Search

CN-122024371-A - Intelligent tuning method and system for vending machine camera

CN122024371ACN 122024371 ACN122024371 ACN 122024371ACN-122024371-A

Abstract

The application discloses an intelligent tuning method and system for a vending machine camera. The method comprises the steps of continuously collecting images of a front area of the vending machine through a camera, outputting image streams, reading the image streams, preprocessing and model reasoning, outputting personnel status signals and distance signals, inquiring volume mapping relation according to the personnel status signals and the distance signals, generating volume control instructions, receiving the volume control instructions, adjusting sound pressure of a loudspeaker, transmitting actual volume sampling values back to a core control step to form closed loop correction, finally providing a parameter configuration interface of the volume mapping relation, and storing configuration results to support subsequent inquiry, so that volume management is completed.

Inventors

  • LIU PANPAN
  • HUANG AIHUA
  • YIN JUEHUI

Assignees

  • 上海趣致网络科技有限公司

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1.A vending machine camera intelligent tuning method, the method comprising: S1, collecting by a camera, continuously collecting an image of a region in front of the vending machine by the camera, and outputting an image stream; S2, image analysis and identification, namely reading the image stream, and outputting personnel state signals and distance signals after preprocessing and model reasoning; s3, core control, namely inquiring a volume mapping relation according to the personnel state signal and the distance signal, and generating a volume control instruction; S4, loudspeaker driving, namely receiving the volume control instruction, adjusting the sound pressure of the loudspeaker, and transmitting the actual volume sampling value back to the core control step to form closed loop correction; And S5, man-machine interaction, namely providing a parameter configuration interface of the volume mapping relation, and storing configuration results to support subsequent inquiry so as to complete volume management.
  2. 2. The vending machine camera intelligent tuning method of claim 1, wherein the S1 camera acquisition specifically comprises: The wide dynamic camera with resolution not lower than 1280 multiplied by 720 and frame rate of 1-8 fps is arranged at the middle part or the top of the front of the vending machine, so that the visual angle of the wide dynamic camera covers the sector area of 1.5-5 m in front, the camera adopts color imaging in daytime, black and white imaging is carried out at night through built-in infrared light supplement, and an image stream is output through a USB or RJ45 interface.
  3. 3. The intelligent tuning method for the vending machine camera according to claim 1, wherein in the step of S2 image analysis and recognition, the preprocessing step specifically includes sequentially performing noise reduction, defogging and edge enhancement on each frame of image to eliminate abrupt light changes, shadows and glass reflection interference, and outputting a standardized image for model reasoning.
  4. 4. The intelligent tuning method for a vending machine camera according to claim 1, wherein in S2 image analysis and recognition, the model reasoning step specifically includes: And adopting an improved YOLOv-nano model to forward infer the standardized image, outputting three state signals of no person, no person staying and no person operating, and returning to the Euclidean distance between the bottom edge center of the human body boundary frame and the vending machine, and controlling the inference result through an internal bus in a JSON frame format.
  5. 5. The intelligent tuning method of a vending machine camera according to claim 1, wherein the volume control command is generated by inquiring a volume mapping relation according to the personnel status signal and the distance signal, and the method specifically comprises the following steps: Pre-storing a five-level volume mapping table, wherein the state priority in the table is abnormal alarm > operation > stay > no person, the corresponding volumes are 80%, 60-70%, 30-40% and 10% of the maximum volume respectively, and when in inquiry, the reference volume is determined according to the state label, and then linear interpolation within +/-10% is carried out according to the distance interval, so that a final PWM duty ratio instruction is generated.
  6. 6. The intelligent tuning method of a vending machine camera of claim 5, wherein the five-level volume mapping table specifically comprises: when the target area is unmanned and the loudspeaker is in a playing state, immediately reducing the volume to 10% of the maximum volume, and closing the output after 3 minutes of unmanned operation; when people stay for no operation and the distance is within the range of 1.5-5m, the volume is adjusted to 30-40%, and if the people stay for more than 10 seconds, the people are automatically lifted to 50%; when the operation action of the personnel is detected, the volume is increased to 60-70%, and the volume is maintained in the payment or selecting stage; When the hand-extending goods taking action is identified, the volume is reduced to 40%, and a goods taking prompt is played; when the payment fails or the commodity card is abnormal, the volume is temporarily increased to 80% to play the warning voice.
  7. 7. The vending machine camera intelligent tuning method of claim 1, wherein the S4 speaker driving step specifically comprises: The method comprises the steps of receiving PWM volume control instructions sent by a core control step through a TDA2030 amplifier, driving a 2-5W loudspeaker through adjusting output gain, connecting an amplifier output loop in series with a sampling resistor, returning a current value to the core control step through an ADC in real time, converting actual power by the core control step according to P=I2R, and correcting a duty ratio in the next PWM period if the deviation from the target power exceeds 5%, so as to form hardware closed loop correction.
  8. 8. The vending machine camera intelligent tuning method of claim 1, wherein S5 human-machine interaction specifically comprises: in a system setting interface of a touch screen of the vending machine, three adjustable parameters of default volume when a person operates, standby volume when no person operates and delay closing time are provided, a worker inputs a modification value through touch, after the modification is finished, a core control step writes the new parameter into a nonvolatile memory and directly calls the new parameter when the next mapping relation is queried, and site updating and persistence of volume mapping logic are realized.
  9. 9. A vending machine camera intelligent tuning system, the system comprising: The camera acquisition module is used for continuously acquiring images of the area in front of the vending machine through the camera and outputting image streams; the image analysis and identification module is used for reading the image stream and outputting personnel state signals and distance signals after preprocessing and model reasoning; The core control module is used for inquiring the volume mapping relation according to the personnel state signal and the distance signal and generating a volume control instruction; The loudspeaker driving module is used for receiving the volume control instruction, adjusting the sound pressure of the loudspeaker and transmitting the actual volume sampling value back to the core control step to form closed loop correction; And the man-machine interaction module is used for providing a parameter configuration interface of the volume mapping relation and storing configuration results to support subsequent inquiry so as to complete volume management.
  10. 10. An electronic device comprising a memory and a processor, the memory storing a computer program that when executed by the processor implements the intelligent tuning method for a vending machine camera of any one of claims 1 to 8.

Description

Intelligent tuning method and system for vending machine camera Technical Field The application relates to the technical field of intelligent retail equipment and automatic control intersection, in particular to an intelligent tuning method and system for a vending machine camera. Background At present, the main stream intelligent vending machine speaker function in the market is mainly used for playing voice contents such as commodity introduction, purchasing operation guidance, successful payment prompt, promotion and propaganda, and the volume adjustment mode of the intelligent vending machine speaker function has obvious limitation: Fixed volume mode, that is, most vending machines adopt preset fixed volume from factory and cannot be adjusted according to the conditions of on-site personnel. When the vending machine is in a dense area (such as a mall hall and a school canteen), the fixed volume can be covered up due to environmental noise, so that consumers can hear unclear operation guide, and in a sparse area (such as an office building corridor and a district corner), the fixed volume can form unnecessary noise to disturb the surrounding environment. The manual adjustment has the defects that few vending machines with the volume adjustment function require staff to manually modify volume parameters through physical keys or background remote, the adjustment frequency is low, the flexibility is poor, and the real-time matching of the change of field staff is impossible. For example, people before vending machines in the morning and evening peak of the workday concentrate, and need higher volume, and when no one is in the morning, the speaker still maintains high volume to play, resulting in electric energy waste and noise pollution. The simple induction technology is insufficient, a part of vending machines try to judge that people are close by adopting an infrared induction module, but the infrared induction has the problems of short detection distance (usually not more than 2 meters), easiness in interference of ambient light (such as direct irradiation of strong light and dim environment), high false triggering rate (such as passing of pets and shielding of articles) and the like, and whether a potential consumer cannot be accurately identified, so that the volume adjustment time deviation is caused, and the use experience is influenced. The prior art can not realize that the volume of a loudspeaker is dynamically and accurately adjusted according to the actual situation of field personnel of the intelligent vending machine, so that the clear acquisition requirement of consumers on voice prompts is not met, and energy waste and hidden danger of noise pollution exist. Disclosure of Invention Based on the above, the embodiment of the application provides an intelligent tuning method and system for a camera of a vending machine, which are used for acquiring images of a target area around the vending machine in real time through the camera, accurately judging whether personnel stay in the area, personnel distance and operation state are combined with an image recognition algorithm, further automatically controlling the opening, closing, increasing or decreasing of the volume of a loudspeaker, realizing the dynamism, intellectualization and accuracy of volume adjustment, improving the definition of operation guidance of consumers, reducing energy consumption and noise pollution in unmanned time periods, optimizing the service efficiency of the intelligent vending machine, and solving the problems of fixed volume, poor adjustment flexibility, low induction detection precision and the like of the loudspeaker of the existing intelligent vending machine. In a first aspect, a vending machine camera intelligent tuning method is provided, the method comprising: S1, collecting by a camera, continuously collecting an image of a region in front of the vending machine by the camera, and outputting an image stream; S2, image analysis and identification, namely reading the image stream, and outputting personnel state signals and distance signals after preprocessing and model reasoning; s3, core control, namely inquiring a volume mapping relation according to the personnel state signal and the distance signal, and generating a volume control instruction; S4, loudspeaker driving, namely receiving the volume control instruction, adjusting the sound pressure of the loudspeaker, and transmitting the actual volume sampling value back to the core control step to form closed loop correction; And S5, man-machine interaction, namely providing a parameter configuration interface of the volume mapping relation, and storing configuration results to support subsequent inquiry so as to complete volume management. Optionally, the S1 camera collects, specifically includes: The wide dynamic camera with resolution not lower than 1280 multiplied by 720 and frame rate of 1-8 fps is arranged at the middle part or the top of the front of t