Payment service auxiliary processing method, device and equipment

Abstract

The embodiment of the specification discloses a payment service auxiliary processing method, device and equipment, and is applied to a near-field cash register device. The method comprises the following steps: performing near-field sensing detection on a user based on sensor data; if it is detected that a user enters a near-field range of the near-field cash register device, obtaining a shop circle selection result provided by a server of the near-field cash register device; and judging whether to enable an auxiliary recommendation model according to whether a shop to which the near-field cash register device belongs is included in the shop circle selection result, wherein the auxiliary recommendation model is trained by the server of the near-field cash register device according to operation data provided by multiple shops and is delivered to the near-field cash register device; if yes, before the user completes payment, the auxiliary recommendation model is used to generate corresponding service auxiliary information for the user and recommend the service auxiliary information to the user.

Description

Technical Field

[0001] This specification relates to the field of electronic payment technology, and in particular to methods, devices and equipment for auxiliary processing of payment transactions. Background Technology

[0002] Near Field Communication (NFC) is a short-range, high-frequency radio technology that operates at a frequency of 13.56 MHz within a range of 20 centimeters. It evolved from contactless radio frequency identification (RFID) and interconnection technologies, providing a highly secure and fast communication method for various electronic products.

[0003] With the widespread use of smartphones supporting Near Field Communication (NFC), NFC technology is increasingly being applied in the payment field. For example, when merchants deploy NFC-enabled POS devices at checkout counters, users can use NFC-enabled phones to make payments by interacting with the NFC device, offering good convenience. However, since NFC POS devices are currently often implemented as extensions, users sometimes don't notice them until they realize that another payment method has already been initiated, requiring them to switch payment methods to utilize the NFC POS device, which may impact the user's payment experience.

[0004] Therefore, solutions are needed to further improve the user payment experience. Summary of the Invention

[0005] This specification provides one or more embodiments of a payment service auxiliary processing method, apparatus, and device to address the following technical problem: the need for a solution that helps to further improve the user payment experience.

[0006] To solve the above-mentioned technical problems, one or more embodiments of this specification are implemented as follows:

[0007] This specification provides one or more embodiments of a payment transaction auxiliary processing method, applied to a near-field cash register device, the method comprising:

[0008] Based on sensor data, perform near-field perception and detection of users;

[0009] If a user is detected entering the near-field range of the near-field POS device, the store selection result provided by the server of the near-field POS device is obtained;

[0010] Whether to enable the auxiliary recommendation model depends on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0011] If so, before the user completes the payment, the auxiliary recommendation model generates corresponding business assistance information for the user and recommends it to the user.

[0012] This specification provides one or more embodiments of a payment transaction auxiliary processing method, applied to a near-field cash register device, the method comprising:

[0013] A near-field perception model is obtained by fine-tuning a specified basic recognition model using upward-view samples, wherein the upward-view samples include real-life photos taken by a camera at an upward angle.

[0014] Using the near-field perception model, the image data dynamically acquired by the camera of the near-field cash register is roughly detected to determine whether a local limb other than the head has entered the near-field range.

[0015] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0016] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0017] Based on the near-field perception results, service assistance information is recommended to the user.

[0018] This specification provides one or more embodiments of a payment transaction auxiliary processing device, applied to a near-field cash register device, the device comprising:

[0019] The near-field perception and detection module performs near-field perception and detection of users based on sensor data;

[0020] The selection result acquisition module acquires the store selection result provided by the server of the near-field POS device if it detects that a user has entered the near-field range of the near-field POS device.

[0021] The recommendation model triggering module determines whether to enable the auxiliary recommendation model based on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0022] If so, the auxiliary information recommendation module generates relevant business auxiliary information for the user before the user completes payment, and recommends it to the user.

[0023] This specification provides one or more embodiments of a payment transaction auxiliary processing device, applied to a near-field cash register device, the device comprising:

[0024] The near-field perception model acquisition module acquires a near-field perception model by fine-tuning a specified basic recognition model using upward-looking samples, wherein the upward-looking samples include real-life photos taken by a camera at an upward-looking angle.

[0025] The near-field perception model detection module uses the near-field perception model to perform a rough detection on the image data dynamically collected by the camera of the near-field cash register to determine whether a local limb other than the head has entered the near-field range.

[0026] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0027] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0028] The business assistance information recommendation module recommends business assistance information to the user based on the near-field perception results.

[0029] This specification provides one or more embodiments of a payment transaction auxiliary processing device, applied to a near-field cash register device, the payment transaction auxiliary processing device comprising:

[0030] At least one processor; and,

[0031] A memory communicatively connected to the at least one processor; wherein,

[0032] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform:

[0033] Based on sensor data, perform near-field perception and detection of users;

[0034] If a user is detected entering the near-field range of the near-field POS device, the store selection result provided by the server of the near-field POS device is obtained;

[0035] Whether to enable the auxiliary recommendation model depends on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0036] If so, before the user completes the payment, the auxiliary recommendation model generates corresponding business assistance information for the user and recommends it to the user.

[0037] This specification provides one or more embodiments of a payment transaction auxiliary processing device, applied to a near-field cash register device, the payment transaction auxiliary processing device comprising:

[0038] At least one processor; and,

[0039] A memory communicatively connected to the at least one processor; wherein,

[0040] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform:

[0041] A near-field perception model is obtained by fine-tuning a specified basic recognition model using upward-view samples, wherein the upward-view samples include real-life photos taken by a camera at an upward angle.

[0042] Using the near-field perception model, the image data dynamically acquired by the camera of the near-field cash register is roughly detected to determine whether a local limb other than the head has entered the near-field range.

[0043] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0044] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0045] Based on the near-field perception results, service assistance information is recommended to the user.

[0046] The above-described technical solutions employed in one or more embodiments of this specification can achieve the following beneficial effects: Near-field payment devices can proactively sense a user entering the near-field range. Based on the store selection results and auxiliary recommendation model provided by the server of the near-field payment device, they can accurately decide whether to proactively attract the current user's attention through recommendations of corresponding business auxiliary information (such as business promotional information, business-appropriate caring remarks, etc.), while providing the user with more payment-related information, thereby helping to improve the user's payment experience. Furthermore, this is also suitable for near-field payment devices placed horizontally. Through fine-grained control of the detection frame, it can more accurately sense a user entering the near-field range and can further assist in the subsequent business auxiliary information process. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A flowchart illustrating a payment transaction auxiliary processing method provided in one or more embodiments of this specification;

[0049] Figure 2 A flowchart illustrating a near-field sensing and detection scheme provided in one or more embodiments of this specification;

[0050] Figure 3 In one application scenario provided by one or more embodiments of this specification, Figure 1 A schematic diagram of the architecture of a specific implementation scheme of the solution;

[0051] Figure 4 A schematic diagram illustrating the working cycle principle of a server-side near-field cash register device provided in one or more embodiments of this specification;

[0052] Figure 5 A schematic diagram illustrating the working principle of a near-field cash register device provided in one or more embodiments of this specification;

[0053] Figure 6 A merchant and user feature data architecture and application framework diagram provided for one or more embodiments of this specification;

[0054] Figure 7 A schematic diagram illustrating the principle of a near-field operation assistant intelligent agent provided in one or more embodiments of this specification;

[0055] Figure 8 A flowchart illustrating another payment transaction auxiliary processing method provided in one or more embodiments of this specification;

[0056] Figure 9 A schematic diagram of a payment transaction auxiliary processing device provided for one or more embodiments of this specification;

[0057] Figure 10 A schematic diagram of another payment transaction auxiliary processing device provided in one or more embodiments of this specification;

[0058] Figure 11 A schematic diagram of the structure of a payment business auxiliary processing device provided in one or more embodiments of this specification;

[0059] Figure 12 This is a schematic diagram of another payment service auxiliary processing device provided in one or more embodiments of this specification. Detailed Implementation

[0060] This specification provides embodiments of payment transaction auxiliary processing methods, apparatus, devices, and storage media.

[0061] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0062] Figure 1 This diagram illustrates a payment processing auxiliary method provided in one or more embodiments of this specification. The method is applied to a near-field payment device. The near-field payment device supports at least one near-field sensing payment method (e.g., near-field communication sensing, acoustic sensing, infrared sensor), and can also support other payment methods. It can be a basic payment device for a store, deployed independently; or it can be an additional extension device connected to the store's existing basic payment device (e.g., typically supporting traditional payment methods such as cash, bank cards, and credit cards), deployed in conjunction with it to extend the basic payment device's support for one or more other payment methods (e.g., near-field communication payment methods).

[0063] For example, basic POS devices can be the merchant POS devices commonly found at the checkout counter, while near-field POS devices can be small near-field communication payment POS devices. Near-field POS devices are usually placed roughly flat on the checkout counter (there may be a slight tilt, but they are generally lying flat, which makes it easier for users to interact with them). Users can make payments by touching their smartphones or smartwatches with near-field communication capabilities to the device.

[0064] Figure 1 The process includes the following steps:

[0065] S102: Based on sensor data, perform near-field perception detection of users.

[0066] Sensor data, including at least image data (pictures, videos, etc.), and potentially more modal data such as audio data, helps to achieve more accurate near-field perception and detection. Sensors include at least a camera, but may also include devices such as microphones, speakers, infrared sensors, Bluetooth modules, or Wi-Fi modules. This allows for the use of multimodal sensing data and multimodal algorithms to fuse information from different modalities, achieving more comprehensive and human-like intelligent reasoning and obtaining more accurate perception results.

[0067] The sensors themselves can be deployed on the near-field POS device itself, making them more convenient to use and facilitating more accurate data collection. Based on these sensors and the corresponding sensing data, many other detection methods can also be supported; here, we are only emphasizing near-field sensing detection.

[0068] Near-field perception detection is used to determine whether a user has entered the near-field range of a near-field checkout device. In addition, it can be used to make more detailed analyses, such as whether the detected user is a real live user, whether the user in the near-field range has the intention to pay, and dynamically continue to infer the specific scenario and interaction stage based on the user's behavior in the near-field range during the checkout process, etc.

[0069] S104: If a user is detected to have entered the near-field range of the near-field POS device, the store selection result provided by the server of the near-field POS device is obtained.

[0070] In one or more embodiments of this specification, there may be one or more near-field cash register devices, of which at least one server can be used to provide business assistance information recommendation services for the near-field cash register device, without necessarily handling payment transactions. If necessary, other servers can handle payment transactions.

[0071] The server side of the near-field POS device can dynamically select some stores from a large number of stores based on the basic business recommendation information and customer flow profile data provided by multiple stores. This selection result triggers the near-field POS devices in the selected stores to perform the business auxiliary information recommendation process.

[0072] Basic business recommendation information includes, for example, promotional information such as best-selling or new product information, limited-time offers, and introductions to new equipment capabilities; intelligent care information matched to the business; and business information from other stores that, while not directly related to this store, are likely to attract or help users; etc. This basic business recommendation information can be used as the starting point for further refinement and optimization to generate more concise, expressive, and impactful, or more targeted information, serving as the final supplementary business recommendation information.

[0073] S106: Determine whether to enable the auxiliary recommendation model based on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0074] Based on whether the store to which the near-field POS device belongs is included in the store selection results, it is determined whether to recommend business assistance information to the user this time.

[0075] The store selection results can be pre-distributed to the near-field POS device, and the judgment action in step S106 can also be performed in advance after the store selection results are distributed. For example, if a user is detected entering the near-field range of the near-field POS device, step S108 can be executed directly. In this case, local computing power can be focused on making more granular and specific predictions.

[0076] The server-side of the near-field POS device, leveraging the vast amount of additional information it can acquire compared to near-field POS devices located in stores, pre-trains an auxiliary recommendation model. This model is then provided to the near-field POS device for accurate and intelligent information recommendations, helping to improve user experience and conversion rates. Therefore, determining whether to enable the auxiliary recommendation model can essentially include deciding whether to perform business-assisted information recommendations based on server-side capabilities.

[0077] The auxiliary recommendation model can selectively acquire business service information that is more appealing or helpful to a particular customer or user based on the store's customer traffic profile or the current user's user profile. This auxiliary recommendation model can be tailored to the specific circumstances of different stores, allowing for the provision of a more suitable model to a particular store, rather than necessarily using a uniform model.

[0078] Operational data can include the store's own business, how users use the store's services, or the impact of information recommended to users on user conversion rates. This operational data can directly or indirectly reflect the effectiveness of recommendations and can also help select or generate new business support information.

[0079] S108: If so, before the user completes the payment, the auxiliary recommendation model generates corresponding business assistance information for the user and recommends it to the user.

[0080] Business support information can include promotional information or caring messages tailored to the current business, user, or situation. This information is presented to users in one or more formats, including at least voice announcements, to support them or guide them through additional actions, such as modifying orders, browsing other content, or becoming a member. In this context, compared to traditional loudspeakers, the promotional capabilities of near-field payment systems (NFC) are more flexible, intelligent, and effective. This encourages users to use NFC more actively and easily, helps them notice more valuable services, and provides a sense of warmth and care. Simultaneously, it effectively reduces the workload of store staff.

[0081] In one or more embodiments of this specification, based on continuous near-field sensing detection or other sensor data, and based on the capabilities of the auxiliary recommendation model, it is also possible to determine one or more appropriate recommendation opportunities before the user completes payment, or even immediately after the payment is completed, and to accurately recommend the corresponding business auxiliary information at the recommendation opportunity, thereby avoiding unintentionally disturbing the user or not disturbing the current business.

[0082] In addition, if needed, near-field POS devices can make more flexible decisions on whether to enable auxiliary recommendation models to recommend business assistance information for the current user, without necessarily being limited by the store selection results. Figure 1 The process within the system is more conducive to making recommendations in a holistic way, so that the recommended information is not limited to the information of the current store itself.

[0083] pass Figure 1 This method enables near-field payment devices to proactively detect when a user enters the near-field range. Based on the store selection results and auxiliary recommendation models provided by the near-field payment device's server, it accurately decides whether to proactively attract the user's attention by recommending relevant business support information (such as business promotional information, or business-appropriate caring remarks), while providing the user with more payment-related information, thereby improving the user's payment experience. Furthermore, it is also suitable for near-field payment devices placed horizontally. Through fine-grained control of the detection frame, it can more accurately detect when a user enters the near-field range and continue to support subsequent business support information processes.

[0084] based on Figure 1 In addition to the method described herein, this specification also provides some specific implementation schemes and extension schemes of this method, which will be further explained below.

[0085] In one or more embodiments of this specification, particularly for a specific scenario, a more accurate near-field perception detection scheme is provided. In this specific scenario, the sensor data includes at least image data dynamically collected by the camera of the near-field POS device (which can directly utilize the barcode scanning camera without adding a dedicated camera to reduce costs). The near-field POS device is placed in the aforementioned flat orientation, and the camera is adapted to the flat orientation of the near-field POS device, generally using an upward-looking perspective to collect image data.

[0086] In this application, we aim to implement a flexible and intelligent means of recommending business assistance information on near-field payment devices. For example, after recognizing that a user is approaching the cashier, we can announce the corresponding business assistance information to guide the customer to complete the payment action or other additional business actions expected by the service side.

[0087] Although human detection is a relatively mature computer vision technology on standard datasets, its application to the retail checkout scenario addressed in this application presents significant challenges. The complexity of implementing human detection to assist near-field perception using the flat-positioned near-field checkout device described in this application exceeds conventional understanding. This is reflected in at least the following five aspects.

[0088] First, there are unique perspective and scale issues. The cameras on near-field checkout devices are adapted to a horizontal, upward-looking perspective. This non-horizontal perspective causes severe perspective distortion in both people and merchandise. A distant customer might only be a few dozen pixels, while a nearby customer can easily occupy most of the frame. This significant scale variation places high demands on the robustness of the model. Simultaneously, the camera's narrow field of view and tricky angles make it difficult to acquire the full-body or half-body images relied upon by standard human detection models. The input data is often incomplete, typically only showing the torso, arms, and hands above the waist. This poses a fundamental challenge to detection models that rely on the overall contours and proportions of the human body.

[0089] Second, there is the widespread and severe issue of occlusion. The checkout environment is dynamic and crowded; customers frequently obstruct each other's view, are blocked by equipment or merchandise on the checkout counter, or are blocked by items in their hands. The model needs strong reasoning capabilities to perform recognition even when only part of the target is visible.

[0090] Third, dynamics and ambiguity. Checkout is a fast, dynamic process; the arms of customers and cashiers quickly move across the camera, causing motion blur. Simultaneously, the hands of both customers and cashiers may appear in the frame at the same time. The model not only needs to detect people but also needs to distinguish the interacting subjects, which increases the complexity of the task.

[0091] Fourth, unconventional postures. Customers' postures at the counter vary greatly, such as leaning forward, turning to the side, reaching out to hand over items, or looking down to find money. These are not common postures in standard pedestrian datasets, further increasing the difficulty of model generalization.

[0092] Fifth, grayscale images. The images from barcode scanners are grayscale images, not normal RGB images, so they contain less information and some unique human image features are lost. Since most detection models are trained on RGB images, they are not easy to adapt.

[0093] To help solve at least one of the above problems, this application provides a basic solution and some extended solutions, see [link to relevant documentation]. Figure 2 , Figure 2 This is a flowchart illustrating a near-field sensing and detection scheme provided in one or more embodiments of this specification, and the near-field sensing and detection scheme is exemplarily used as the basic scheme.

[0094] Figure 2 The process includes the following steps:

[0095] S202: Obtain a near-field perception model by fine-tuning a specified basic recognition model using upward-looking samples, wherein the upward-looking samples include real-life photos taken by a camera at an upward-looking angle.

[0096] For near-field perception models, the process involves data acquisition, model training, and model deployment.

[0097] Regarding data acquisition, exposure selection is of particular concern. In one or more embodiments of this specification, a single capture by the barcode scanner will generate multiple (e.g., 6) images with different exposures. Each image has a different exposure value, resulting in different image quality. If each image is inspected, the process is time-consuming, and the inspection results for each image will vary, requiring additional fusion strategies, which is detrimental to efficiency and accuracy. Therefore, it is advisable to select an image with relatively better image quality for inspection, which can also serve as a baseline sample. For example, the RGB histogram of each image can be calculated, and images with relatively fewer overexposed (e.g., pixel value > 240) and underexposed (e.g., pixel value < 50) pixels can be selected. This ensures that the selected image does not have excessive overexposed / underexposed areas, guaranteeing image clarity.

[0098] For data retrieval, to ensure data diversity, a series of retrieval strategies are employed, including: ensuring diversity in the devices used to retrieve images, thereby ensuring diversity in image backgrounds and enhancing the generalization of the model; and interval sampling to avoid sampling too many similar images.

[0099] For model training, an existing model can be specified as the base recognition model, for example, a base recognition model based on the YOLO architecture. A batch of images of customers in front of the camera can be further annotated, including the customer's bounding box (e.g., box coordinates). Then, the base recognition model can be fine-tuned to obtain a near-field perception model with higher recognition accuracy.

[0100] Regarding model deployment, the scenario described in this application needs to be considered, as the model needs to be used on a near-field payment device, meaning it needs to be deployed on the device itself. Therefore, high efficiency is required for the model. Based on this, assuming the training framework used is PyTorch, the trained PyTorch model can be converted into an ONNX model, reducing recognition time and improving efficiency.

[0101] S204: Using the near-field perception model, perform a rough detection on the image data dynamically acquired by the camera of the near-field cash register to determine whether a local limb other than the head has entered the near-field range.

[0102] In one or more embodiments of this specification, image distortion occurs when viewed from a low angle, which can make human detection more sensitive to help trigger the model more comprehensively and avoid missed recognition. However, if human detection is too sensitive, a detection box will be generated when only a part of the human body, such as a hand or half a head, is visible in front of the camera, which may waste detection resources. Therefore, in order to balance sensitivity and reasonable use of resources, on the one hand, it can be sensitive to parts of the limbs other than the head, and on the other hand, it avoids formally and carefully identifying the specific parts of the limbs.

[0103] S206: If so, then for the local limb, suppress the generation of the first fine detection box.

[0104] For limbs other than the head, the corresponding detection bounding box can be left blank, thus avoiding entering a detailed detection state for those limbs. At the same time, preprocessing actions such as resource allocation and model pre-start can be performed in advance to prepare for the appearance of the head.

[0105] S208: During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0106] If only part of the head enters the near-field range, the generation of the first fine-grained detection box can still be suppressed. If it is determined that the head has completely entered the near-field range, a corresponding detection box can be generated for the head to enter the fine-grained detection state. This detection serves two purposes: firstly, it confirms that a real user has indeed entered the near-field range of the near-field POS device, suggesting that such a user intends to conduct business through the near-field POS device, such as making a payment; secondly, if needed, this fine-grained detection can be used to obtain other features such as user expression, gender, and identity, generating corresponding user profile data, and helping to assist in subsequent recommendations of business assistance information more suitable for the user. For example, based on the head features, the gender of the user can be identified using a classification model. If the user is female, then information on products or promotional activities that women may be interested in and that are specifically designed for women can be recommended.

[0107] Furthermore, based on the head detection results, it is determined whether a user with payment intent has entered the near-field range, and corresponding user profile data is generated. If so, the above inhibition can be lifted, and a second fine-grained detection box can be generated for the local limbs other than the head. The local limbs are dynamically detected using the second fine-grained detection box (for example, by obtaining the user's real-time gestures or other habitual or indicative limb movements). Based on the detection results of the second fine-grained detection box, the user profile data is further improved. Before the user completes the payment, the timing of recommending the corresponding business auxiliary information is determined, so that the recommendation can be made more accurately based on the user profile data at the recommendation time.

[0108] This allows the model to prioritize head features, improving the robustness of head recognition in areas that are relatively far away and have greater distortion. Furthermore, after accurately recognizing head features, it also focuses on other local limb features, resulting in better anti-distortion recognition performance in scenarios where the camera is viewed from below. At the same time, it can also recognize features efficiently, making it more friendly to the performance of on-device models.

[0109] To improve detection accuracy, before generating the first fine detection box to detect the head, it can be determined whether the size of the fine detection box to be generated meets the set false alarm object size. If so, the fine detection box to be generated is filtered out and not actually used for detection, in order to avoid misidentification caused by interference objects and scale confusion factors that may arise in dynamic and crowded checkout scenarios. For example, under the setting that each side of the image data is no less than 320 pixels and no more than 640 pixels (generally, the image size of a barcode camera is 640*640 pixels; to improve model efficiency, for example, the input image size used by the model is reduced to 320*320 pixels), the false alarm object size can exemplary include at least one of the following two: the long side / short side is greater than 2 pixels and the short side is less than 60 pixels; the short side is greater than 260 pixels. This exemplary false alarm object size is particularly preferred when the model input image size is reduced to 320*320 pixels.

[0110] In one or more embodiments of this specification, the near-field POS device is connected to the basic POS device of the store, which further extends the basic POS device's support for one or more other payment methods; before performing near-field perception detection on the user based on sensor data, it interacts with the server of the near-field POS device in advance to provide data support for the server to select stores or train models.

[0111] In this scenario, store clerks can perform operations on the basic POS device to provide basic business recommendation information to the near-field POS device. The near-field POS device then responds to this operation by receiving the basic business recommendation information transmitted from the basic POS device.

[0112] Similarly, basic business recommendation information and store customer flow profile data can be synchronized from the basic POS device to its server. Then, the basic POS device's server sends the same basic business recommendation information and customer flow profile data (which may also include relevant information from other stores) to the near-field POS device's server. The near-field POS device's server trains an auxiliary recommendation model based on operational data from multiple stores, including basic business recommendation information and store customer flow profile data, and distributes this data to the near-field POS devices. The near-field POS device's server can also determine the store selection results based on the basic business recommendation information and customer flow profile data provided by multiple stores.

[0113] In this way, the server-side capabilities of near-field POS devices can be fully utilized, and the data limitations of a single store can be overcome, enabling more flexible recommendations. This not only helps users use services in the current store, but also helps guide users' movement tendencies between different stores.

[0114] Based on this approach, one or more embodiments of this specification provide an auxiliary recommendation model initiation decision-making scheme, the process of which includes the following steps:

[0115] S302: Determine whether the store to which the near-field POS device belongs is included in the store selection results.

[0116] If the store is not included in the store selection results, then no recommendation will be made this time, and the auxiliary recommendation model will not be activated accordingly.

[0117] S304: If so, determine the leading merchant corresponding to the store selection result (which may have a corresponding offline store or may only operate online without an offline store). The leading merchant is a merchant that is highly relevant to the specified leading operation activity. The business auxiliary information will be generated based on the leading operation activity through the auxiliary recommendation model.

[0118] In one or more embodiments of this specification, the store to which the business belongs is not a dominant merchant. Preferably, the business of the dominant merchant and the business of the store to which the business belongs have a certain degree of continuity. In this case, the goal is to naturally guide the user to visit the store to which the business belongs (and thus use the business of the store to complete the corresponding checkout operation) and then continue to use the business of the dominant merchant. Regarding the continuity of business, for example, the former business could be clothing retail, and since accessories for clothing can be matched with clothing, the latter business could be accessories retail; another example is that the former business could be children's picture book retail, and since children may subsequently read picture books or dine, the latter business could be a coffee and dessert business that provides a reading space for children, or a children's restaurant business suitable for children; and so on. The specific degree of continuity between different businesses can be calculated through matching calculations of business features or model inference, combined with scene features (such as time, season, etc.).

[0119] The leading operational activity can be primarily provided by the leading merchant, be highly relevant to the leading merchant's business, or, although not directly provided, have sufficient control, such as being an important partner in the leading operational activity.

[0120] In one or more embodiments of this specification, the store selection action based on the dominant merchant can be dynamically and frequently executed. This allows for the simultaneous execution of different store selection actions suitable for different stores, and enables more stores to become dominant merchants, indirectly utilizing the recommendation resources of other stores to assist in recommending their own relevant business information. This breaks through the limitations of the dominant merchant's own store space, effectively increasing dissemination. Moreover, for users, such recommended business assistance information is not only not abrupt, but also brings them continuous, predictable, and proactive convenience.

[0121] S306: Obtain conflict risk data between the store and the dominant merchant or the dominant operational activity.

[0122] In real life, there may be competition between different shops or merchants, and there may also be conflicts between the specific business of a shop and its main operating activities. Therefore, conflict risk data that measures the degree of such competition or conflict can be calculated by setting corresponding factors (for example, shops and / or merchants can set pre-defined avoidance labels, business red lines, competitors, operational preferences, etc.) and weights.

[0123] S308: If the conflict risk data reflects a sufficiently high conflict risk, then it is determined that the auxiliary recommendation model will not be activated.

[0124] If the conflict risk data reflects a sufficiently high conflict risk, it indicates that recommending the product to the current store might actually be detrimental to it. Therefore, the auxiliary recommendation model can be omitted in this instance.

[0125] If the conflict risk data reflects a sufficiently low conflict risk, then normal recommendations can be made. This helps guide users to move on to the main operational activities or the main merchant's services after completing payment at the current store. In this case, the recommended business support information can effectively guide users across stores.

[0126] In one or more embodiments of this specification, consideration is given to choosing an appropriate time to make recommendations so as not to disrupt normal business operations, thereby enabling proactive monitoring of device status.

[0127] Specifically, when the near-field POS device is interconnected with the basic POS device, the first business status information is synchronized from the basic POS device; based on the first business status information and the second business status information of the near-field POS device itself; based on the first business status information and the second business status information, it is determined whether the recommendation timing conflicts with the predetermined business sensitive status. If so, the recommendation timing is re-determined to avoid conflict; otherwise, business auxiliary information can be recommended normally according to the original recommendation timing.

[0128] Business status information, such as whether the system is in checkout mode, on standby screen, playing voice prompts, or engaging in other key interactions with the user, etc., allows for timely awareness of the status of higher-level business processes that hold a primary position.

[0129] In one or more embodiments of this specification, fatigue control can be implemented to avoid excessive user disruption from recommendations. Specifically, for example, the server-side data of the near-field POS device is used to calculate a baseline fatigue level based on store transaction characteristics; based on the baseline fatigue level and auxiliary fatigue data from the near-field POS device itself, a decision is made on whether to temporarily suspend further recommendations to the user.

[0130] Store transaction characteristics can include, for example, peak transaction times, number of transactions, or number of near-field payment transactions. They can also include negative indicators, such as a decrease in volume.

[0131] Furthermore, in the scenario described in this application, proactive recommendations across stores are achieved. Therefore, when controlling fatigue, not only is the current store considered, but also other related stores are taken into account at a global level. Following this approach, the server-side of the near-field POS device can calculate the basic fatigue level. For example, based on the user's recent dynamic trajectory obtained from multiple different stores within the same store selection result, the fatigue level accumulated by the user in each of these stores along the way is determined based on the dynamic trajectory. This, combined with other factors, allows for a more comprehensive, sustained, and accurate measurement of whether a user may have developed fatigue from the corresponding recommendations.

[0132] Based on the foregoing description, and more intuitively, one or more embodiments of this specification also provide an application scenario where... Figure 1 One specific implementation plan of the scheme, see Figure 3 , Figure 3 This is a schematic diagram of the architecture of the solution.

[0133] The near-field device side mainly includes: near-field POS devices and their servers.

[0134] This server can handle two aspects: operation frequency and operation material management, which can specifically include recommendation model management, experiment management, and operation plan management.

[0135] Recommendation model management: The server can obtain user or store traffic profiles through the data platform, and use different profiles to train the recommendation model in a targeted manner. For example, it can predict which operational materials have a higher conversion rate in the store, and then increase the push of operational materials (as the above-mentioned business auxiliary information) to the store.

[0136] Experimental management includes store selection and model control. Since different stores typically conduct different operational activities, relying solely on manual methods makes rapid selection and configuration difficult. The server provides a T+1 rapid store selection capability, leveraging the data platform to obtain ongoing operational activities and determine whether to implement near-field recommendation operations. In addition, the server also utilizes the data platform to provide comparative testing and data evaluation capabilities for different recommendation models, facilitating rapid verification of model effectiveness and the selection of better solutions.

[0137] Operational Plan Management: Operational plans, specifically, can be domain models provided by the server to manage store operational materials and frequency, including fatigue management and operational material management. Fatigue management primarily calculates appropriate fatigue levels based on store transaction characteristics, such as peak transaction periods, number of transactions, number of near-field payment transactions, and negative indicators like decreased volume, and dynamically adjusts near-field POS devices accordingly. Operational material management includes uploading, managing, and deciding on operational materials, using models or rules to determine the operational materials and display methods that should be shown on the client side at any given time.

[0138] Near-field payment devices have capabilities such as small commands, voice, and image recognition, and also provide corresponding UIs to enable interaction with users.

[0139] For recommendation models, an exemplary model architecture and training scheme are provided here.

[0140] Regarding the architecture, for example, using the YOLOv8-n network architecture, CSP-DARKNET can be used as the feature extraction backbone network, and PAFPN as the feature fusion network. Through the processing of these networks, the coordinates and category of the target box are output. The original network input is a 640*640 pixel value, but to speed up the operation on the device, the input is changed to a 320*320 pixel value.

[0141] For the training scheme: For the training data, collect a set of relevant images and annotate the human bounding boxes (e.g., the coordinates of the four corners); For training fine-tuning, fine-tune the original model, for example, setting the batch size to 32 and reducing the learning rate to 1 / 10 of the original setting.

[0142] Training frameworks, such as PyTorch, can be used. To improve efficiency, PyTorch models can be converted into ONNX models using the ONNX engine.

[0143] The solution also includes, for example, a merchant server and a merchant POS device (as the basic POS equipment mentioned above). The execution process of the solution can be understood by referring to the modules and steps in the diagram and the preceding explanation; some non-critical parts are not described in detail here.

[0144] Figure 4 This is a schematic diagram illustrating the working cycle principle of a server-side near-field payment device provided in one or more embodiments of this specification.

[0145] A single cycle mainly includes the following aspects: deploying the plan, collecting business data after the plan takes effect, analyzing the business data after the plan takes effect, and generating an optimization plan based on rules or algorithms. For each aspect, the diagram provides some specific processing actions for reference.

[0146] Figure 5 This is a schematic diagram illustrating the working principle of a near-field payment device provided in one or more embodiments of this specification.

[0147] exist Figure 5 In its working principle, compared to the original payment chain used by near-field POS devices, modifications have been made before payment, adding a near-field operation recommendation display step, which includes at least:

[0148] State synchronization: In order to avoid affecting the original payment chain, the near-field operation recommendation module will sense the status of the upper-layer business, including whether it is in the cashier state, whether it is in the standby page, whether the voice is playing, etc., and determine whether near-field operation recommendation can be performed at this time through state synchronization.

[0149] Near-field perception detection: Near-field perception detection is one of the keys to near-field operation and recommendation. It uses algorithm models to detect images or speech. The near-field operation and recommendation module can dynamically switch between different algorithm models through server-side model management.

[0150] Fatigue Management: To avoid negative impacts from excessive operational recommendations, the near-field operational recommendation module implements detailed controls on the frequency of operational recommendations, including the hourly playback limit, the interval between two consecutive operational recommendations, and the interval after a transaction.

[0151] Figure 6 This specification provides a merchant and user feature data architecture and application framework diagram for one or more embodiments. Near-field POS devices and their servers can extract the required feature data based on this framework and apply it to near-field operational recommendations.

[0152] Within this framework, for example, data collected from devices / sensors is processed through ETL to obtain data related to merchant operations, which is then further aggregated into merchant, customer, and product characteristic data. Based on this characteristic data, it is packaged into data services through middleware such as data analytics, and made available for consumption by downstream devices, large and small models, and near-field POS devices.

[0153] Figure 7This is a schematic diagram illustrating the principle of a near-field operation assistant intelligent agent provided in one or more embodiments of this specification. The solution proposed in this application can be implemented by deploying this near-field operation assistant intelligent agent in a near-field POS device (which may also include its server for integrated deployment).

[0154] exist Figure 7 In the diagram, the upper layer represents merchants and customers, the middle layer represents the near-field operations assistant intelligent agent, and the lower layer represents infrastructure such as devices, models, and databases. The near-field operations assistant intelligent agent is the main component, whose capabilities are divided into perception, intent recognition, reasoning, decision-making, and action in sequence.

[0155] The perception and intent recognition part corresponds to the near-field perception detection part mentioned earlier. The reasoning part is exemplified by several aspects, including an operational data analysis agent, an activity generation agent, and an operational copywriting generation agent. Based on the identified different intents, the corresponding agent can be triggered and invoked. Based on the agent's reasoning, corresponding decision-making schemes are generated, such as promotional copy, decision-making prompts, and activity suggestions. These decision-making schemes are then executed through the action part, for example, providing one-click configuration of loudspeakers and posters to achieve the execution of the decision-making schemes.

[0156] The above specific solutions enable intelligent and precise near-field operation recommendation capabilities, which help merchants operate better and more efficiently, improve user experience, and enhance the effective utilization of internet resources.

[0157] Based on the same approach, one or more embodiments of this specification also provide another payment transaction auxiliary processing method, applied to near-field cash register devices, see [link to documentation]. Figure 8 It focuses on the process of near-field perception. Figure 8 This is a flowchart illustrating the method.

[0158] Figure 8 The process includes the following steps:

[0159] S802: Obtain a near-field perception model by fine-tuning a specified basic recognition model using upward-looking samples, wherein the upward-looking samples include real-life photos taken by a camera at an upward-looking angle.

[0160] S804: Using the near-field perception model, perform a rough detection on the image data dynamically acquired by the camera of the near-field cash register to determine whether a local limb other than the head has entered the near-field range.

[0161] S806: If so, then for the local limb, suppress the generation of the first fine detection box.

[0162] S808: During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0163] S810: Based on the near-field perception results, recommend service assistance information to the user.

[0164] This can be understood based on the previous explanation. Figure 8 The process and corresponding effects are not detailed here.

[0165] Based on the same idea, one or more embodiments of this specification also provide apparatus and devices corresponding to the above methods, such as... Figures 9-12 As shown. The apparatus and equipment are capable of performing the above methods and related alternatives accordingly.

[0166] Figure 9 This is a schematic diagram of a payment transaction auxiliary processing device provided in one or more embodiments of this specification, applied to a near-field cash register device. The device includes:

[0167] The near-field perception and detection module 902 performs near-field perception and detection of the user based on sensor data;

[0168] The selection result acquisition module 904 acquires the store selection result provided by the server of the near-field POS device if it detects that a user has entered the near-field range of the near-field POS device.

[0169] The recommendation model triggering module 906 determines whether to enable the auxiliary recommendation model based on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0170] If so, the auxiliary information recommendation module 908 generates corresponding business auxiliary information for the user before the user completes payment, and recommends it to the user.

[0171] Optionally, the sensor data includes image data dynamically acquired by the camera of the near-field POS device, wherein the camera is adapted to the flat position of the near-field POS device and acquires the image data from an upward viewing angle;

[0172] The near-field perception detection module 902 acquires a near-field perception model by fine-tuning a specified basic recognition model using upward-looking samples, wherein the upward-looking samples include real-life photos taken by a camera at an upward-looking angle.

[0173] Using the near-field perception model, the image data dynamically acquired by the camera of the near-field cash register is roughly detected to determine whether a local limb other than the head has entered the near-field range.

[0174] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0175] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0176] Optionally, before generating the first fine detection box to detect the head, the near-field sensing detection module 902 determines whether the size of the fine detection box to be generated conforms to the set size of the false alarm object.

[0177] If so, the fine detection bounding box to be generated will be filtered out and will not be used for actual detection.

[0178] Optionally, with the image data having at least 320 pixels and no more than 640 pixels on each side, the size of the false alarm object includes at least one of the following two:

[0179] The ratio of the longer side to the shorter side is greater than 2 pixels, and the shorter side is less than 60 pixels;

[0180] The shorter side is greater than 260 pixels.

[0181] Optionally, the near-field perception and detection module 902 determines whether a user with payment intent has entered the near-field range based on the detection result of the head, and generates corresponding user profile data;

[0182] If the auxiliary information recommendation module 908 is such that the inhibition is lifted, a second fine detection box is generated for the local limbs other than the head.

[0183] The second fine detection box is used to dynamically detect the local limbs. Based on the detection results of the second fine detection box, the timing for recommending the corresponding business auxiliary information is determined before the user completes the payment, so that recommendations can be made based on the user profile data at the recommended timing.

[0184] Optionally, the near-field POS device is connected to the store's basic POS device, which in turn extends the basic POS device's ability to support one or more other payment methods.

[0185] The recommendation to the user specifically includes:

[0186] Synchronize the first business status information from the basic POS device;

[0187] Based on the first business status information and the second business status information of the near-field POS device itself, it is determined whether the recommendation timing conflicts with the predetermined business sensitive status. If so, the recommendation timing is re-determined to avoid the conflict.

[0188] Optionally, the near-field POS device is connected to the store's basic POS device, which further expands the basic POS device's support for one or more other payment methods;

[0189] Before performing near-field perception detection on the user based on sensor data, the auxiliary information recommendation module 908 receives the basic business recommendation information transmitted from the basic POS device in response to the operation performed by the store clerk on the basic POS device. The basic business recommendation information, along with the store's customer flow profile data, is also sent to the server of the near-field POS device across servers through the server of the basic POS device.

[0190] The server receiving the near-field POS device trains and distributes the auxiliary recommendation model based on operational data provided by multiple stores. The operational data includes basic business recommendation information and store customer flow profile data.

[0191] Optionally, the store selection result is determined by the server of the near-field POS device based on business-based recommendation information and customer flow profile data provided by multiple stores.

[0192] Optionally, the recommendation model triggering module 906 determines whether the store to which the near-field POS device belongs is included in the store selection results;

[0193] If so, the leading merchant corresponding to the store selection result is determined, wherein the leading merchant is a merchant that is sufficiently relevant to the designated leading operation activity, and the business auxiliary information will be generated according to the leading operation activity through the auxiliary recommendation model;

[0194] Obtain conflict risk data between the store and the leading merchant or the leading operational activity;

[0195] If the conflict risk data reflects a sufficiently high conflict risk, then it is determined that the auxiliary recommendation model will not be activated.

[0196] Optionally, the auxiliary information recommendation module 908 obtains the basic fatigue level calculated by the server of the near-field POS device based on the store's transaction characteristics;

[0197] Based on the baseline fatigue level and the fatigue auxiliary data of the near-field POS device, it is determined whether to temporarily suspend further recommendations to the user.

[0198] Optionally, the basic fatigue level is calculated by the server of the near-field POS device based on the user's recent dynamic trajectory obtained from multiple different stores in the same store selection result.

[0199] Optionally, the business support information includes business promotion information;

[0200] The auxiliary information recommendation module 908 recommends information to the user in one or more forms, including at least voice broadcast.

[0201] Figure 10 This is a schematic diagram of another payment transaction auxiliary processing device provided in one or more embodiments of this specification, applied to a near-field cash register device. The device includes:

[0202] The near-field perception model acquisition module 1002 acquires a near-field perception model obtained by fine-tuning a specified basic recognition model using upward-looking samples, wherein the upward-looking samples include real-life photos taken by a camera at an upward-looking angle.

[0203] The near-field perception model detection module 1004 uses the near-field perception model to perform a rough detection on the image data dynamically collected by the camera of the near-field cash register to determine whether a local limb other than the head has entered the near-field range.

[0204] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0205] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0206] The business assistance information recommendation module 1006 recommends business assistance information to the user based on the near-field perception results.

[0207] Figure 11 This is a schematic diagram of the structure of a payment transaction auxiliary processing device provided in one or more embodiments of this specification, applied to a near-field cash register device. The payment transaction auxiliary processing device includes:

[0208] At least one processor; and,

[0209] A memory communicatively connected to the at least one processor; wherein,

[0210] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform:

[0211] Based on sensor data, perform near-field perception and detection of users;

[0212] If a user is detected entering the near-field range of the near-field POS device, the store selection result provided by the server of the near-field POS device is obtained;

[0213] Whether to enable the auxiliary recommendation model depends on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0214] If so, before the user completes the payment, the auxiliary recommendation model generates corresponding business assistance information for the user and recommends it to the user.

[0215] Figure 12 This is a schematic diagram of the structure of a payment transaction auxiliary processing device provided in one or more embodiments of this specification, applied to a near-field cash register device. The payment transaction auxiliary processing device includes:

[0216] At least one processor; and,

[0217] A memory communicatively connected to the at least one processor; wherein,

[0218] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform:

[0219] A near-field perception model is obtained by fine-tuning a specified basic recognition model using upward-view samples, wherein the upward-view samples include real-life photos taken by a camera at an upward angle.

[0220] Using the near-field perception model, the image data dynamically acquired by the camera of the near-field cash register is roughly detected to determine whether a local limb other than the head has entered the near-field range.

[0221] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0222] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0223] Based on the near-field perception results, service assistance information is recommended to the user.

[0224] Based on the same idea, one or more embodiments of this specification also provide a non-volatile computer storage medium for use in near-field payment devices, wherein the medium stores computer-executable instructions, the computer-executable instructions being configured as follows:

[0225] Based on sensor data, perform near-field perception and detection of users;

[0226] If a user is detected entering the near-field range of the near-field POS device, the store selection result provided by the server of the near-field POS device is obtained;

[0227] Whether to enable the auxiliary recommendation model depends on whether the store to which the near-field POS device belongs is included in the store selection results. The auxiliary recommendation model is trained by the server of the near-field POS device based on the operational data provided by multiple stores and then distributed to the near-field POS device.

[0228] If so, before the user completes the payment, the auxiliary recommendation model generates corresponding business assistance information for the user and recommends it to the user.

[0229] One or more embodiments of this specification also provide another non-volatile computer storage medium for use in near-field point-of-sale (POS) devices, the medium storing computer-executable instructions configured as follows:

[0230] A near-field perception model is obtained by fine-tuning a specified basic recognition model using upward-view samples, wherein the upward-view samples include real-life photos taken by a camera at an upward angle.

[0231] Using the near-field perception model, the image data dynamically acquired by the camera of the near-field cash register is roughly detected to determine whether a local limb other than the head has entered the near-field range.

[0232] If so, then for the local limb, the generation of the first fine detection box is suppressed;

[0233] During the suppression process, if it is determined that the head has completely entered the near field range, a first fine detection box is generated to detect the head, and near field perception of the user is performed based on the detection result of the head.

[0234] Based on the near-field perception results, service assistance information is recommended to the user.

[0235] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0236] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0237] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0238] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0239] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0240] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0241] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0242] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0243] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0244] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0245] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0246] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0247] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0248] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0249] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0250] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A payment service auxiliary processing method applied to a near-field cash register device, the method comprising: detecting, based on sensor data, whether a user enters a near field of the near-field cash register device; if it is detected that a user enters the near field of the near-field cash register device, obtaining a store circle selection result provided by a server of the near-field cash register device; determining whether to enable an auxiliary recommendation model according to whether a store to which the near-field cash register device belongs is included in the store circle selection result, wherein the auxiliary recommendation model is trained by the server of the near-field cash register device according to operation data provided by a plurality of stores and is distributed to the near-field cash register device; if yes, generating, by the auxiliary recommendation model, corresponding service auxiliary information for the user before the user completes payment, and recommending the service auxiliary information to the user. The sensor data includes image data dynamically collected by a camera of the near-field cash register device. The detection of whether a user enters a near field based on sensor data specifically comprises: obtaining a near field perception model obtained by fine-tuning a specified basic recognition model using a downward-looking sample, wherein the downward-looking sample includes a photo of a real person taken by the camera at a downward-looking angle; performing rough detection on the image data dynamically collected by the camera of the near-field cash register device using the near field perception model to determine whether a local limb other than a head enters the near field; if yes, suppressing generation of a first fine detection frame for the local limb; during the suppression, if it is determined that the head has completely entered the near field, generating a first fine detection frame to detect the head, and performing near field perception of the user according to a detection result of the head.

2. The method of claim 1, wherein the camera is adapted to a flat posture of the near-field cash register device and collects the image data at a downward-looking angle.

3. The method of claim 1, wherein before the generation of the first fine detection frame to detect the head, the method further comprises: determining whether a size of a fine detection frame to be generated conforms to a set false alarm object size; if yes, filtering out the fine detection frame to be generated without actually using it for detection.

4. The method of claim 3, wherein under a condition that each side pixel of the image data is not less than 320 pixels and not higher than 640 pixels, the false alarm object size includes at least one of the following two conditions: a long side / short side is greater than 2 pixels, and the short side is less than 60 pixels; and the short side is greater than 260 pixels.

5. The method of claim 1, wherein the near field perception of the user according to the detection result of the head specifically comprises: determining, according to the detection result of the head, whether a user with a payment intention enters the near field, and generating corresponding user portrait data; the generation, by the auxiliary recommendation model, of corresponding service auxiliary information for the user before the user completes payment, and the recommendation of the service auxiliary information to the user specifically comprises: if yes, removing the suppression and generating a second fine detection frame for the local limb other than the head. The second fine detection frame is used to dynamically detect the local limb, and a recommendation opportunity for the corresponding business auxiliary information is determined according to a detection result of the second fine detection frame before the user completes payment, so that the recommendation is made according to the user portrait data at the recommendation opportunity.

6. The method of claim 1 or 5, wherein the near-field cash register device is connected to a basic cash register device of a store, and the near-field cash register device additionally extends the support capability of one or more other payment methods to the basic cash register device. The recommendation to the user specifically includes: Synchronizing first business state information from the basic cash register device; According to the first business state information and second business state information of the near-field cash register device itself, it is judged whether the recommendation opportunity conflicts with a predetermined business sensitive state, and if so, the recommendation opportunity is re-determined to avoid the conflict.

7. The method of claim 1, wherein the near-field cash register device is connected to a basic cash register device of a store, and the near-field cash register device additionally extends the support capability of one or more other payment methods to the basic cash register device. Before the near-field sensing detection of the user based on the sensor data, the method further includes: Receiving the business basic recommendation information transmitted from the basic cash register device in response to the operation of providing the business basic recommendation information performed by the store clerk on the basic cash register device, wherein the business basic recommendation information and the customer flow portrait data of the store are also sent to the service end of the near-field cash register device across the service end through the service end of the basic cash register device. Receiving that the service end of the near-field cash register device trains and issues the auxiliary recommendation model according to the operation data provided by multiple stores, wherein the operation data includes business basic recommendation information and customer flow portrait data of stores.

8. The method of claim 1 or 7, wherein the store circle selection result is determined by the service end of the near-field cash register device according to the business basic recommendation information and the customer flow portrait data provided by multiple stores.

9. The method of claim 8, wherein the determination of whether to enable the auxiliary recommendation model according to whether the store to which the near-field cash register device belongs is included in the store circle selection result specifically includes: Determining whether the store to which the near-field cash register device belongs is included in the store circle selection result; If so, determining a dominant merchant corresponding to the store circle selection result, wherein the dominant merchant is a merchant with a high enough correlation with a specified dominant operation activity, and the business auxiliary information is generated by the auxiliary recommendation model according to the dominant operation activity; Obtaining conflict risk data between the store and the dominant merchant or the dominant operation activity; If the conflict risk data reflects that the conflict risk is high enough, it is determined that the auxiliary recommendation model is not started.

10. The method of claim 7, wherein the generation of the corresponding business auxiliary information for the user by the auxiliary recommendation model and the recommendation to the user further include: Obtaining a basic fatigue degree calculated by the service end of the near-field cash register device according to store transaction characteristics; According to the basic fatigue degree and fatigue degree auxiliary data local to the near-field cash register device, it is determined whether to temporarily suspend the recommendation to the user.

11. The method of claim 10, wherein the basic fatigue degree is calculated by a server of the near-field cash register device based on the user's recent dynamic track obtained from a plurality of different stores in the same store circle selection result.

12. The method of claim 1, wherein the service auxiliary information comprises service promotion information. The recommendation to the user specifically comprises: recommending to the user in one or more forms including at least voice broadcast.

13. A payment service auxiliary processing method applied to a near-field cash register device, the method comprising: obtaining a near-field perception model obtained by fine-tuning a specified basic recognition model using a downward-looking sample, the downward-looking sample comprising a photo of a real person taken by a camera at a downward-looking angle; using the near-field perception model to perform rough detection on image data dynamically collected by a camera of the near-field cash register device to determine whether a local limb other than a head enters a near-field range; if yes, suppressing generation of a first fine detection frame for the local limb; in the process of the suppression, if it is determined that the head has completely entered the near-field range, generating a first fine detection frame to detect the head, and performing near-field perception on a user according to a detection result of the head; according to the near-field perception result, recommending service auxiliary information to the user.

14. A payment service auxiliary processing device applied to a near-field cash register device, the device comprising: a near-field perception detection module that performs near-field perception detection on a user based on sensor data; a circle selection result acquisition module that acquires a store circle selection result provided by a server of the near-field cash register device if it is detected that a user has entered a near-field range of the near-field cash register device; a recommendation model triggering module that determines whether to enable an auxiliary recommendation model according to whether a store to which the near-field cash register device belongs is included in the store circle selection result, wherein the auxiliary recommendation model is trained by the server of the near-field cash register device according to operation data provided by a plurality of stores and is distributed to the near-field cash register device; an auxiliary information recommendation module that, if yes, generates corresponding service auxiliary information for the user by the auxiliary recommendation model before the user completes payment and recommends the service auxiliary information to the user; the sensor data comprises image data dynamically collected by a camera of the near-field cash register device; the near-field perception detection module acquires a near-field perception model obtained by fine-tuning a specified basic recognition model using a downward-looking sample, the downward-looking sample comprising a photo of a real person taken by a camera at a downward-looking angle; the near-field perception model is used to perform rough detection on the image data dynamically collected by the camera of the near-field cash register device to determine whether a local limb other than a head enters a near-field range; if yes, suppressing generation of a first fine detection frame for the local limb; In the process of the inhibition, if it is judged that the head has completely entered the near field range, a first fine detection frame is generated to detect the head, and a near field awareness of the user is performed according to the detection result of the head.

15. The apparatus of claim 14, wherein the camera is adapted to a flat posture of the near field cash register device, and collects the image data by using a downward view angle.

16. The apparatus of claim 14, wherein the near field awareness detection module, before the first fine detection frame is generated to detect the head, judges whether the size of the fine detection frame to be generated conforms to a set false alarm object size. If yes, the fine detection frame to be generated is filtered out and not actually used for detection.

17. The apparatus of claim 16, wherein, under the setting that each side pixel of the image data is not less than 320 pixels and not higher than 640 pixels, the false alarm object size includes at least one of the following two: a long side / short side greater than 2 pixels, and a short side less than 60 pixels; a short side greater than 260 pixels.

18. The apparatus of claim 14, wherein the near field awareness detection module, according to the detection result of the head, judges whether a user with a payment intention has entered the near field range, and generates corresponding user portrait data. The auxiliary information recommendation module, if yes, releases the inhibition, and generates a second fine detection frame for a local body part other than the head. The local body part is dynamically detected by using the second fine detection frame, and according to the detection result of the second fine detection frame, a recommendation time of the corresponding service auxiliary information is determined before the user completes payment, so as to recommend at the recommendation time according to the user portrait data.

19. The apparatus of claim 14 or 18, wherein the near field cash register device is connected with a basic cash register device of the store, and additionally extends support capability of one or more other payment methods for the basic cash register device. The recommendation to the user specifically includes: synchronizing first service state information from the basic cash register device; judging whether the recommendation time conflicts with a predetermined service sensitive state according to the first service state information and second service state information of the near field cash register device itself, and if yes, re-determining the recommendation time to avoid the conflict.

20. The apparatus of claim 14, wherein the near field cash register device is connected with a basic cash register device of the store, and additionally extends support capability of one or more other payment methods for the basic cash register device. the auxiliary information recommendation module, before the near field sensing detection on the user based on the sensor data, receives the service-based recommendation information transferred from the base cash register device in response to the operation of providing the service-based recommendation information performed by the store clerk on the base cash register device, wherein The service basic recommendation information and the store customer flow portrait data are also sent to the service end of the near field cash register device by the service end of the basic cash register device in a cross-service end manner. The service end of the near field cash register device receives the auxiliary recommendation model trained and distributed according to operation data provided by multiple stores, and the operation data includes service basic recommendation information and store customer flow portrait data. 21.The apparatus of claim 14 or 20, wherein the store circle selection result is determined by a server of the near field cash register based on business basis recommendation information and customer flow portrait data provided by a plurality of stores. 22.The apparatus of claim 21, wherein the recommendation model triggering module determines whether a store to which the near field cash register belongs is included in the store circle selection result. If yes, a dominant merchant corresponding to the shop circle selection result is determined, wherein, The dominant merchant is a merchant with a high enough correlation with a specified dominant operation activity, and the business auxiliary information is to be generated by the auxiliary recommendation model based on the dominant operation activity. Obtain conflict risk data between the store and the dominant merchant or the dominant operation activity. If the conflict risk data reflects a high enough conflict risk, it is determined that the auxiliary recommendation model is not started. 23.A payment business auxiliary processing apparatus applied to a near field cash register, the apparatus comprising: A near field perception model obtaining module obtains a near field perception model obtained by fine-tuning a specified basic recognition model using a downward-looking sample, wherein the downward-looking sample includes a photo of a real person taken by a camera at a downward-looking angle. A near field perception model detection module uses the near field perception model to roughly detect image data dynamically collected by a camera of the near field cash register to determine whether a local body part other than a head has entered a near field range. If yes, the generation of a first fine detection frame is suppressed for the local body part. During the suppression, if it is determined that the head has completely entered the near field range, a first fine detection frame is generated to detect the head, and the user is perceived based on the detection result of the head. A business auxiliary information recommendation module recommends business auxiliary information to the user based on the near field perception result. 24.A payment business auxiliary processing device applied to a near field cash register, the payment business auxiliary processing device comprising: At least one processor; And A memory in communication connection with the at least one processor; wherein The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform: Based on sensor data, a near field perception detection is performed on a user. If it is detected that a user has entered a near field range of the near field cash register, a store circle selection result provided by a server of the near field cash register is obtained. Based on whether a store to which the near field cash register belongs is included in the store circle selection result, it is determined whether to enable an auxiliary recommendation model, wherein the auxiliary recommendation model is trained by a server of the near field cash register based on operation data provided by a plurality of stores and is delivered to the near field cash register. If yes, before the user completes payment, the auxiliary recommendation model is used to generate corresponding business auxiliary information for the user and recommend it to the user. The sensor data includes image data dynamically collected by a camera of the near field cash register. The near field perception detection based on the sensor data specifically includes: The near-field perception model is obtained by fine-tuning a specified base recognition model by using a downward-looking sample, and the downward-looking sample includes a real person photo taken by a camera at a downward-looking angle; the image data dynamically collected by the camera of the near-field cash register device is coarsely detected by using the near-field perception model to determine whether a local limb other than a head enters a near-field range; if yes, the generation of a first fine detection box is inhibited for the local limb; in the process of the inhibition, if it is judged that the head has completely entered the near-field range, the first fine detection box is generated to detect the head, and the user is perceived according to the detection result of the head.

25. A payment service auxiliary processing device applied to a near-field cash register device, the payment service auxiliary processing device comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform: obtaining a near-field perception model obtained by fine-tuning a specified base recognition model by using a downward-looking sample, and the downward-looking sample includes a real person photo taken by a camera at a downward-looking angle; coarsely detecting image data dynamically collected by the camera of the near-field cash register device by using the near-field perception model to determine whether a local limb other than a head enters a near-field range; if yes, the generation of a first fine detection box is inhibited for the local limb; in the process of the inhibition, if it is judged that the head has completely entered the near-field range, the first fine detection box is generated to detect the head, and the user is perceived according to the detection result of the head; according to the near-field perception result, the user is recommended service auxiliary information.

Images