An automatic vending machine recognition system based on image processing and its recognition method

Abstract

The present invention provides an automatic vending machine recognition system based on image processing and its recognition method. The recognition system includes: an authorization terminal for obtaining a high-definition face image of a buyer, performing authorization authentication on the buyer based on the high-definition face image, and obtaining an authorization authentication result; a monitoring terminal for controlling the vending machine to open a purchase window based on the authorization authentication result, and obtaining a full-process video of the buyer taking goods in the vending machine; a settlement terminal for generating a settlement order based on the goods to be settled identified in the full-process video, and performing an automatic deduction operation based on the settlement order and the authorization authentication result; thereby realizing face authentication and authorization of users through the combination of face recognition function and cloud authorization authentication, then dynamically recognizing the taking process based on the face and generating a settlement order corresponding to the buyer, improving the transaction efficiency and making the transaction process more intelligent.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and vending machine integration, and particularly relates to an automatic vending machine recognition system based on image processing and its recognition method. Background Art

[0002] Currently, the development of unmanned retail is very rapid. The technologies of traditional vending machines, intelligent vending machines, and open shared shelves are already quite mature. There are currently four common ways to identify vending machines:

[0003] Static recognition. This method requires a large number of camera devices to be set in the unmanned vending machine, which is not only costly and takes up a large amount of space, but also has a fatal technical problem. Since the camera angle is fixed, the recognition effect is poor when identifying the goods blocked during the artificial taking process.

[0004] RFID recognition. Although this method is relatively easy to design, because products need to be attached with RFID identification tags, it requires a certain amount of manpower and material resources, and the later cost is relatively high.

[0005] Gravity recognition. This method cannot distinguish goods with very similar weights. In addition, if the goods are tilted, or the vending machine is tilted, or the goods lean on each other, it will affect the gravity sensing, thus affecting the system's judgment of the goods taken by the customer.

[0006] Dynamic recognition. Although this method overcomes the setting cost and has a good recognition effect, the existing dynamic recognition of vending machines is only applied to the recognition of goods during the user's taking process. The payment transaction function of vending machines is not intelligent enough. Users need to perform self-checkout operations before each transaction, resulting in low transaction efficiency, and it cannot identify the taking behavior and the attribution of the goods order, and the transaction process is not intelligent enough.

[0007] Therefore, the present invention proposes an automatic vending machine recognition system based on image processing and its recognition method. Summary of the Invention

[0008] The present invention provides an automatic vending machine recognition system based on image processing and its recognition method, which realizes the face authentication and authorization of the corresponding user by combining the face recognition function and cloud authorization authentication. Then, based on the face dynamic recognition during the taking process, a settlement order to be settled for the corresponding buyer is generated, realizing that the buyer does not need to perform self-checkout before each transaction, improving the transaction efficiency, and also realizing the identification of the taking behavior and the attribution of the goods order, making the transaction process more intelligent.

[0009] The present invention provides an automatic vending machine recognition system based on image processing, including:

[0010] An authorization terminal, configured to obtain a high-definition face image of a buyer, perform authorization authentication on the buyer based on the high-definition face image, and obtain an authorization authentication result;

[0011] A monitoring terminal, configured to control the vending machine to open a purchase window based on the authorization authentication result, and obtain a full-process video of the buyer taking goods in the vending machine;

[0012] A settlement terminal, configured to generate a to-be-settled order based on the goods to be settled identified in the full-process video, and perform an automatic deduction operation based on the to-be-settled order and the authorization authentication result.

[0013] Preferably, the authorization terminal includes:

[0014] An instruction receiving module, configured to receive a purchase request instruction input by the buyer based on a touch display screen arranged outside the vending machine;

[0015] An image acquisition module, configured to obtain all high-definition face images within a preset range in front of the vending machine when receiving the purchase request instruction;

[0016] An authorization authentication module, configured to perform authorization authentication on the buyer based on the purchase request instruction and all high-definition face images, and obtain an authorization authentication result.

[0017] Preferably, the image acquisition module includes:

[0018] An image acquisition unit, configured to obtain all personnel images within a preset range in front of the vending machine when receiving the purchase request instruction;

[0019] A dynamic tracking unit, configured to perform dynamic tracking on the personnel images to obtain a dynamic tracking video of each personnel image;

[0020] A face recognition unit, configured to identify all high-definition face images within a preset range in front of the vending machine based on the dynamic tracking video.

[0021] Preferably, the face recognition unit includes:

[0022] A face determination subunit, configured to identify a dynamic face video of each personnel image in the dynamic tracking video, and determine a face area and a face contour in each face image frame in the dynamic face video;

[0023] A frame determination subunit, configured to determine the bone point positions corresponding to the positioning bone points included in a preset bone point list in the face area, and connect the bone point positions included in the face area based on a preset connection relationship between the positioning bone points to obtain a bone point distribution frame corresponding to the face area;

[0024] An angle determination subunit, configured to perform surrounding matching on the bone point distribution framework and a standard three-dimensional bone point distribution framework, and determine the shooting angle of the face region;

[0025] An optimal determination subunit, configured to determine whether there is more than one face region corresponding to each shooting angle. If so, obtain the face region set and the face contour set corresponding to the shooting angle, and fit the optimal face region and the optimal face contour corresponding to the shooting angle based on the face region set and the face contour set. Otherwise, use the corresponding face region as the optimal face region corresponding to the shooting angle, and at the same time, use the corresponding face contour as the optimal face contour corresponding to the shooting angle;

[0026] A region determination subunit, configured to determine a first small-amplitude change region in the optimal face region based on the shooting angle and a small-amplitude change region in the standard three-dimensional bone point distribution framework;

[0027] A facial feature determination subunit, configured to regard the remaining region in the optimal face region except the first small-amplitude change region as the facial feature region, identify the facial feature contours in the facial feature region, calculate the movement amplitude of the facial feature region based on the facial feature contours and the bone point positions included in the facial feature region, and use the facial feature region with the smallest movement amplitude as the optimal facial feature region corresponding to the shooting angle;

[0028] A model restoration subunit, configured to perform contour restoration on the standard three-dimensional bone point distribution framework based on the optimal face contour corresponding to each shooting angle, and perform local restoration on the standard three-dimensional bone point distribution framework based on the first small-amplitude change region and the corresponding optimal facial feature region corresponding to each shooting angle, to obtain a three-dimensional face model of the corresponding person's image;

[0029] An image processing subunit, configured to screen out the optimal face image corresponding to each person's image in the dynamic face video, and perform local restoration on the optimal face image based on the corresponding three-dimensional face model, to obtain a high-definition face image within a preset range in front of the vending machine.

[0030] Preferably, the authorization and authentication module includes:

[0031] A terminal determination unit, configured to determine an authorization and authentication terminal based on the purchase request instruction;

[0032] An instruction sending unit, configured to generate an authorization and authentication instruction based on all the high-definition face images and the purchase request instruction, and send the authorization and authentication instruction to the authorization and authentication terminal;

[0033] An account determination unit, configured to receive an authorization authentication feedback instruction from the authorization authentication terminal, and determine the buyer's face image and the automatic deduction account of the buyer based on the authorization authentication feedback instruction;

[0034] An authorization authentication unit, configured to perform authorization authentication on the buyer based on the automatic deduction account, and obtain an authorization authentication result in combination with the buyer's face image.

[0035] Preferably, the authorization authentication unit includes:

[0036] A balance acquisition subunit, configured to acquire the deductible balance in the automatic deduction account;

[0037] An authorization authentication subunit, configured to, when the deductible balance is not less than the authorizable balance, take the buyer's successful authorization authentication and the buyer's face image as the authorization authentication result; otherwise, take the buyer's failed authorization authentication as the authorization authentication result, and send a failed authorization authentication prompt instruction to the authorization authentication terminal.

[0038] Preferably, the monitoring end includes:

[0039] An opening control module, configured to control the purchase window of the vending machine to open when the authorization authentication result is that the buyer passes the authorization authentication;

[0040] A face acquisition module, configured to acquire the buyer's face image based on the authorization authentication result when the authorization authentication result is that the buyer passes the authorization authentication;

[0041] A video screening module, configured to acquire an internal taking video based on a camera disposed inside the vending machine, and screen out the whole-process video of the buyer taking goods in the vending machine from the internal taking video based on the buyer's face image.

[0042] Preferably, the settlement end includes:

[0043] A commodity identification module, configured to identify the goods to be settled in the current purchase process from the whole-process video;

[0044] An order commodity module, configured to generate a to-be-settled order based on the goods to be settled;

[0045] An automatic settlement module, configured to obtain the deduction permission for the corresponding buyer's account based on the authorization authentication result, and perform an automatic deduction operation on the buyer's account based on the deduction permission and the to-be-settled order.

[0046] Preferably, the commodity identification module includes:

[0047] An area tracking unit, configured to determine a difference area between adjacent video frames in the whole video, perform video tracking on the difference area based on the whole video, and obtain a difference area sequence;

[0048] A weight determination unit, configured to perform frame alignment processing on the difference area sequence and the whole video to obtain a corresponding alignment result, determine a frame interval between the last frame of the difference area sequence and the last frame of the whole video based on the alignment result, and determine a screening weight of the corresponding difference area based on the total number of frames of the difference area sequence and the frame interval;

[0049] A contour restoration unit, configured to use the difference area sequence with a screening weight not less than a weight threshold as a corresponding suspected commodity area sequence, extract edges of the suspected commodity areas in the suspected commodity area sequence to obtain a corresponding set of suspected commodity edges, and restore a corresponding three-dimensional suspected commodity contour based on the set of suspected commodity edges;

[0050] A sequence discrimination unit, configured to calculate a matching degree between the suspected commodity contour and each commodity contour included in a preset commodity contour list, determine whether there is a commodity contour in the preset commodity contour list whose matching degree with the suspected commodity contour is greater than a matching degree threshold. If so, use the corresponding suspected commodity area sequence as a corresponding sequence to be recognized. Otherwise, determine that the suspected commodity area sequence is not a sequence to be recognized;

[0051] A matrix determination unit, configured to determine a corresponding gray threshold based on the gray distribution data corresponding to each area to be recognized in the sequence to be recognized, perform binary processing on the corresponding area to be recognized based on the gray threshold to obtain a binary area sequence, divide each binary area in the binary area sequence into multiple sub-areas to be recognized based on a preset division method, determine a texture gradient characterization value of the sub-areas to be recognized, and determine a texture gradient feature matrix of the binary area based on the texture gradient characterization values of each sub-areas to be recognized;

[0052] A matrix splicing unit, configured to sort the texture gradient feature matrices of all binary areas based on the frame sequence of the binary area sequence to obtain a corresponding texture gradient feature matrix sequence, determine an overlapping part between adjacent texture gradient feature matrices in the texture gradient feature matrix sequence, and splice all texture gradient feature matrices in the texture gradient feature matrix sequence based on the overlapping part and the frame sequence to obtain a corresponding comprehensive texture gradient matrix;

[0053] A commodity determination unit, configured to use the commodity corresponding to the standard texture gradient matrix with the largest similarity to the comprehensive texture gradient matrix as the commodity to be settled.

[0054] The present invention provides an identification method for a vending machine based on image processing, including:

[0055] S1: Obtain a high-definition face image of the buyer, perform authorization authentication on the buyer based on the high-definition face image, and obtain an authorization authentication result;

[0056] S2: Control the vending machine to open the purchase window based on the authorization authentication result, and obtain the full-process video of the buyer picking up goods in the vending machine;

[0057] S3: Generate a settlement order for the goods to be settled identified in the full-process video, and perform an automatic deduction operation based on the settlement order and the authorization authentication result.

[0058] Other features and advantages of the present invention will be described in the subsequent specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification, claims, and drawings.

[0059] The technical solutions of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0060] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0061] Figure 1 is a schematic diagram of an identification system for a vending machine based on image processing in an embodiment of the present invention;

[0062] Figure 2 is a schematic diagram of an authorization terminal in an embodiment of the present invention;

[0063] Figure 3 is a schematic diagram of an image acquisition module in an embodiment of the present invention;

[0064] Figure 4 is a schematic diagram of a face recognition unit in an embodiment of the present invention;

[0065] Figure 5 is a schematic diagram of an authorization authentication module in an embodiment of the present invention;

[0066] Figure 6 is a schematic diagram of an authorization authentication unit in an embodiment of the present invention;

[0067] Figure 7 is a schematic diagram of a monitoring terminal in an embodiment of the present invention;

[0068] Figure 8 Schematic diagram of a settlement terminal in an embodiment of the present invention;

[0069] Figure 9 Schematic diagram of a commodity recognition module in an embodiment of the present invention;

[0070] Figure 10 Flowchart of an automatic vending machine recognition method based on image processing in an embodiment of the present invention. Specific embodiments

[0071] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not used to limit the present invention.

[0072] Embodiment 1:

[0073] The present invention provides an automatic vending machine recognition system based on image processing. Refer to Figure 1 , including:

[0074] An authorization terminal, configured to obtain a high-definition face image of a buyer, perform authorization authentication on the buyer based on the high-definition face image, and obtain an authorization authentication result;

[0075] A monitoring terminal, configured to control the vending machine to open a purchase window based on the authorization authentication result, and obtain a full-process video of the buyer taking goods in the vending machine;

[0076] A settlement terminal, configured to generate a settlement order based on the goods to be settled identified in the full-process video, and perform an automatic deduction operation based on the settlement order and the authorization authentication result.

[0077] In this embodiment, the high-definition face image is the high-definition face image of the buyer obtained based on a camera set outside the vending machine.

[0078] In this embodiment, the authorization authentication result is the result obtained after performing authorization authentication on the buyer based on the high-definition face image, including the result of whether the buyer passes the authorization authentication and the high-definition face image of the buyer.

[0079] In this embodiment, controlling the vending machine to open a purchase window based on the authorization authentication result means: when the authorization authentication result is that the buyer passes the authorization authentication, controlling the vending machine to open the purchase window.

[0080] In this embodiment, the purchase window is the window for taking goods in the vending machine.

[0081] In this embodiment, the full-process video is the video monitoring the whole process of the buyer taking goods in the vending machine.

[0082] In this embodiment, the goods to be settled are the goods taken by the buyer identified in the whole-process video.

[0083] In this embodiment, the order to be settled is the order that needs to be settled by the buyer generated based on the goods to be settled.

[0084] In this embodiment, the automatic deduction operation is the deduction operation automatically executed in the buyer's account based on the order to be dissolved and the authorization authentication result.

[0085] The beneficial effects of the above technology are as follows: By combining the face recognition function and cloud authorization authentication, the face authentication and authorization of the corresponding user are realized. Then, based on the face dynamic tracking of the taking process and the recognition of the taken goods, the order to be settled for the corresponding buyer is generated. It realizes that the buyer does not need to perform self-service settlement between each transaction, improves the transaction efficiency, and also realizes the recognition of the taking behavior and the attribution of the goods order, making the transaction process more intelligent.

[0086] Embodiment 2:

[0087] Based on Embodiment 1, the authorization terminal, referring to Figure 2 , includes:

[0088] An instruction receiving module, configured to receive the purchase request instruction input by the buyer based on the touch display screen arranged outside the vending machine;

[0089] An image acquisition module, configured to acquire all high-definition face images within a preset range in front of the vending machine when receiving the purchase request instruction;

[0090] An authorization authentication module, configured to perform authorization authentication on the buyer based on the purchase request instruction and all high-definition face images to obtain an authorization authentication result.

[0091] In this embodiment, the touch display screen is a touch screen device arranged outside the vending machine and used to receive the purchase request instruction input by the buyer.

[0092] In this embodiment, the purchase request instruction is the instruction input by the buyer on the touch display screen for requesting to purchase goods from the vending machine.

[0093] In this embodiment, the preset range is the preset shooting range of the camera arranged outside the vending machine (specifically set according to the actual situation).

[0094] The beneficial effects of the above technology are as follows: all high-definition face images in front of the vending machine are obtained based on the purchase request instruction input by the user, realizing personalized authorization authentication of the user based on the face images, providing a basis for determining the order attribution based on the face images and the whole-process video, and also realizing that except for the first time, there is no need for the buyer to perform authentication operations or self-checkout operations every time before purchasing, which not only improves the transaction efficiency but also improves the intelligence level of the transaction process.

[0095] Embodiment 3:

[0096] Based on Embodiment 2, the image acquisition module, with reference to Figure 3 , includes:

[0097] An image acquisition unit, configured to acquire all personnel images within a preset range in front of the vending machine when receiving the purchase request instruction;

[0098] A dynamic tracking unit, configured to perform dynamic tracking on the personnel images to obtain a dynamic tracking video of each personnel image;

[0099] A face recognition unit, configured to recognize all high-definition face images within a preset range in front of the vending machine based on the dynamic tracking video.

[0100] In this embodiment, the personnel image is the image of all human bodies within a preset range in front of the vending machine.

[0101] In this embodiment, the dynamic tracking video is the tracking video corresponding to the personnel image obtained by dynamically tracking the personnel image based on a camera set outside the vending machine.

[0102] The beneficial effects of the above technology are as follows: when receiving the purchase request instruction, all personnel images in front of the vending machine are acquired and dynamically tracked, providing an image basis for obtaining the high-definition face images of the buyer.

[0103] Embodiment 4:

[0104] Based on Embodiment 3, the face recognition unit, with reference to Figure 4 , includes:

[0105] A face determination subunit, configured to recognize the dynamic face video of each personnel image in the dynamic tracking video, and determine the face area and face contour in each face image frame in the dynamic face video;

[0106] A framework determination subunit, configured to determine the bone point positions corresponding to the positioning bone points included in a preset bone point list in the face region, and connect the bone point positions included in the face region based on a preset connection relationship between the positioning bone points, so as to obtain a bone point distribution framework corresponding to the face region;

[0107] An angle determination subunit, configured to perform surrounding matching between the bone point distribution framework and a standard bone point distribution three-dimensional framework to determine the shooting angle of the face region;

[0108] An optimal determination subunit, configured to determine whether there is more than one face region corresponding to each shooting angle. If so, obtain a face region set and a face contour set corresponding to the shooting angle, and fit an optimal face region and an optimal face contour corresponding to the shooting angle based on the face region set and the face contour set. Otherwise, use the corresponding face region as the optimal face region corresponding to the shooting angle, and at the same time, use the corresponding face contour as the optimal face contour corresponding to the shooting angle;

[0109] A region determination subunit, configured to determine a first small-variation region in the optimal face region based on the shooting angle and a small-variation region with a smaller amplitude in the standard bone point distribution three-dimensional framework;

[0110] A facial feature determination subunit, configured to regard the remaining region in the optimal face region except the first small-variation region as a facial feature region, identify the facial feature contours in the facial feature region, calculate the motion amplitude of the facial feature region based on the facial feature contours and the bone point positions included in the facial feature region, and use the facial feature region with the smallest motion amplitude as the optimal facial feature region corresponding to the shooting angle;

[0111] A model restoration subunit, configured to perform contour restoration on the standard bone point distribution three-dimensional framework based on the optimal face contour corresponding to each shooting angle, and perform local restoration on the standard bone point distribution three-dimensional framework based on the first small-variation region and the corresponding optimal facial feature region corresponding to each shooting angle, so as to obtain a three-dimensional face model of the corresponding person's image;

[0112] An image processing subunit, configured to screen out the optimal face image corresponding to each person's image in the dynamic face video, and perform local restoration on the optimal face image based on the corresponding three-dimensional face model, so as to obtain a high-definition face image within a preset range in front of the vending machine.

[0113] In this embodiment, the dynamic face video is the face video of each person's image identified in the dynamic tracking video.

[0114] In this embodiment, the face region is the image region of the face determined in each face image frame of the dynamic face video.

[0115] In this embodiment, the facial contour is the regional contour of the face determined in each facial image frame of the dynamic facial video.

[0116] In this embodiment, the bone point distribution framework is a framework obtained by connecting the positions of the bone points included in the facial region, which represents the bone point distribution in the facial region.

[0117] In this embodiment, the preset bone point list is a list containing the positioning bone points, such as the eyebrow arch, nasal bone, cheekbone, etc.

[0118] In this embodiment, the bone point position is the position of the positioning bone point in the facial region.

[0119] In this embodiment, the preset connection relationship is the preset connection relationship between the positioning bone points, such as the connection relationship between two eyebrow arches, and the connection relationship between the eyebrow arch and the nasal bone.

[0120] In this embodiment, surrounding matching the bone point distribution framework with the standard three-dimensional bone point distribution framework to determine the shooting angle of the facial region is as follows:

[0121] Performing multi-angle matching between the bone point distribution framework and the standard three-dimensional bone point distribution framework to determine the rotation angle between the position of the bone point distribution framework in the standard three-dimensional bone point distribution framework and the facial center position, and taking the rotation angle as the shooting angle of the facial region.

[0122] In this embodiment, the standard three-dimensional bone point distribution framework is a three-dimensional facial bone point framework containing all the positioning bone points in the preset bone point list.

[0123] In this embodiment, fitting the best facial region and the best facial contour corresponding to the shooting angle based on the facial region set and the facial contour set includes:

[0124] Calculating the first visual effect value of each facial region in the facial region set, which is:

[0125] The average value of the gray value, chromaticity value, and contrast value of all pixel points in the facial region is used as the first visual effect value of the facial region.

[0126] Taking the facial region corresponding to the maximum first visual effect value as the best facial region corresponding to the shooting angle;

[0127] Calculating the second visual effect value of each facial contour in the facial contour set, which is:

[0128] The average value of the gray value, chromaticity value, and contrast value of all pixel points in the facial contour is used as the second visual effect value of the facial contour.

[0129] Take the facial contour corresponding to the maximum second visual effect value as the best facial contour for the corresponding shooting angle.

[0130] In this embodiment, the set of facial regions is the set composed of facial regions at the same shooting angle.

[0131] In this embodiment, the set of facial contours is the set composed of facial contours at the same shooting angle.

[0132] In this embodiment, the best facial region is the facial region with the best visual effect for the determined corresponding shooting angle.

[0133] In this embodiment, the best facial contour is the facial contour with the best visual effect for the determined corresponding shooting angle.

[0134] In this embodiment, the regions with small amplitude variation, such as: forehead region, chin region, temple region, cheek region.

[0135] In this embodiment, the first region with small amplitude variation is the corresponding region in the best facial region of the region with small amplitude variation in the standard bone point distribution three-dimensional framework.

[0136] In this embodiment, based on the facial feature contours and the bone point positions included in the facial feature regions, calculating the movement amplitude of the facial feature regions includes:

[0137] Divide the facial feature region into an eye region, a nose region, and a mouth region, and regard the eye region, the nose region, and the mouth region as sub-facial feature regions;

[0138] Take the coordinate value mean of all contour pixels of each sub-facial feature region as the reference position of the sub-facial feature region;

[0139] Take the bone point position closest to the reference position as the reference bone point position of the sub-facial feature region;

[0140] Take the vector from the reference position to the reference bone point position as the reference vector, and determine the pointing vector from each pixel point in the sub-facial feature region to the reference bone point position;

[0141] Based on the reference vectors and all pointing vectors in all sub-facial feature regions, determine the movement amplitude of the facial feature region:

[0142]

[0143] In the formula, α is the movement amplitude of the facial feature region, j is the jth sub-facial feature region, n is the total number of sub-facial feature regions, k is the kth pointing vector in the sub-facial feature region, b is the total number of pointing vectors included in the sub-facial feature region, is the reference vector of the jth sub-facial feature region, is the average vector of the reference vectors for all sub - facial feature regions, is the absolute value of, is the absolute value of, is the k - th pointing vector in the j - th sub - facial feature region, is the absolute value of;

[0144] For example, the facial feature region contains two sub - facial feature regions. The pointing vectors in the first sub - facial feature region are (1, 1) and (3, 3) in sequence, and the reference vector is (2, 2). The pointing vectors in the second sub - facial feature region are (4, 4) and (8, 8) in sequence, and the reference vector is (6, 6). is (4, 4), then α is 0.5.

[0145] In this embodiment, the facial feature region is the remaining region of the best facial region except for the first region with a smaller amplitude change.

[0146] In this embodiment, the facial feature contour is the contour of the facial feature region.

[0147] In this embodiment, the best facial feature region is the facial feature region corresponding to the minimum movement amplitude.

[0148] In this embodiment, the three - dimensional facial model is a facial model of the corresponding person's image obtained by restoring the contour of the standard bone point distribution three - dimensional framework based on the best facial contour corresponding to each shooting angle, and locally restoring the standard bone point distribution three - dimensional framework based on the first region with a smaller amplitude change and the corresponding best facial feature region corresponding to each shooting angle.

[0149] In this embodiment, locally restoring the best facial image based on the corresponding three - dimensional facial model to obtain the high - definition face image within the preset range in front of the vending machine means:

[0150] Locally restoring the best facial image based on the local chromaticity value, brightness value, and contrast in the three - dimensional facial model to obtain the high - definition face image within the preset range in front of the vending machine.

[0151] In this embodiment, the best facial image is the facial image with the best visual effect, that is, the facial image with the largest average value of the gray - scale values, chromaticity values, and brightness values of all pixel points it contains.

[0152] The beneficial effects of the above technology are as follows: By determining the positions of the positioning bone points in the facial area and the connection relationships between the bone points, a bone point distribution framework is constructed. By performing a surrounding match between the bone point distribution framework and the standard three-dimensional bone point distribution framework, the shooting angle of the facial area can be determined, thereby providing an important restoration standard for restoring a high-definition face image. Then, the optimal facial area, the optimal facial contour, and the first area with a relatively small amplitude change corresponding to each shooting angle are determined. Based on the bone points and the facial feature contours in the facial feature area of the optimal facial area, the movement amplitudes of the facial features are determined. The facial feature area corresponding to the minimum movement amplitude is used as the optimal facial feature area. Based on the determined optimal facial contour, the first area with a relatively small amplitude change, and the facial feature area corresponding to each shooting angle, and in combination with the standard three-dimensional bone point distribution framework, the restoration and construction of the facial model of the corresponding person are realized, thereby providing an important basis for subsequent local restoration on the basis of the optimal facial image, and achieving higher-definition facial image extraction and correction.

[0153] Embodiment 5:

[0154] Based on Embodiment 2, the authorization and authentication module refers to Figure 5 , and includes:

[0155] A terminal determination unit, configured to determine an authorization and authentication terminal based on the purchase request instruction;

[0156] An instruction sending unit, configured to generate an authorization and authentication instruction based on all high-definition face images and the purchase request instruction, and send the authorization and authentication instruction to the authorization and authentication terminal;

[0157] An account determination unit, configured to receive an authorization and authentication feedback instruction from the authorization and authentication terminal, and determine the buyer's facial image and the automatic deduction account of the buyer based on the authorization and authentication feedback instruction;

[0158] An authorization and authentication unit, configured to perform authorization and authentication on the buyer based on the automatic deduction account, and obtain an authorization and authentication result in combination with the buyer's facial image.

[0159] In this embodiment, the authorization and authentication terminal is a communication terminal device through which the buyer can receive the authorization and authentication instruction, such as: a mobile phone.

[0160] In this embodiment, the authorization and authentication terminal is determined based on the purchase request instruction. For example, the buyer's Alipay account, WeChat account, or mobile phone account is input on the touch display screen, a purchase request instruction is generated based on the buyer's Alipay account, WeChat account, or mobile phone number, and the authorization and authentication terminal (i.e., the mobile phone logging in to the buyer's Alipay account or WeChat account or the terminal device to which the mobile phone number belongs) is determined based on the purchase request instruction.

[0161] In this embodiment, the authorization and authentication instruction is an instruction generated by combining a high-definition face image and a purchase request instruction for authorizing and authenticating the buyer.

[0162] In this embodiment, the authorization and authentication feedback instruction is the feedback instruction input by the user at the authorization and authentication terminal, including: the high-definition face image dynamically tracked during this pick-up process selected by the user from all the high-definition face images included in the authorization and authentication instruction, the feedback information agreeing to execute the authorization and authentication, and the automatic deduction account.

[0163] In this embodiment, the buyer's face image is the high-definition face image dynamically tracked during this pick-up process selected by the user from all the high-definition face images included in the authorization and authentication instruction.

[0164] In this embodiment, the automatic deduction account is the account selected by the user for automatic deduction by the vending machine in the authorization and authentication feedback instruction.

[0165] The beneficial effects of the above technology are as follows: Based on the purchase request instruction, the corresponding authorization and authentication terminal is determined, and then the authorization and authentication instruction generated based on the purchase request instruction and all the high-definition face images obtained in front of the vending machine is sent to the authorization and authentication terminal, and the authorization and authentication feedback instruction of the user is received. This not only realizes the authentication of the buyer's identity, but also obtains the automatic deduction account and the automatic deduction permission, and further determines the face image to be tracked during the pick-up process based on the feedback instruction of the buyer, improving the accuracy, security and convenience of the transaction.

[0166] Embodiment 6:

[0167] Based on Embodiment 5, the authorization and authentication unit refers to Figure 6 , and includes:

[0168] A balance acquisition subunit for acquiring the deductible balance in the automatic deduction account;

[0169] An authorization and authentication subunit for, when the deductible balance is not less than the authorizable balance, taking the buyer's successful authorization and authentication and the buyer's face image as the authorization and authentication result; otherwise, taking the buyer's failed authorization and authentication as the authorization and authentication result, and sending a failed authorization and authentication prompt instruction to the authorization and authentication terminal.

[0170] In this embodiment, the deductible balance is the balance in the automatic deduction account that can be automatically deducted by the vending machine.

[0171] In this embodiment, the authorizable balance is the minimum deductible balance when determining that the buyer has passed the authorization and authentication.

[0172] The beneficial effects of the above technology are as follows: authorizing and authenticating McGee based on the deductible balance in the automatic deduction account ensures the sufficiency of the balance for automatic settlement and avoids the situation of deduction failure due to insufficient account balance, improving the intelligence of the transaction process of the vending machine.

[0173] Embodiment 7:

[0174] Based on Embodiment 1, the monitoring end refers to Figure 7 , and includes:

[0175] An opening control module, configured to control the vending machine to open the purchase window when the authorization authentication result is that the buyer passes the authorization authentication;

[0176] A face acquisition module, configured to acquire the buyer's face image of the buyer based on the authorization authentication result when the authorization authentication result is that the buyer passes the authorization authentication;

[0177] A video screening module, configured to acquire an internal taking video based on a camera disposed inside the vending machine, and screen out a full-process video of the buyer taking goods in the vending machine from the internal taking video based on the buyer's face image.

[0178] In this embodiment, the internal taking video is a monitoring video of all personnel taking goods inside the vending machine acquired based on a camera disposed inside the vending machine.

[0179] The beneficial effects of the above technology are as follows: when the authorization authentication is passed, the purchase window is opened for the buyer, avoiding the situation where goods are taken but cannot be deducted and settled, ensuring the safety of the unattended vending process of the vending machine, and extracting a full-process video of the buyer taking goods in the vending machine from the internal taking video based on the buyer's face image, providing a key reference basis for subsequent determination of the attribution of the goods to be settled and the order to be settled.

[0180] Embodiment 8:

[0181] Based on Embodiment 1, the settlement end refers to Figure 8 , and includes:

[0182] A commodity recognition module, configured to recognize the goods to be settled in the current purchase process from the full-process video;

[0183] An order commodity module, configured to generate a settlement order based on the goods to be settled;

[0184] An automatic settlement module, configured to obtain the deduction authority of the corresponding buyer's account based on the authorization authentication result, and perform an automatic deduction operation on the buyer's account based on the deduction authority and the settlement order.

[0185] In this embodiment, the deduction permission is the permission for the vending machine to perform an automatic deduction operation on the buyer's account based on the authorization authentication result.

[0186] The beneficial effects of the above technology are as follows: It realizes the dynamic recognition of the taken goods during the taking process, also realizes the automatic generation of orders and the accurate determination of ownership, and through obtaining the deduction permission, it realizes the automatic settlement based on the face video without the need for the buyer to perform a settlement operation, which not only improves the transaction efficiency but also improves the degree of intelligence in the transaction process.

[0187] Embodiment 9:

[0188] Based on Embodiment 8, the commodity recognition module refers to Figure 9 , and includes:

[0189] An area tracking unit, configured to determine the difference area between adjacent video frames in the whole-process video, perform video tracking on the difference area based on the whole-process video, and obtain a difference area sequence;

[0190] A weight determination unit, configured to perform frame alignment processing on the difference area sequence and the whole-process video to obtain a corresponding alignment result, determine the frame interval between the last frame of the difference area sequence and the last frame of the whole-process video based on the alignment result, and determine the screening weight of the corresponding difference area based on the total number of frames of the difference area sequence and the frame interval;

[0191] A contour restoration unit, configured to use the difference area sequence with a screening weight not less than the weight threshold as the corresponding suspected commodity area sequence, perform edge extraction on the suspected commodity areas in the suspected commodity area sequence to obtain a corresponding set of suspected commodity edges, and restore the corresponding three-dimensional suspected commodity contour based on the set of suspected commodity edges;

[0192] A sequence discrimination unit, configured to calculate the matching degree between the suspected commodity contour and each commodity contour included in the preset commodity contour list, determine whether there is a commodity contour in the preset commodity contour list whose matching degree with the suspected commodity contour is greater than the matching degree threshold. If so, use the corresponding suspected commodity area sequence as the corresponding sequence to be recognized; otherwise, determine that the suspected commodity area sequence is not a sequence to be recognized;

[0193] A matrix determination unit, configured to determine a corresponding grayscale threshold based on the grayscale distribution data corresponding to each region to be recognized in the sequence to be recognized, perform binarization processing on the corresponding region to be recognized based on the grayscale threshold to obtain a binarized region sequence, divide each binarized region in the binarized region sequence into multiple sub-regions to be recognized based on a preset division method, determine the texture gradient characterization value of the sub-regions to be recognized, and determine the texture gradient feature matrix of the binarized region based on the texture gradient characterization value of each sub-region to be recognized;

[0194] A matrix splicing unit, configured to sort the texture gradient feature matrices of all binarized regions based on the frame sequence of the binarized region sequence to obtain a corresponding texture gradient feature matrix sequence, determine the overlapping part between adjacent texture gradient feature matrices in the texture gradient feature matrix sequence, and splice all texture gradient feature matrices in the texture gradient feature matrix sequence based on the overlapping part and the frame sequence to obtain a corresponding comprehensive texture gradient matrix;

[0195] A commodity determination unit, configured to use the commodity corresponding to the standard texture gradient matrix with the largest similarity to the comprehensive texture gradient matrix as the commodity to be settled.

[0196] In this embodiment, the difference region sequence is the region sequence obtained by performing video tracking on the difference regions based on the entire video.

[0197] In this embodiment, the difference region is the different region between adjacent video frames in the entire video.

[0198] In this embodiment, the alignment result is the result obtained by performing frame alignment processing on the difference region sequence and the entire video.

[0199] In this embodiment, the frame interval is the number of frames included between the last frame of the difference region sequence (i.e., the last frame in the difference region sequence) and the last frame of the entire video (i.e., the last frame in the entire video) in the alignment result.

[0200] In this embodiment, the screening weight is the ratio between the frame interval of the corresponding difference region and the total number of frames (the larger the screening weight, the more forward the difference region is in the entire video, and the smaller the total number of frames. The normal taking process should be that the total number of frames is large and continues to the last few frames in the entire video. Therefore, the screening weight represents the possibility that the difference region may be a commodity region, and the larger the screening weight, the greater the possibility that the difference region is a commodity region).

[0201] In this embodiment, the suspected commodity three-dimensional contour is the three-dimensional contour of the suspected commodity restored based on the set of suspected commodity edges.

[0202] In this embodiment, the weight threshold is the minimum screening weight corresponding to the differential region sequence regarded as the suspected commodity region sequence.

[0203] In this embodiment, the set of suspected commodity edges is a set composed of the suspected commodity edges obtained by edge extraction of the suspected commodity regions in the suspected commodity region sequence.

[0204] In this embodiment, calculating the matching degree between the suspected commodity contour and each commodity contour included in the preset commodity contour list includes:

[0205] Determining the coordinate values of each pixel point in the suspected commodity contour and the coordinate values of each pixel point in each commodity contour included in the preset commodity contour list;

[0206] Taking the ratio of the difference between the coordinate value of each pixel point in the suspected commodity contour and the coordinate value of the corresponding pixel point in the corresponding commodity contour to the coordinate value of the corresponding pixel point in the commodity contour as the difference value of the corresponding pixel point, taking the average value of the difference values of all pixel points in the suspected commodity contour as the difference degree between the suspected commodity contour and the commodity contour, and taking the difference between 1 and the difference degree as the matching degree between the suspected commodity contour and the commodity contour.

[0207] In this embodiment, the sequence to be recognized is the suspected commodity region sequence corresponding to the suspected commodity contour whose matching degree with the commodity contours in the preset commodity contour list is greater than the matching degree threshold.

[0208] In this embodiment, the matching degree threshold is the minimum matching degree corresponding to the suspected commodity region sequence corresponding to the suspected commodity contour being regarded as the sequence to be recognized.

[0209] In this embodiment, determining the corresponding gray threshold based on the gray distribution data corresponding to each region to be recognized in the sequence to be recognized includes:

[0210]

[0211] In the formula, H is the gray threshold, i is the i-th pixel point in the region to be recognized, and h i is the gray value of the i-th pixel point in the region to be recognized, and n is the total number of pixel points included in the region to be recognized;

[0212] For example, if there are three pixel points included in the region to be recognized, and the gray values are 10, 20, and 30 in sequence, then H is 11.

[0213] In this embodiment, the binary region sequence is the region sequence obtained by performing binary processing on the corresponding region to be recognized based on the gray threshold.

[0214] In this embodiment, the sub-regions to be recognized are the multiple sub-regions obtained by dividing each binary region in the binary region sequence based on a preset division method.

[0215] In this embodiment, determining the texture gradient characterization value of the sub-regions to be recognized is:

[0216] Calculating the average of the interval distances between the pixels with a gray value of 255 and all adjacent pixels, and taking the average of the average interval distances corresponding to the pixels with a gray value of 255 in the sub-regions to be recognized as the texture gradient characterization value of the sub-regions to be recognized.

[0217] In this embodiment, the texture gradient feature matrix is a matrix constructed based on the texture gradient characterization value of each sub-region to be recognized and the position of each sub-region to be recognized in the binary region.

[0218] In this embodiment, the gray-scale distribution data is the data containing the gray values of each pixel in the region to be recognized.

[0219] In this embodiment, the region to be recognized is the image contained in the sequence to be recognized.

[0220] In this embodiment, the preset division method is, for example: dividing the binary region evenly into sub-regions of M rows and N columns (where the values of M and N are set according to specific situations).

[0221] In this embodiment, the texture gradient feature matrix sequence is a matrix sequence obtained by sorting the texture gradient feature matrices of all binary regions based on the frame sequence of the binary region sequence.

[0222] In this embodiment, determining the overlapping part between adjacent texture gradient feature matrices in the texture gradient feature matrix sequence is: the part with the same values between adjacent texture gradient feature matrices.

[0223] In this embodiment, the comprehensive texture gradient matrix is a matrix obtained by splicing all the texture gradient feature matrices in the texture gradient feature matrix sequence based on the overlapping part and the frame sequence.

[0224] In this embodiment, taking the commodity corresponding to the standard texture gradient matrix with the highest similarity to the comprehensive texture gradient matrix as the commodity to be settled includes:

[0225] Calculating the similarity between each comprehensive texture gradient matrix and the standard texture gradient matrix, which is:

[0226] The ratio of the total number of values in the comprehensive texture gradient matrix that are consistent with the values and positions (rows and columns where the values are located) in the standard texture gradient matrix to the total number of values in the standard texture gradient matrix is used as the similarity between the comprehensive texture gradient matrix and the standard texture gradient matrix;

[0227] The product corresponding to the standard texture gradient matrix with the maximum similarity is used as the product to be settled.

[0228] In this embodiment, the standard texture gradient matrix is the standard comprehensive texture gradient matrix corresponding to each product.

[0229] The beneficial effects of the above technology are as follows: Determine the difference regions between adjacent video frames in the entire video, obtain the difference region sequence, align the difference region sequence with the entire video to obtain the corresponding result, determine the screening weight based on the frame interval between the last frame of the difference region sequence and the last frame of the entire video and the total number of frames of the difference region sequence, screen out the difference region sequence with a higher possibility of being a product region sequence based on the screening weight as the suspected product region sequence, and further screen the suspected product region sequence based on the matching of the contour of the suspected product region sequence and the product contours in the preset product list. Based on the matching degree, the difference region sequence is gradually screened and the product to be settled is gradually determined. Then, based on the binarization processing of the latest screened sequence to be recognized and the determination of the texture gradient characterization value, the texture gradient feature matrix corresponding to the binarized region is determined, and then the texture gradient feature matrix is spliced based on the binarized region sequence to obtain the matrix representing the texture gradient features of the object corresponding to the sequence to be recognized, providing important comparison data for accurately identifying the product to be settled. Based on the similarity comparison between the spliced comprehensive texture gradient matrix and the standard texture gradient matrices corresponding to different products, the final product to be settled is determined, realizing the accurate determination of the product to be settled based on frame number comparison, contour matching, and texture gradient feature comparison, with higher recognition accuracy and smaller error than traditional dynamic recognition algorithms.

[0230] Embodiment 10:

[0231] The present invention provides an automatic vending machine recognition method based on image processing, referring to Figure 10 , including:

[0232] S1: Obtain a high-definition face image of the buyer, perform authorization authentication on the buyer based on the high-definition face image, and obtain an authorization authentication result;

[0233] S2: Control the vending machine to open the purchase window based on the authorization authentication result, and obtain the entire video of the buyer taking products in the vending machine;

[0234] S3: Generate a settlement pending order based on the goods to be settled identified in the full-course video, and perform an automatic deduction operation based on the settlement pending order and the authorization authentication result.

[0235] The beneficial effects of the above technology are as follows: By combining the face recognition function with cloud authorization authentication, the face authentication and authorization of the corresponding user are realized. Then, based on the face dynamic tracking of the taking process and the identification of the taken goods, a settlement pending order for the corresponding buyer is generated, which realizes that the buyer does not need to perform self-settlement between each transaction, improves the transaction efficiency, and also realizes the identification of the taking behavior and the attribution of the goods order, making the transaction process more intelligent.

[0236] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. An automatic vending machine recognition system based on image processing, characterized in that, Including: An authorization end, which is used to obtain the high-definition face image of the buyer, perform authorization authentication on the buyer based on the high-definition face image, and obtain an authorization authentication result; A monitoring end, which is used to control the vending machine to open the purchase window based on the authorization authentication result, and obtain the whole-process video of the buyer taking goods in the vending machine; A settlement end, which is used to generate a to-be-settled order based on the goods to be settled identified in the whole-process video, and perform an automatic deduction operation based on the to-be-settled order and the authorization authentication result, including: A goods recognition module, which is used to identify the goods to be settled in the current purchase process in the whole-process video, including: A region tracking unit, which is used to determine the difference region between adjacent video frames in the whole-process video, perform video tracking on the difference region based on the whole-process video, and obtain a difference region sequence; A weight determination unit, which is used to perform frame alignment processing on the difference region sequence and the whole-process video to obtain a corresponding alignment result, determine the frame interval between the last frame of the difference region sequence and the last frame of the whole-process video based on the alignment result, and determine the screening weight of the corresponding difference region based on the total number of frames of the difference region sequence and the frame interval; A contour restoration unit, which is used to use the difference region sequence with a screening weight not less than the weight threshold as the corresponding suspected goods region sequence, extract the edges of the suspected goods regions in the suspected goods region sequence to obtain a corresponding set of suspected goods edges, and restore the corresponding three-dimensional contour of the suspected goods based on the set of suspected goods edges; A sequence discrimination unit, which is used to calculate the matching degree between the suspected goods contour and each goods contour included in the preset goods contour list, determine whether there is a goods contour in the preset goods contour list whose matching degree with the suspected goods contour is greater than the matching degree threshold, and if so, use the corresponding suspected goods region sequence as the corresponding to-be-identified sequence, otherwise, determine that the suspected goods region sequence is not a to-be-identified sequence; A matrix determination unit, which is used to determine the corresponding gray threshold based on the gray distribution data corresponding to each to-be-identified region in the to-be-identified sequence, perform binaryzation processing on the corresponding to-be-identified region based on the gray threshold to obtain a binaryzation region sequence, divide each binaryzation region in the binaryzation region sequence into multiple to-be-identified sub-regions based on a preset division method, determine the texture gradient characterization value of the to-be-identified sub-regions, and determine the texture gradient feature matrix of the binaryzation region based on the texture gradient characterization value of each to-be-identified sub-region; A matrix splicing unit, which is used to sort the texture gradient feature matrices of all binaryzation regions based on the frame sequence of the binaryzation region sequence to obtain a corresponding texture gradient feature matrix sequence, determine the overlapping part between adjacent texture gradient feature matrices in the texture gradient feature matrix sequence, and splice all the texture gradient feature matrices in the texture gradient feature matrix sequence based on the overlapping part and the frame sequence to obtain a corresponding comprehensive texture gradient matrix; A product determination unit, configured to use the product corresponding to the standard texture gradient matrix with the highest similarity to the comprehensive texture gradient matrix as the product to be settled; An order product module, configured to generate a to-be-settled order based on the product to be settled; An automatic settlement module, configured to obtain the deduction authority for the corresponding buyer's account based on the authorization authentication result, and perform an automatic deduction operation on the buyer's account based on the deduction authority and the to-be-settled order.

2. The automatic vending machine recognition system based on image processing according to claim 1, characterized in that, The authorization terminal includes: An instruction receiving module, configured to receive a purchase request instruction input by the buyer based on a touch display screen arranged outside the vending machine; An image acquisition module, configured to, when receiving the purchase request instruction, acquire all high-definition face images within a preset range in front of the vending machine; An authorization authentication module, configured to perform authorization authentication on the buyer based on the purchase request instruction and all high-definition face images, and obtain an authorization authentication result.

3. The automatic vending machine recognition system based on image processing according to claim 2, characterized in that, The image acquisition module includes: An image acquisition unit, configured to, when receiving the purchase request instruction, acquire all personnel images within a preset range in front of the vending machine; A dynamic tracking unit, configured to perform dynamic tracking on the personnel images to obtain a dynamic tracking video of each personnel image; A face recognition unit, configured to recognize all high-definition face images within a preset range in front of the vending machine based on the dynamic tracking video.

4. The automatic vending machine recognition system based on image processing according to claim 3, characterized in that, The face recognition unit includes: A face determination subunit, configured to recognize a dynamic face video of each personnel image in the dynamic tracking video, and determine a face area and a face contour in each face image frame in the dynamic face video; A frame determination subunit, configured to determine the bone point positions corresponding to the positioning bone points included in a preset bone point list in the face area, and connect the bone point positions included in the face area based on a preset connection relationship between the positioning bone points, to obtain a bone point distribution frame corresponding to the face area; An angle determination subunit, configured to perform surrounding matching on the bone point distribution frame and a standard bone point distribution three-dimensional frame to determine the shooting angle of the face area; An optimal determination subunit, configured to determine whether there is more than one face area corresponding to each shooting angle. If so, obtain a face area set and a face contour set corresponding to the shooting angle, and fit an optimal face area and an optimal face contour corresponding to the shooting angle based on the face area set and the face contour set. Otherwise, use the corresponding face area as the optimal face area corresponding to the shooting angle, and at the same time, use the corresponding face contour as the optimal face contour corresponding to the shooting angle; A region determination subunit, configured to determine a first small-variation region in the optimal face area based on the shooting angle and a small-variation region with a smaller amplitude in the standard bone point distribution three-dimensional frame. The facial feature determination sub-unit is configured to regard the remaining area in the best facial area except the first area with a small amplitude change as the facial feature area, identify the facial feature contours in the facial feature area, calculate the movement amplitude of the facial feature area based on the facial feature contours and the positions of the bone points included in the facial feature area, and use the facial feature area corresponding to the minimum movement amplitude as the best facial feature area for the corresponding shooting angle; The model restoration sub-unit is configured to perform contour restoration on the standard bone point distribution three-dimensional framework based on the best facial contour corresponding to each shooting angle, and perform local restoration on the standard bone point distribution three-dimensional framework based on the first area with a small amplitude change and the corresponding best facial feature area corresponding to each shooting angle, to obtain a three-dimensional facial model of the corresponding person's image; The image processing sub-unit is configured to screen out the best facial image corresponding to each person's image in the dynamic facial video, and perform local restoration on the best facial image based on the corresponding three-dimensional facial model, to obtain a high-definition face image within a preset range in front of the vending machine.

5. The automatic vending machine recognition system based on image processing according to claim 2, characterized in that, The authorization and authentication module includes: The terminal determination unit is configured to determine the authorization and authentication terminal based on the purchase request instruction; The instruction sending unit is configured to generate an authorization and authentication instruction based on all the high-definition face images and the purchase request instruction, and send the authorization and authentication instruction to the authorization and authentication terminal; The account determination unit is configured to receive the authorization and authentication feedback instruction from the authorization and authentication terminal, and determine the buyer's facial image and the automatic deduction account of the buyer based on the authorization and authentication feedback instruction; The authorization and authentication unit is configured to perform authorization and authentication on the buyer based on the automatic deduction account, and combine the buyer's facial image to obtain the authorization and authentication result.

6. The automatic vending machine recognition system based on image processing according to claim 5, characterized in that, The authorization and authentication unit includes: The balance acquisition sub-unit is configured to acquire the deductible balance in the automatic deduction account; The authorization and authentication sub-unit is configured to, when the deductible balance is not less than the authorized balance, regard the buyer as having passed the authorization and authentication and the buyer's facial image as the authorization and authentication result, otherwise, regard the buyer as not having passed the authorization and authentication as the authorization and authentication result, and send an authorization and authentication failure prompt instruction to the authorization and authentication terminal.

7. The automatic vending machine recognition system based on image processing according to claim 1, characterized in that, The monitoring end includes: The opening control module is configured to control the vending machine to open the purchase window when the authorization and authentication result is that the buyer has passed the authorization and authentication; The facial image acquisition module is configured to acquire the buyer's facial image based on the authorization and authentication result when the authorization and authentication result is that the buyer has passed the authorization and authentication; The video screening module is configured to acquire the internal taking video based on the camera arranged inside the vending machine, and screen out the whole-process video of the buyer taking goods in the vending machine from the internal taking video based on the buyer's facial image.

8. An automatic vending machine recognition method based on image processing, characterized in that, Applied to any one of the image processing-based vending machine recognition systems described in claims 1 to 7, including: S1: Acquire the high-definition face image of the buyer, perform authorization and authentication on the buyer based on the high-definition face image, and obtain the authorization and authentication result; S2: Control the vending machine to open the purchase window based on the authorization and authentication result, and obtain the full-process video of the buyer picking up goods in the vending machine; S3: Generate a settlement order based on the goods to be settled identified in the full-process video, and perform an automatic deduction operation based on the settlement order and the authorization and authentication result.

Images