角范围拍摄的图像中选取最佳图像，将最佳图像对应的视角范围确定为最佳视角范围，并将摄像头103的视角调整至最佳视角范围，其中，最佳图像中包括摄像头103所在放置空间102内的所有物品，且物品在图像中的面积最大。
具体的说，可以通过调节摄像头的焦距或者通过调节摄像头与镜面之间的距离，改变摄像头103视角范围。若摄像头为变焦摄像头，主控模块106控制摄像头103在不同焦距下对放置空间102内的镜子进行拍摄，获得摄像头103在不同焦距下拍摄的图像；从摄像头在不同焦距下拍摄的图像中选取最佳图像，将最佳图像对应的焦距确定为最佳焦距，调节摄像头的焦距至最佳焦距。当然，拍摄的每一个图像都应当清晰。
主控模块106可以以固定焦距间隔调节摄像头的视角，直至达到该摄像头的最大焦距或者最小焦距，其中，当摄像头103处于一种视角范围并拍摄时，获取拍摄的图像以及该图像对应的焦距。主控模块106比对每一个图像，通过比对图像中物品的影像信息，选出最佳图像。确定了最佳图像即可获知最佳图像对应的焦距，主控模块控制摄像头103将焦距调整至最佳焦距。
下面以一个具体的例子进行说明。
例如，假设摄像头焦距的取值范围为3mm至12mm，那么主控模块控制摄像头的焦距在3mm时对摄像头所在放置空间的镜子进行拍摄，得到图像1并记录图像1对应的焦距，之后控制摄像头的焦距在4.0mm时对该放置空间的镜子进行拍摄，得到图像2并记录图像2对应的焦距，主控模块以1mm的焦距间隔调整摄像头的焦距，并对当前放置空间中的镜子进行拍摄，得到对应的10个清晰的图像以及记录10个对应的焦距信息，对获得的10个图像进行比较（如判断图像中是否包含了当前放置空间内的所有物品，以及比较每个物品在当前图像中的像素面积大小），若图像2包括当前放置空间内所有物品，且物品在图像2中的面积最大的图像，那么选取图像2为最佳图像，则最佳焦距为4.0mm，主控模块将摄像头的焦距调整至4.0mm。
需要说明的是，主控模块106接收到柜体内物品放置完毕的信号之后，控制摄像头103处于不同视角范围时对放置空间102内的镜子进行拍摄。例如，工作人员向智能售货柜内放置物品，在放置完毕后，工作人员通过终端向售货柜发送柜内物品放置完毕的信号，主控模块接收到该信号后，即可控制摄像头处于不同视角范围时对放置空间102内的镜子的镜面进行拍摄。当然，本实施例中，主控模块106可以定时控制摄像头103处于不同视角范围对放置空间102内的镜子进行拍摄，还可以在检测其他触发信号时控制摄像头103处于不同视角范围时对放置空间102内的镜子进行拍摄，本实施例不再一一例举。
此外，值得一提的是，在主控模块106将摄像头103调整至最佳视角范围后，可以定时对当前放置空间内的镜子的镜面进行拍摄，或者在接收到拍摄信号后对当前放置空间102内的镜子的镜面进行拍摄，并将拍摄图像传输至物品识别装置进行图像识别，确定当前柜内的物品种类及对应的数量。
相对于现有技术而言，本申请部分实施例中通过在每个放置空间顶部设置镜子，并将摄像头设置在放置空间底部且镜头朝向镜子的镜面，摄像头通过拍摄镜子中的成像，进而拍摄到整个放置空间内的物品，由于从镜子中获取物品的图像，增加了物品与摄像头之间的物距，减小拍摄图像中物品发生畸变的可能性，提高对图像的识别能力；同时由于减小了摄像头与物品之间的物距，进而增大了放置空间的可利用空间，提高放置空间放置物品的利用率。
本申请的第二实施例涉及一种智能售货柜，本实施例是在第一实施例的基础上做了进一步改进，具体改进之处为：本实施例中，智能售货柜还包括设置在每个放置空间102的底部中心位置处且高度可调节的支架107，摄像头103固定于该支架107上。该智能售货柜10的具体结构如图2所示。
一个具体的实现中，主控模块106用于调节支架107的高度，并在摄像头103处于不同视角范围时控制摄像头103对放置空间102内的镜子进行拍摄，获得摄像头103在不同视角范围拍摄的图像；以及通过调节支架107的高度将摄像头103的视角调整至最佳视角范围。
具体的说，支架中包括电机1071，主控模块106具体用于通过控制电机1071调节支架的高度。电机1071安装在支架107内，且电机1071与主控模块106连接（图2中未示出），主控模块106通过控制电机1071转动，带动可调节高度的支架，进而带动安装在支架107上的摄像头103上下运动。通过调节摄像头103的高度，改变摄像头103与镜子105之间的距离，从而改变摄像头103的视角范围。
需要说明的是，电机1071可以安装在支架107的其他部位，并不限于本实施例中列举的位置。
主控模块106可以是以固定高度间隔对支架107进行调节，直至支架107的高度达到该支架107的最大高度或者最小高度。当摄像头103每次拍摄时，记录当前支架处于的高度，并获取当前拍摄的图像。主控模块106获取每次调节支架107后拍摄的图像以及记录图像对应的支架高度，主控模块对每个图像进行比对分析，如，可以通过深度学习的方法比对每个图像中图形特征、颜色特征，从而确定出包括当前放置空间102内所有物品的图像，再从包括当前放置空间102内所有物品的图像中选出物品的面积最大的图像，同样，选取的过程可以采用深度学习的方法对图像中图形特征、颜色特征进行学习，从而选出符合条件的最佳图像，此处不对深度学习的方法进行赘述。确定出最佳图像后，即可获知最佳图像对应的支架高度，主控模块106调整支架107的高度到最佳图像对应的支架高度，同样，每个拍摄的图像应当为清晰图像。
下面以一个具体的例子进行说明。
例如，假设摄像头为定焦摄像头，支架高度的可调节范围为10厘米至15厘米，主控模块以1厘米的高度间隔对支架进行调节，那么，支架在10厘米时，对当前放置空间拍摄图像，得到图像1并记录当前拍摄图像时支架的高度；支架在11厘米高时，对当前放置空间拍摄图像，得到图像2并记录当前拍摄图像时支架的高度；依次调节高度，直至支架高度到15厘米。此时，获取到6个图像以及对应的支架高度，若通过深度学习后，确定最佳图像为图像3，那么主控模块将支架高度调整至12厘米。
需要说明的是，摄像头103可以为变焦摄像头，或者摄像头为定焦摄像头。若摄像头为变焦摄像头且安装在支架上，那么主控模块可以先调整变焦摄像头的焦距，再调整支架高度；或者先调整支架高度，再调整摄像头的焦距。
与现有技术相比，本实施例提供的智能售货柜，通过调整可调节高度的支架，从而调节摄像头的视角范围到最佳视角范围，提高智能售货柜拍摄的图像的质量，提高通过图像进行物品识别的准确性。
本申请的第三实施例涉及一种智能售货柜，本实施例是在第二实施例的基础上做了进一步改进，具体改进之处为：本实施例中，智能售货柜还包括柜门108，柜门108与柜体101铰接。该智能售货柜10的具体结构如图3所示。
一个具体的实现中，主控模块106在检测到柜门108关闭后，控制摄像头103在最佳视角范围对每个放置空间内的镜子进行拍摄，并将拍摄的图像传输至物品识别装置进行图像识别。
具体的说，主控模块106与摄像头103之间可以采用有线连接，如，网线、USB数据线等；或者采用无线连接，如无线保真（wifi）、无线通信网络（3G/4G/5G）等方式；或者既采用有线连接又采用无线连接的方式，确保数据的无障碍传输。
主控模块106可以通过柜门108上安装的传感器，或者柜体两个侧面中任一面安装的传感器，来实现对柜门108是否关闭的检测，主控模块106与传感器通信连接（图3中未示出主控模块与传感器之间的连接），主控模块106通过传感器检测到柜门108关闭后，主控模块106控制摄像头103处于最佳视角范围并对当前放置空间102内镜子中成像的物品进行拍摄，并将拍摄的图像传输至物品识别装置进行图像识别，确定拍摄图像中的物品种类和对应的数量。
可以理解的是，主控模块106对柜门108的检测方法并不限制于本实施例中列举的方法，还可以采用其他方式进行检测，此处不再一一进行列举。
本实施例提供的智能售货柜，通过设置柜门可以确保放置在智能售货柜中物品的安全，同时在检测到柜门关闭时，对放置空间内的镜子进行拍摄，可以保证智能售货柜及时的清点柜内的物品及数量。
本申请的第四实施例涉及一种智能售货柜，本实施例是在第三实施例的基础上做了进一步改进，具体改进之处为：本实施例中，放置空间的任意侧面设置有限高线109，限高线109用于标识允许放置的放置物的最高高度，如图4所示（图中仅示意了一根限高线）。
具体的说，可以在柜体101的任意侧面设置限高线，或者在柜门108上设置限高线109，限高线109的根数不做限制，可以每个侧面均设置一根限高线109，也可以仅设置一根限高线109。限高线109可以画在柜体101的侧面，或者以贴线的方式设置在柜体101的侧面，限高线109的颜色可以采用明显的颜色，如红色。例如，假设放置空间的高为20厘米，为了便于从放置空间中拿出物品，那么限高线可以画在放置空间的高的五分之四处，即限高线画在距放置空间底部16厘米处。可以理解的是，每个放置空间内的限高线109距离当前放置空间底部的高度可以不同，也可以相同。
值得一提的是，上述镜子可以是平面镜、凸面镜或者凹面镜，具体可以根据实际需要选择对应的镜子。
本实施例中提供的智能售货柜，通过在智能售货柜的任意侧面设置限高线，防止放置在放置空间的物品过高而影响摄像头的拍摄效果。
本申请的第五实施例涉及一种物品识别方法，该物品识别方法用于识别智能售货柜中的物品，该物品识别方法的具体流程如图5所示。
步骤501：接收智能售货柜传输的每个放置空间的图像。
具体的说，智能售货柜和物品识别装置通信连接，智能售货柜的主控模块控制摄像头对柜内的每个放置空间内镜子进行拍摄，并向物品识别装置发送拍摄的图像，物品识别装置接收每个放置空间的图像，当然，为了便于对图像进行识别，对每个放置空间拍摄至少一个图像并传输至物品识别装置。
步骤502：分别对每个放置空间的图像进行抗干扰处理。
具体的说，柜体内的摄像头拍摄的图像中可能存在其他与当前放置空间内物品无关的影像，为了减少对图像识别的干扰，需要对接收的图像进行抗干扰处理。抗干扰处理的具体过程至少包括三种方法，下面将详细说明三种抗干扰处理方法。
抗干扰处理方法一：
从每个图像中去除干扰影像，干扰影像包括摄像头的影像以及放置空间外的影像。
具体的说，由于镜子的镜面面向放置空间底部，而放置空间底部设置有摄像头，那么镜中成像中存在摄像头。摄像头的外形与黑色的瓶盖相近，因而，为了避免在图像识别过程中将摄像头误识别为瓶盖或者识别成其他物品，确保图像识别的准确性，需要从图像中去除摄像头影像。可以预先存储摄像头各个角度的图像，将包含摄像头的图像与预存的摄像头图像进行比对，并根据摄像头的位置，确定需要去除的摄像头影像。
当然，镜中可能存在放置空间外的影像，因此，可以将放置空间外的区域从图像中抠除，例如，以放置空间的四个侧壁（可以包括柜门）作为目标区域的边界，贴合目标区域边界将边界外的影像去除。
抗干扰处理方法二：
判断图像中是否包含限高线；若是，且确定图像中物品的高度小于限高线的高度，则按照限高线裁切图像，否则，按照图像中镜子的边缘裁切图像。
具体的说，预先存储限高线的有效特征（如，红色的直线，虚点线），便于通过特征比对判断图像中是否包含限高线，若是未检测到限高线，或者检测到限高线且存在高度高于限高线的物品，则直接以图像中镜子的边缘作为裁切边界，对图像进行裁切。若是检测到限高线，且图像中物品的高度均低于限高线，则以限高线为边界对图像进行裁切。可以理解的是，若放置空间的4个侧面并非都有限高线，可以通过其他限高线计算获得，例如，若只有一个限高线，通过计算可以获得该限高线在图像中所处高度，继而可以在图像中画出其他侧面的限高线。
通过裁切图像的方式可以避免实际影像与镜中的影像同时出现图像中，减少图像识别的难度，同时避免识别出非放置空间内物品的概率。
抗干扰处理方法三：
具体的说，将抗干扰处理方法一与抗干扰处理方法二进行任意组合，即可以先进行抗干扰处理方法一，再进行抗干扰处理方法二，或者先进行抗干扰处理方法二，再进行抗干扰处理方法一。
步骤503：识别抗干扰处理后的图像中的物品。
一个具体的实现中，识别抗干扰处理后的图像中每个物品的种类以及在图像中的位置。
具体的说，可以采用深度学习的方式对抗干扰处理后的图像进行识别，确定出图中物品的种类以及物品在图像中的位置。还可以采用比对的方式识别图中物品，该方法需要预先存储物品各个角度的图像，将抗干扰处理后的图像和预先存储的物品图像进行匹配，根据相似度确定图像中的物品种类。
本实施例相对于现有技术而言，通过对接收的图像进行抗干扰处理，消除通过拍摄镜子中的成像中的干扰影像，通过裁切图像的方式可以避免实际影像与镜中的影像同时出现图像中，减少图像识别的难度，同时避免识别出非放置空间内物品的概率。
本申请的第六实施例涉及一种物品识别装置，该物品识别装置与智能售货柜通信连接，物品识别装置60包括：通信模块601、图像处理模块602和图像识别模块603，该物品识别装置的具体结构如图6所示。
具体的说，通信模块601用于接收智能售货柜传输的每个放置空间的图像；图像处理模块602用于分别对每个放置空间的图像进行抗干扰处理；图像识别模块603用于识别抗干扰处理后图像中的物品。
本实施例是与上述物品识别方法对应的虚拟装置实施例，上述方法实施例中技术细节在本实施例中依然适用，此处不再赘述。
需要说明的是，以上所述的装置实施例仅仅是示意性的，并不对本申请的保护范围构成限定，在实际应用中，本领域的技术人员可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的，此处不做限制。
本申请的第七实施例涉及一种服务器70，其结构如图7所示。包括：至少一个处理器701；以及，与至少一个处理器701通信连接的存储器702。存储器702存储有可被至少一个处理器701执行的指令。指令被至少一个处理器701执行，以使至少一个处理器701能够执行上述的物品识别方法。
存储器702和处理器701采用总线方式连接，总线可以包括任意数量的互联的总线和桥，总线将一个或多个处理器701和存储器702的各种电路链接在一起。总线还可以将诸如外围设备、稳压器和功率管理电路等之类的各种其他电路链接在一起，这些都是本领域所公知的，因此，本文不再对其进行进一步描述。总线接口在总线和收发机之间提供接口。收发机可以是一个元件，也可以是多个元件，比如多个接收器和发送器，提供用于在传输介质上与各种其他装置通信的单元。经处理器701处理的数据通过天线在无线介质上进行传输，进一步，天线还接收数据并将数据传送给处理器701。
处理器701负责管理总线和通常的处理，还可以提供各种功能，包括定时，外围接口，电压调节、电源管理以及其他控制功能。而存储器702可以被用于存储处理器在执行操作时所使用的数据。
需要说明的是，本实施例中的处理器能够执行上述的方法实施例中实施步骤，具体的执行功能并未详细说明，可参见方法实施例中的技术细节，此处不再赘述。
本申请的第八实施例涉及一种计算机可读存储介质，该可读存储介质为计算机可读存储介质，该计算机可读存储介质中存储有计算机指令，该计算机指令使计算机能够执行本申请第五实施例中涉及的物品识别方法。
需要说明的是，本领域的技术人员能够理解，上述实施例中显示方法是通过程序来指令相关的硬件来完成的，该程序存储在一个存储介质中，包括若干指令用以使得一个设备（可以是单片机，芯片等）或处理器（processor）执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器（ROM，Read-Only Memory）、随机存取存储器（RAM，Random-Access Memory）、磁碟或者光盘等各种可以存储程序代码的介质。
本领域的普通技术人员可以理解，上述各实施例是实现本申请的具体实施例，而在实际应用中，可以在形式上和细节上对其作各种改变，而不偏离本申请的精神和范围。

Claims (18)

  1. 一种智能售货柜，其中，包括：设置有至少一个放置空间的柜体，设置在每个放置空间底部的摄像头，覆盖放置空间顶部且镜面朝向放置空间底部的镜子，以及与所述摄像头通信连接的主控模块；
    所述摄像头的镜头正对所述摄像头所在放置空间内的镜子，所述摄像头的视角范围覆盖所述摄像头所在放置空间的镜子的镜面；
    所述主控模块控制所述摄像头对所述放置空间内的所述镜子进行拍摄，并将拍摄获得的图像传输至物品识别装置进行处理。
  2. 根据权利要求1所述的智能售货柜，其中，所述主控模块具体用于：
    控制所述摄像头处于不同视角范围并对所述放置空间内的所述镜子进行拍摄，获得所述摄像头在不同视角范围拍摄的图像；
    从所述摄像头在不同视角范围拍摄的图像中选取最佳图像，将所述最佳图像对应的视角范围确定为最佳视角范围，并将所述摄像头的视角调整至所述最佳视角范围，其中，所述最佳图像中包括所述摄像头所在放置空间内的所有物品，且物品在图像中的面积最大。
  3. 根据权利要求2所述的智能售货柜，其中，所述主控模块还用于：
    在控制所述摄像头处于不同视角范围时对所述放置空间内的所述镜子进行拍摄之前，接收到所述柜体内物品放置完毕的信号。
  4. 根据权利要求2或3所述的智能售货柜，其中，所述摄像头为定焦摄像头或变焦摄像头。
  5. 根据权利要求4所述的智能售货柜，其中，若所述摄像头为变焦摄像头，所述主控模块具体用于：
    控制所述摄像头在不同的焦距下对所述放置空间内的所述镜子进行拍摄，获得所述摄像头在不同焦距下拍摄的图像；
    从所述摄像头在不同焦距下拍摄的图像中选取所述最佳图像，将所述最佳图像对应的焦距确定为最佳焦距，调节所述摄像头的焦距至所述最佳焦距。
  6. 根据权利要求4所述的智能售货柜，其中，所述智能售货柜还包括设置在每个所述放置空间底部中心位置处且高度可调节的支架，所述摄像头固定于所述支架上；
    所述主控模块具体用于：
    调节所述支架的高度，并在所述摄像头处于不同视角范围时控制所述摄像头对所述放置空间内的所述镜子进行拍摄，获得所述摄像头在不同视角范围拍摄的图像；
    以及通过调节所述支架的高度将所述摄像头的视角调整至所述最佳视角范围。
  7. 根据权利要求6所述的智能售货柜，其中，所述支架中包括电机，所述主控模块具体用于通过控制电机调节所述支架的高度。
  8. 根据权利要求1至7中任一项所述的智能售货柜，其中，所述智能售货柜还包括柜门，所述柜门与所述柜体铰接；
    所述主控模块具体用于：
    在检测到柜门关闭后，控制所述摄像头在最佳视角范围对所述放置空间内的所述镜子进行拍摄，并将所述拍摄的图像传输至物品识别装置进行图像识别。
  9. 根据权利要求8所述的智能售货柜，其中，所述放置空间的任意侧面设置有限高线，所述限高线用于标识允许放置的放置物的最高高度。
  10. 根据权利要求1至9中任一项所述的智能售货柜，其中，所述镜子为平面镜、凸面镜或者凹面镜。
  11. 根据权利要求1至10中任一项所述的智能售货柜，其中，所述摄像头与所述主控模块采用有线和/或无线通信的方式连接。
  12. 根据权利要求1至11中任一项所述的智能售货柜，其中，所述放置空间由层隔板分隔所述柜体获得。
  13. 一种物品识别方法，其中，用于识别如权利要求1至12中任一项所述的智能售货柜中的物品，包括：
    接收所述智能售货柜传输的每个放置空间的图像；
    分别对每个所述放置空间的图像进行抗干扰处理；
    识别所述抗干扰处理后的图像中的物品。
  14. 根据权利要求13所述的物品识别方法，其中，所述分别对每个所述放置空间的图像进行抗干扰处理，具体包括：
    从所述每个图像中去除干扰影像，所述干扰影像包括摄像头的影像以及放置空间外的影像；
    和/或，
    判断所述图像中是否包含限高线，若是，且确定所述图像中物品的高度小于所述限高线的高度，则按照所述限高线裁切所述图像，否则，按照所述图像中镜子的边缘裁切所述图像。
  15. 根据权利要求13至14中任一项所述的物品识别方法，其中，所述识别所述抗干扰处理后的图像中的物品，具体包括：
    识别所述抗干扰处理后的图像中每个物品的种类以及在图像中的位置。
  16. 一种物品识别装置，其中，所述物品识别装置与权利要求1至12中任一项所述的智能售货柜通信连接，所述物品识别装置包括：通信模块、图像处理模块和图像识别模块；
    所述通信模块用于接收所述智能售货柜传输的每个放置空间的图像；
    所述图像处理模块用于分别对每个所述放置空间的图像进行抗干扰处理；
    所述图像识别模块用于识别所述抗干扰处理后图像中的物品。
  17. 一种服务器，包括：
    至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行如权利要求13至15任一项所述的物品识别方法。
  18. 一种计算机可读存储介质，存储有计算机程序，其中，所述计算机程序被处理器执行时实现权利要求13至15任一项所述的物品识别方法。
PCT/CN2018/087304 2018-05-17 2018-05-17 智能售货柜、物品识别方法、装置、服务器和存储介质 WO2019218303A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2018/087304 WO2019218303A1 (zh) 2018-05-17 2018-05-17 智能售货柜、物品识别方法、装置、服务器和存储介质
CN201880001294.6A CN108780505B (zh) 2018-05-17 2018-05-17 智能售货柜、物品识别方法、装置、服务器和存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/087304 WO2019218303A1 (zh) 2018-05-17 2018-05-17 智能售货柜、物品识别方法、装置、服务器和存储介质

Publications (1)

Publication Number Publication Date
WO2019218303A1 true WO2019218303A1 (zh) 2019-11-21

Family

ID=64029104

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/087304 WO2019218303A1 (zh) 2018-05-17 2018-05-17 智能售货柜、物品识别方法、装置、服务器和存储介质

Country Status (2)

Country Link
CN (1) CN108780505B (zh)
WO (1) WO2019218303A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112270324A (zh) * 2020-10-26 2021-01-26 成都平凡谷科技有限公司 一种通过采集客显数据的货品信息分析方法和系统
CN113723383A (zh) * 2021-11-03 2021-11-30 武汉星巡智能科技有限公司 异视角同步识别同区域商品的订单生成方法及智能售货机

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109191668A (zh) * 2018-08-20 2019-01-11 上海了物网络科技有限公司 基于镜面反射的物品视觉识别售货柜
CN109671227B (zh) * 2018-11-13 2024-04-26 达闼机器人股份有限公司 智能货柜消费行为识别方法、装置、存储介质和电子设备
CN109753881B (zh) * 2018-12-13 2023-05-12 小黄狗环保科技有限公司 一种防止强光干扰饮料瓶回收机识别投递物的方法
CN109509293A (zh) * 2018-12-29 2019-03-22 哈尔滨新中新电子股份有限公司 新型餐品自动售卖机
CN109686104B (zh) * 2018-12-31 2020-12-11 武汉静磁栅机电制造有限公司 一种车辆超高检测方法
CN109815896B (zh) * 2019-01-23 2022-12-27 武汉嗨柜技术有限公司 一种针对智能售货柜的参数可配置的虚拟货道划分方法
DE102019204893A1 (de) * 2019-04-05 2020-10-08 Sielaff GmbH & Co. KG Automatenbau Herrieden Warenverkaufsautomat und Verfahren
CN110211187A (zh) * 2019-04-28 2019-09-06 上海小萌科技有限公司 一种多摄像头位置标定方法
CN112990138B (zh) * 2019-11-18 2022-04-29 支付宝(杭州)信息技术有限公司 层板调控设备、方法及装置
CN111260299B (zh) * 2020-02-18 2023-07-18 中国联合网络通信集团有限公司 货品盘点及管理方法、装置、电子设备及存储介质
CN111166057A (zh) * 2020-02-18 2020-05-19 枣庄学院 图书馆运输车
CN114762010A (zh) * 2020-10-29 2022-07-15 京东方科技集团股份有限公司 监控设备、监控方法
CN114022999A (zh) * 2021-10-27 2022-02-08 北京云迹科技有限公司 一种自动售货机的缺货检测方法、装置、设备和介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3115924A1 (en) * 2015-07-09 2017-01-11 NCR Corporation Scanner including a switchable mirror
CN107909719A (zh) * 2017-12-11 2018-04-13 晋江市出奇爱食品有限公司 一种自动售货柜
CN107992820A (zh) * 2017-11-29 2018-05-04 北京伟景智能科技有限公司 基于双目视觉的货柜自助售货方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN202841321U (zh) * 2012-10-19 2013-03-27 吴平 一种利用镜像获取纸质文档图像的高速摄像扫描装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3115924A1 (en) * 2015-07-09 2017-01-11 NCR Corporation Scanner including a switchable mirror
CN107992820A (zh) * 2017-11-29 2018-05-04 北京伟景智能科技有限公司 基于双目视觉的货柜自助售货方法
CN107909719A (zh) * 2017-12-11 2018-04-13 晋江市出奇爱食品有限公司 一种自动售货柜

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112270324A (zh) * 2020-10-26 2021-01-26 成都平凡谷科技有限公司 一种通过采集客显数据的货品信息分析方法和系统
CN112270324B (zh) * 2020-10-26 2024-03-01 成都平凡谷科技有限公司 一种通过采集客显数据的货品信息分析方法和系统
CN113723383A (zh) * 2021-11-03 2021-11-30 武汉星巡智能科技有限公司 异视角同步识别同区域商品的订单生成方法及智能售货机

Also Published As

Publication number Publication date
CN108780505B (zh) 2022-11-08
CN108780505A (zh) 2018-11-09

Similar Documents

Publication Publication Date Title
WO2019218303A1 (zh) 智能售货柜、物品识别方法、装置、服务器和存储介质
CN110738142B (zh) 一种自适应改善人脸图像采集的方法、系统及存储介质
US9922238B2 (en) Apparatuses, systems, and methods for confirming identity
CN100502471C (zh) 图像处理装置、图像处理方法和成像装置
WO2020078440A1 (zh) 采集高清晰度面部图像的装置和摄像头云台自动俯仰调节的方法
WO2021027537A1 (zh) 一种拍摄证件照的方法、装置、设备及存储介质
CN110688914A (zh) 一种手势识别的方法、智能设备、存储介质和电子设备
EP2616993A1 (en) Smile detection systems and methods
KR20140013407A (ko) 객체 추적 장치 및 방법
CN111062234A (zh) 一种监控方法、智能终端及计算机可读存储介质
JP5271742B2 (ja) サングラス着用検出装置
CN109451233B (zh) 一种采集高清晰度面部图像的装置
KR20090086891A (ko) 적외선을 이용한 얼굴인식 시스템 및 방법
KR101640014B1 (ko) 허위 안면 이미지 분류가 가능한 홍채 인식 장치
WO2021204267A1 (zh) 身份识别
US9779506B2 (en) Method for processing information and electronic device
KR101754152B1 (ko) 복수 대역 카메라를 이용한 발열환자 감시 시스템 및 이를 이용한 발열환자 감시 방법
TWI509466B (zh) 物件辨識方法與裝置
TWI731503B (zh) 活體臉部辨識系統與方法
CN107844764B (zh) 图像处理方法、装置、电子设备和计算机可读存储介质
WO2021204211A1 (zh) 人脸和虹膜图像采集方法、装置、可读存储介质及设备
WO2016004819A1 (zh) 一种拍摄方法、拍摄装置和计算机存储介质
CN113591526A (zh) 人脸活体检测方法、装置、设备及计算机可读存储介质
JP2018081402A (ja) 画像処理装置、画像処理方法、及びプログラム
CN106934349B (zh) 双摄像头成像及虹膜采集识别一体化设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18919287

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 19.04.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18919287

Country of ref document: EP

Kind code of ref document: A1