Intelligent welcoming and guest soliciting system based on human visual feature recognition

Abstract

The application discloses an intelligent welcome and guest inviting system based on human visual feature recognition and belongs to the technical field of artificial intelligence store intelligent reception. The core of the application comprises a data acquisition module, a visual recognition module, a control module and an audio playing module, the data acquisition module continuously acquires video image data of a passerby passing area at a store entrance, the visual recognition module recognizes the gender and age of a passerby through pure visual recognition of human body posture features such as facial features, hairstyle, shape and posture proportion, the control module identifies at least a passerby scene according to a passerby track and a stay state and outputs a guest inviting trigger instruction, and the audio playing module completes intelligent guest inviting voice broadcast of a passerby. The application can further expand to realize functions such as in-store welcome, out-of-store send-off, intelligent voice consultation and offline operation. The application discards a traditional voiceprint recognition scheme, has strong anti-environmental interference capability, initiatively invites a passerby through a pure visual core scheme and solves technical pain points such as single scene of an existing welcome device, incapability of active flow guiding, susceptibility to environmental interference and simple function reduction avoidance. The application has wide protection range, strong anti-avoidance property and is suitable for all kinds of offline physical stores for all-weather intelligent flow guiding reception.

Description

Technical Field

[0001] This invention belongs to the field of intelligent store reception and artificial intelligence visual recognition technology. Specifically, it relates to an intelligent welcoming and customer-attracting system based on human visual feature recognition, which is applicable to the intelligent reception operations of offline physical stores in all scenarios, including intelligent customer attraction of passersby, welcoming of customers arriving at the store, and seeing off of customers leaving the store. Background Technology

[0002] The customer traffic generation and daily customer service of physical stores directly impact their operational efficiency and revenue conversion. Traditional stores rely on manual staff to attract and greet customers, which results in high labor costs, limited working hours, and inconsistent service standards, making it easy to miss customers or delay in providing timely service.

[0003] Existing smart greeting devices have limited functionality, mostly providing fixed voice announcements for customers entering and leaving the store. They lack the ability to proactively identify and attract passersby at the store entrance, resulting in a large untapped potential customer base and low conversion rates. Furthermore, some existing smart devices rely on voiceprint recognition to determine user attributes, which is highly susceptible to interference from noisy environments, crowds, and multiple people speaking simultaneously. This leads to poor recognition stability, a high false trigger rate, and an inability to adapt to complex store operating environments.

[0004] In addition, the existing welcoming equipment has a single recognition dimension, which cannot accurately distinguish the gender and age characteristics of pedestrians. The broadcast messages are highly homogenized, making it impossible to achieve refined and differentiated reception. Moreover, it has poor light adaptability, and recognition fails in backlight, low light, and low light environments at night, making it impossible to achieve stable operation in all weather conditions.

[0005] In summary, existing technologies have many shortcomings, such as the inability to actively attract customers, weak identification and anti-interference capabilities, incomplete scene coverage, lack of differentiated reception, and poor environmental adaptability. They cannot meet the needs of modern physical stores for intelligent, refined, and 24 / 7 customer attraction and reception. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an intelligent welcoming and customer-attracting system based on human visual feature recognition. This system solves the technical pain points of traditional equipment, such as the inability to actively attract customers from passersby, reliance on voiceprint recognition which is easily interfered with, limited scene coverage, and lack of differentiated reception. It enables intelligent unmanned reception in all scenarios, including actively attracting passersby to the store, welcoming customers upon entry, and seeing customers off upon departure.

[0007] To achieve the above technical objectives, the present invention adopts the following technical solution: A smart welcoming and customer-attracting system based on human visual feature recognition includes a data acquisition module, a visual recognition module, a control module, and an audio playback module. The data acquisition module is used to continuously collect real-time video image data of the store entrance area, and the image data at least covers the pedestrian passage area at the store entrance. The visual recognition module is used to analyze human features in the collected real-time video image data. By fusing facial features and human body posture features, it identifies the gender and age range of pedestrians. The human body posture features include hairstyle features, body outline, and body proportion appearance features. It completes pedestrian attribute recognition entirely based on pure visual images. The control module is electrically connected to the visual recognition module and the audio playback module, respectively. It is used to receive pedestrian gender and age group information output by the visual recognition module, and combine the pedestrian movement trajectory and dwelling state to at least identify the scene of pedestrians passing by the store entrance and output a customer attraction trigger command. The audio playback module is used to receive trigger commands from the control module, match corresponding voice content according to the pedestrian's gender and age group, and play it in real time to realize intelligent voice broadcasting for attracting customers from pedestrians.

[0008] Furthermore, the control module can also identify customers entering and leaving the store, and output corresponding welcome and farewell trigger commands. The audio playback module will then execute differentiated voice broadcasts for welcoming customers entering and farewell customers leaving.

[0009] Furthermore, the data acquisition module includes a wide-angle image acquisition device, which is arranged above the store entrance to cover the area in front of the store, the entrance area, and the exit area. It is equipped with an adaptive supplementary lighting function, which can stably acquire clear human images in complex lighting environments such as low light, backlight, and low light at night.

[0010] Furthermore, the visual recognition module employs a facial feature key point extraction combined with a human body contour feature fusion algorithm. Through a preset gender feature model and age group stratification model, it comprehensively compares and judges the pedestrian's facial features and body features, accurately outputting the pedestrian's gender and the age group range of youth, middle-aged, and elderly.

[0011] Furthermore, the control module has a built-in scene judgment logic unit that tracks pedestrian movement trajectory, dwell time, and direction of entry and exit through continuous frame images, distinguishing between pedestrians who pass by without stopping, pedestrians who stop to look around, pedestrians entering the store, and pedestrians leaving the store, so as to achieve precise trigger control for different scenarios.

[0012] Furthermore, the system also includes an audio generation module, which has a built-in scene dialogue adaptation unit. It can receive instructions from the control module and automatically generate differentiated audio for attracting customers, welcoming guests, and seeing them off based on the pedestrian's gender, age group, and scene type. It also supports custom editing of exclusive voice content based on the store's business type, event information, and store style.

[0013] Furthermore, the system also includes an intelligent dialogue module, which includes a speech-to-text unit, a semantic parsing unit, and an intelligent response unit. It can recognize customers' daily inquiries, product questions, and event inquiries, accurately extract inquiry keywords, match knowledge base content to generate natural, conversational responses, and achieve unmanned intelligent customer service in stores.

[0014] Furthermore, the system also includes a storage module, which stores human feature recognition models, age and gender matching rules, scene classification judgment logic, voice broadcast materials, script generation rules, store Q&A knowledge base, and system operating parameters. The storage module can de-identify the collected image data, retaining only the recognition feature results and not storing the original face images, thus preventing the leakage of customer privacy.

[0015] Furthermore, the system also includes a backend management module, which supports remote or local parameter configuration. It allows users to customize parameters such as pedestrian solicitation sensitivity, dwell time determination threshold, broadcast interval, script content, and age group classification standards. It also supports knowledge base updates, operation log viewing, and equipment fault diagnosis.

[0016] Furthermore, the system as a whole adopts local edge AI computing. All human feature recognition, scene judgment, and voice dispatch logic are completed on the local device without relying on cloud servers, making it suitable for various offline physical stores to carry out unattended intelligent customer attraction and reception scenarios around the clock. Detailed Implementation

[0017] The present invention will be further described in detail below with reference to specific embodiments.

[0018] This embodiment is a preferred implementation of the present invention, which is an intelligent customer-attracting system based on human visual feature recognition. It is applied to offline physical stores and relies on pure visual AI human feature recognition technology to realize intelligent customer attraction for passersby at the store entrance. At the same time, it can be expanded to realize the functions of welcoming customers entering the store, seeing off customers leaving the store, and intelligent voice consultation interaction. No voiceprint recognition assistance is required throughout the process. It has strong anti-interference ability and wide applicability.

[0019] The data acquisition module employs a wide-angle, high-definition image acquisition device, fixedly installed above the store entrance in a central position. The shooting angle at least covers the pedestrian traffic area outside the store, and can be extended to cover both the entrance and exit areas. The device is equipped with an adaptive supplementary lighting module that monitors ambient light levels in real time via photosensitive elements and automatically adjusts the supplementary lighting intensity. It can acquire clear and stable pedestrian video images in complex lighting conditions, including backlighting during the day, low light on cloudy days, and complete darkness at night, ensuring the accuracy of subsequent visual recognition.

[0020] The visual recognition module is equipped with a trained multi-feature fusion AI recognition model. It continuously receives real-time video streams from the data acquisition module and extracts multi-dimensional visual features such as facial key points, hairstyle features, body contours, and body proportions frame by frame. Through comprehensive comparison and analysis using a built-in gender determination model and age group stratification model, it accurately distinguishes the male or female gender attributes of pedestrians and their age groups (youth, middle-aged, and elderly). This invention adopts a pure visual recognition solution throughout, eliminating the need to collect voice signals and completely avoiding recognition failures and false triggers caused by store environment noise and mixed voices, thus significantly improving recognition stability.

[0021] As the core scheduling unit of the system, the control module receives pedestrian gender and age group data from the visual recognition module in real time. Combined with continuous video frame tracking of pedestrian movement trajectories and dwell times, it prioritizes identifying pedestrians passing by the store entrance, triggering corresponding intelligent customer-attracting broadcast commands to proactively guide potential customer traffic. Simultaneously, the control module can further identify customers entering and leaving the store, triggering welcome and farewell broadcast commands respectively, achieving differentiated customer service across all scenarios.

[0022] The audio playback module receives various trigger commands from the control module and matches exclusive voice content to complete real-time broadcasts based on the pedestrian's gender, age group, and corresponding scenario. It outputs lightweight customer-attracting scripts to passersby to encourage them to stop and enter the store; personalized greetings to customers entering the store; and thank-you and farewell messages to departing customers, thus refining the store's customer service experience and improving customer conversion rates.

[0023] As a preferred expansion method, this system can be configured with an audio generation module, which supports customizing the script rules according to the store's business type, offline promotional activities, and store characteristics, and automatically generating exclusive voice audio that is suitable for different groups of people and different scenarios. There is no need for manual pre-recording of audio, and the adaptability and iteration convenience are higher.

[0024] As a preferred expansion method, this system can be configured with an intelligent dialogue module. After customers enter the store, they can initiate inquiries by voice. The system completes voice-to-text conversion, semantic parsing, keyword matching, and retrieves the built-in question-and-answer knowledge base to generate natural, conversational replies, replacing manual answers to basic inquiries and reducing the pressure on store staff.

[0025] As a preferred expansion method, this system can be configured with a storage module to locally store AI recognition models, scene judgment logic, broadcast materials, and question-and-answer knowledge bases. This supports independent offline operation of devices without relying on cloud networks, avoiding device failure caused by network lag or outages. Simultaneously, the collected image data undergoes real-time anonymization processing, retaining only the recognition results data and not storing original facial or human images, strictly protecting user privacy and complying with data security standards.

[0026] The backend management module supports local and remote dual-mode operation and maintenance management. Administrators can customize and adjust the sensitivity of pedestrian recognition, dwell time, voice broadcast interval, age group classification standards, and various scenario scripts. At the same time, they can view the device operation logs in real time, troubleshoot device faults, and update the question and answer knowledge base, making device operation and maintenance extremely convenient.

[0027] The system adopts local AI computing on the device side. All recognition, scene judgment and voice scheduling logic are completed on the local device. It has a fast response speed and low running latency, and can work stably 24 / 7. It is suitable for the all-weather intelligent customer attraction and reception needs of various offline physical stores such as catering, retail, clothing, and beauty. Beneficial effects of the invention

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. Unique core capability for proactively attracting customers from passersby, blocking and simplifying circumvention paths: This invention establishes passerby recognition and proactive customer attraction at the store entrance as core essential functions, with welcoming guests upon entry and seeing them off upon departure as optional extension functions. Whether competitors only focus on attracting customers from passersby or also cover all scenarios, they all fall entirely within the scope of this patent protection, completely blocking the industry's common practice of competitors reducing functions to circumvent infringement, creating an extremely high barrier to legal action.

[0030] 2. Pure visual multi-feature fusion recognition with strong anti-interference ability: Abandoning the traditional voiceprint recognition solution, it adopts multi-dimensional visual feature fusion recognition of face, hairstyle and body shape, which completely avoids the recognition failure problem caused by noisy store environment and human voice interference. With the adaptive supplementary light structure, it can be adapted to all weather and all lighting complex scenes, and the recognition stability and accuracy are greatly improved.

[0031] 3. Refined and differentiated customer service to improve store conversion: It can accurately identify the gender and age characteristics of pedestrians and match exclusive personalized words for different scenarios such as attracting passers-by, welcoming customers into the store, and thanking customers upon leaving the store. It gets rid of the drawbacks of the single mechanical broadcast of traditional equipment, enhances the brand quality of the store and the customer consumption experience, and effectively improves the store's customer traffic conversion rate.

[0032] 4. Modular and optional architecture with extremely wide protection scope: The core protection architecture only retains the essential modules for attracting customers on the street, while all other value-added functions are optional extension features. There are no restrictions on chip models, specific algorithms, or hardware brands. Competitors cannot avoid infringement by changing hardware, simplifying configurations, or removing additional functions.

[0033] 5. Stable offline operation and privacy and security compliance: Supports local offline computing without relying on cloud networks, ensuring stable operation without downtime; image data is anonymized throughout the process, with no original privacy data retained, meeting data security and compliance requirements.

[0034] 6. Cost reduction and efficiency improvement, wide adaptability: It can realize 24 / 7 uninterrupted unattended intelligent customer attraction and reception, replacing basic manual reception work, significantly reducing store operating costs, and adapting to the intelligent upgrade needs of all categories of offline physical stores.

Claims

1. An intelligent welcoming and customer-attracting system based on human visual feature recognition, characterized in that, It includes a data acquisition module, a visual recognition module, a control module, and an audio playback module; The data acquisition module is used to continuously collect real-time video image data of the store entrance area, and the image data at least covers the pedestrian passage area at the store entrance. The visual recognition module is used to analyze human features in the collected real-time video image data. By fusing facial features and human body posture features, it identifies the gender and age range of pedestrians. The human body posture features include hairstyle features, body outline, and body proportion appearance features. It completes pedestrian attribute recognition entirely based on pure visual images. The control module is electrically connected to the visual recognition module and the audio playback module, respectively. It is used to receive pedestrian gender and age group information output by the visual recognition module, and combine the pedestrian movement trajectory and dwelling state to at least identify the scene of pedestrians passing by the store entrance and output a customer attraction trigger command. The audio playback module is used to receive trigger commands from the control module, match corresponding voice content according to the pedestrian's gender and age group, and play it in real time to realize intelligent voice broadcasting for attracting customers from pedestrians.

2. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: The control module can also identify customers entering and leaving the store, and output corresponding welcome and farewell trigger commands. The audio playback module executes differentiated voice broadcasts for welcoming customers entering the store and farewell customers leaving the store.

3. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: The data acquisition module includes a wide-angle image acquisition device, which is placed above the store entrance to cover the area in front of the store, the entrance area, and the exit area. It is equipped with an adaptive supplementary lighting function, which can stably acquire clear human images in complex lighting environments such as low light, backlight, and low light at night.

4. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: The visual recognition module uses a facial feature key point extraction combined with human body contour feature fusion algorithm. Through preset gender feature model and age group stratification model, it comprehensively compares and judges the facial features and body features of pedestrians, and accurately outputs the pedestrian's gender and the age group range of youth, middle age and old age.

5. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: The control module has a built-in scene judgment logic unit. By tracking the pedestrian's movement trajectory, dwell time, and direction of entry and exit through continuous frame images, it can distinguish between pedestrians who are passing by without stopping, pedestrians who are standing and looking around, pedestrians entering the store, and pedestrians leaving the store, so as to achieve precise trigger control for different scenarios.

6. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: It also includes an audio generation module, which has a built-in scene dialogue adaptation unit. It can receive instructions from the control module and automatically generate differentiated audio for attracting customers, welcoming guests, and seeing them off based on the pedestrian's gender, age group, and scene type. It also supports custom editing of exclusive voice content based on the store's business type, event information, and store style.

7. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: It also includes an intelligent dialogue module, which comprises a speech-to-text unit, a semantic parsing unit, and an intelligent response unit. This module can recognize customers' daily inquiries, product questions, and event inquiries, accurately extract inquiry keywords, match them with knowledge base content to generate natural, conversational responses, and achieve unmanned intelligent customer service in stores.

8. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: It also includes a storage module, which is used to store human feature recognition models, age and gender matching rules, scene classification judgment logic, voice broadcast materials, script generation rules, store Q&A knowledge base, and system operating parameters; the storage module can de-identify the collected image data, retain only the recognition feature results, and not store the original face image, thus preventing the leakage of customer privacy.

9. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: It also includes a back-end management module, which supports remote or local parameter configuration. Users can customize the sensitivity of soliciting customers from passersby, the threshold for determining dwell time, the broadcast interval, the script content, and the age group classification standards. It also supports knowledge base updates, viewing operation logs, and troubleshooting equipment malfunctions.

10. The intelligent welcoming and customer-attracting system based on human visual feature recognition according to claim 1, characterized in that: The system adopts local edge AI computing. All human feature recognition, scene judgment and voice dispatch logic are completed on the local device without relying on cloud servers. It is suitable for various offline physical stores to carry out unattended intelligent customer attraction and reception scenarios around the clock.