A top-down imaging illumination device, a method and system for acquiring and controlling dining images.
By designing a top-down imaging illumination device with the lens and light source arranged coaxially, and an image acquisition control method that is confirmed by the server after local testing, the problems of viewing angle adjustment accuracy, glare, uneven shadows, and privacy leakage in the existing technology are solved, achieving uniform light source shadows, equipment security, and image acquisition accuracy.
Patent Information
- Application Number
- CN202611527308.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-29
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies for image acquisition of desktop areas at fixed heights and angles suffer from problems such as reduced accuracy due to viewing angle adjustments, glare, uneven shadows, heat from light sources affecting the lifespan of electronic components and the stability of image sensors, device misjudgments, and privacy leaks. Furthermore, there is a lack of effective image acquisition control methods.
Design an overhead imaging illumination device with a lens and light source arranged coaxially, a light-blocking structure cutting off the light path, and a control unit that detects and generates a probe signal locally. Image uploading is only initiated after confirmation by the server, separating perception and decision-making to avoid the device making its own judgments and continuously transmitting data.
It achieves uniform light source shadows, avoids lens glare, and effectively dissipates heat from electronic components, reducing the computing power requirements of the device, avoiding misjudgments and privacy leaks, and ensuring the accuracy and security of image acquisition.
Smart Images

Figure CN123075256A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of lighting and image acquisition technology, and in particular to a top-down imaging lighting device installed above a desktop for acquiring top-down images of a desktop area from a fixed perspective, as well as a method and system for controlling the acquisition of dining images using the device. Background Technology
[0002] Recording the daily diet of family members is the foundation of nutrition management and health care. There are two main types of existing dietary record-keeping methods: one relies on users to manually take photos or input data, which is difficult to maintain in the long term, and the shooting distance, angle, and lighting vary, making it difficult to use the images for quantitative analysis; the other relies on wearable devices or special tableware, which has limited coverage and makes it difficult to record the meal situations of multiple family members at the same time.
[0003] Existing technologies already include solutions that combine cameras with light bulbs. US patents US8562158B2 and US9228731B2 disclose devices for mounting cameras inside screw-in LED light bulbs, where the camera is retractable, rotatable, or its orientation can be adjusted via a ball joint; US patent US9172917B1 discloses a webcam powered by a light bulb socket, where the lens's orientation is adjusted via a flexible hose connector; PCT application WO2009129664A1 discloses a camera device with a lens embedded in the center of a spotlight cup and a ring of LEDs mounted on the inner wall of the cup; US patent US9497421B2 discloses a camera light with multiple LEDs distributed around a lens, the lens being substantially flush with the surface of the light source surrounding it.
[0004] In applications requiring repeated close-range image acquisition of a tabletop area directly below a light fixture (e.g., tableware and food on a dining table) at a fixed height and angle, and the identification of object outlines and estimation of dimensions, the applicant has identified the following unresolved issues: First, the relationship between the angle of view and the tabletop changes with each adjustment of the adjustable camera, necessitating recalibration. Furthermore, perspective distortion occurs when the angle is not perpendicular, affecting the accuracy of estimations using objects of known size on the tabletop as a reference. Second, when the light-emitting element is close to the lens, the arrangement of the lens flush with the light source surface or the lens offset allows light from the light-emitting element to directly enter the lens, creating glare and reducing image contrast. Third, when the lens images through a lens cover, the scattering of light from the lens and stains on the cover are superimposed on the image. Fourth, when the light source and lens are not on the same axis, the shadow of the subject is biased to one side, resulting in uneven illumination distribution. Fifth, the power of the light source inside the bulb is much greater than the power of the imaging and communication electronics. The heat from the light source is conducted to the imaging and communication electronics within the enclosed casing, shortening the lifespan of electronic components and affecting the operational stability of the image sensor.
[0005] Furthermore, security cameras either continuously record and upload video streams, resulting in large data volumes, high power consumption, and a persistent threat to home privacy; or they may perform motion detection and record only when triggered. For dining records, having the device automatically determine and initiate recording means a single misjudgment could lead to unwarranted image capture; while continuously transmitting images to a server for further judgment is both bandwidth and privacy-incompatible. Current technology has not yet provided a capture control method that avoids complex judgments on the device, continuous image transmission to the server, and prevents image sequence capture from being initiated by misjudgment.
[0006] To address this, we propose an overhead imaging lighting device, a method and system for controlling the acquisition of dining images. Summary of the Invention
[0007] The main objective of this invention is to provide an overhead imaging lighting device, a method and system for capturing and controlling dining images, in order to solve the problems mentioned in the background art.
[0008] To achieve the above objectives, the present invention proposes a top-down imaging lighting device, comprising a mounting part, a housing, a light source assembly, a camera module, a control unit, and a wireless communication module; the mounting part is located at one end of the housing and is used to fix the device to a mounting surface above a table and to draw power; the light source assembly is located on the side of the housing opposite to the mounting part, and includes a lamp plate and multiple light-emitting elements located on the side of the lamp plate opposite to the mounting part; the camera module is located at the end of the housing opposite to the mounting part, and the lens of the camera module is located in the area surrounded by the multiple light-emitting elements, with the optical axis of the lens pointing away from the mounting part; a distance is provided between the lens and the light-emitting elements. The device has a light-shielding structure that protrudes from the light-emitting surface of the light-emitting element along the optical axis and surrounds the lens, so that the straight line between the light-emitting surface of any light-emitting element and the light-incident surface of the lens is interrupted by the light-shielding structure. The control unit and the wireless communication module are located inside the housing and electrically connected to the camera module. The control unit is configured to: detect whether there is a person in front of the table locally on the device based on the downsampled image output by the camera module, generate a start test signal or end test signal accordingly, and report it to the server via the wireless communication module. The control unit will only open or close the upload channel of the image captured by the camera module after receiving confirmation of the corresponding test signal from the server.
[0009] Optionally, the mounting part is a lamp holder, which is used to screw into or insert into the lamp socket to draw power and to fix the device to the lamp socket.
[0010] Optionally, the mounting part is a ceiling mount, an embedded downlight mounting structure, or a hanging structure, and the device is powered by a power cord or an external power adapter.
[0011] Optionally, the optical axis of the lens coincides with the central axis of the device, and the direction of the optical axis of the lens is not adjustable relative to the housing.
[0012] Optionally, the system further includes a heat sink and a lens housing. The heat sink is disposed on the side of the housing opposite to the mounting portion and is connected to the housing. The lamp plate is attached to the side of the heat sink opposite to the mounting portion. The lamp plate has a central through hole at its center. The lens housing is located on the side of the lamp plate opposite to the mounting portion and is disposed opposite to the central through hole. The lens housing is fixed relative to the lamp plate. The camera module is fixed inside the lens housing. The light-shielding structure is a light-shielding section of the lens housing that protrudes from the light-emitting surface.
[0013] Optionally, it also includes a lampshade, which is disposed on the side of the light-emitting element away from the mounting portion. The bottom wall of the lampshade is provided with a through hole, and the end of the light-shielding section extends into the through hole, so that the imaging light path of the lens does not pass through the light-transmitting wall of the lampshade.
[0014] Optionally, the control unit and the wireless communication module are mounted on a functional motherboard. The functional motherboard is located on the side of the heat sink near the mounting portion and has a gap between it and the heat sink. A heat insulation layer is provided in the gap. The outer peripheral wall of the housing has multiple lateral openings on the side of the heat insulation layer near the mounting portion. The lateral openings communicate with the space where the functional motherboard is located. The end of the housing near the mounting portion has multiple heat dissipation grooves. An airflow channel is formed between the lateral openings and the heat dissipation grooves through the inner cavity of the heat insulation layer near the mounting portion. The heat insulation layer separates the airflow channel from the space where the heat sink and the lamp panel are located.
[0015] Optionally, the control unit is further configured to: when no person is detected in the detection of the entire downsampled image, divide the downsampled image into multiple overlapping sub-regions, detect each sub-region separately, and merge the detection results of each sub-region with the detection results of the entire image to remove duplicates.
[0016] This invention also proposes another overhead imaging lighting device, including a mounting part, a housing, a lamp plate, multiple light-emitting elements, a lens housing, a camera module, and a lampshade; the mounting part is located at one end of the housing and is used to fix the device to a mounting surface above a table and to draw power; the lamp plate is located on the side of the housing opposite to the mounting part, and the multiple light-emitting elements are located on the side of the lamp plate opposite to the mounting part and distributed around the center of the lamp plate; the lamp plate has a central through hole in its center, the lens housing is located on the side of the lamp plate opposite to the mounting part and is opposite to the central through hole, the lens housing is fixed relative to the lamp plate, and the camera module... Fixed within the lens housing, the lens of the camera module is located within the area surrounded by the plurality of light-emitting elements, and the optical axis of the lens faces away from the mounting portion. The lens housing is an opaque component and has a light-shielding section that protrudes from the light-emitting surface of the light-emitting element along the optical axis. The light-shielding section surrounds the lens, so that the straight line between the light-emitting surface of any of the light-emitting elements and the light-incident surface of the lens is interrupted by the light-shielding section. The lampshade is disposed on the side of the light-emitting element away from the mounting portion, and the bottom wall of the lampshade has a through hole. The end of the light-shielding section extends into the through hole, so that the imaging light path of the lens does not pass through the light-transmitting wall of the lampshade.
[0017] Optionally, the optical axis of the lens coincides with the central axis of the device and is not adjustable relative to the housing; the mounting part is a lamp holder, a ceiling mount, an embedded downlight mounting structure, or a hanging structure.
[0018] This invention also proposes a method for controlling the acquisition of dining images executed by an acquisition device. The method includes: continuously acquiring downsampled images output by the camera module and locally detecting whether there is someone at the table in front of the acquisition device, wherein the continuous acquisition and local detection are not affected by the state of the image upload gate; generating a start test event when someone is detected, and generating an end test event when the number of consecutive determinations of no one reaches a preset threshold; reporting the test signal corresponding to the test event to the server via the wireless communication module; opening the image upload gate only when the server confirms the start test signal, and closing the image upload gate only when the server confirms the end test signal, wherein unconfirmed test signals do not change the state of the image upload gate; and uploading full-frame images for meal analysis one by one to the storage path specified by the server for this meal only when the image upload gate is open, and not uploading full-frame images for meal analysis when the image upload gate is closed.
[0019] Optionally, the newly generated start trial event and end trial event are strictly alternated; after the image upload gate is closed, a process must be performed to determine that no one is present for a certain number of consecutive times before a start trial event can be generated again; the encoding of the whole frame image used for meal analysis is only performed during the period when the image upload gate is open.
[0020] Optionally, the probe signal may include at most one frame of the image from the camera module at that time for the server to make a decision; the downsampled image used for the local detection is not stored or uploaded.
[0021] Optionally, when a probe signal transmission fails, the type of the probe event is recorded, and the probe signal is resent when a corresponding detection result is obtained again. The transmission failure includes the request not being delivered to the server, the request being delivered but no response being received within a preset time, and the received response being unparseable. If the detection result is reversed before resentment, the probe event to be resent is cleared and not resent. When the server explicitly refuses to start a probe signal, no retry is made until the number of consecutive times no one is detected reaches the preset threshold.
[0022] Optionally, after receiving confirmation of the start probe signal from the server, if opening the image upload gate fails, the meal number returned by the server is saved, and the image upload gate is retried using the meal number in the next probe event, without reporting a new start probe signal.
[0023] Optionally, the image-by-image upload includes: assigning a monotonically increasing and non-repeating sequence number to each image, the sequence number being persisted before transmission, and the sequence number of images that are not successfully uploaded being disused and not re-uploaded; immediately clearing local temporary files on the acquisition device after upload and verification are completed; the local detection also includes: when no person is detected in the detection of the entire downsampled image, dividing the downsampled image into multiple overlapping sub-regions for separate detection and merging to remove duplicates.
[0024] This invention also proposes a server-executed method for controlling the acquisition of dining images, comprising: receiving a start probe signal or an end probe signal reported by a acquisition device installed above the table, wherein the probe signal is generated by the acquisition device based on local detection of whether there is someone at the table; and adjudicating the received probe signal, the adjudication including at least: for the start probe signal, if the acquisition device already has an ongoing meal, returning the meal number without creating a new meal; if the acquisition device already has a meal record to be confirmed, reusing that record; for the end probe signal, if the meal it points to does not exist... If the request is rejected, and the meal it points to has already ended, only confirmation is returned without repeating the termination; only one meal record to be confirmed is allowed to exist in the same acquisition device at any given time; a confirmation is returned to the acquisition device for a successful test signal, so that the acquisition device can open or close the image upload gate accordingly; when the test signal is confirmed, a dedicated storage path is issued for this meal and provided to the acquisition device; images uploaded by the acquisition device to the storage path are received; before starting the recognition and analysis of the images of this meal, the image list under the storage path is frozen so that one analysis corresponds to one definite input.
[0025] Optionally, the decision also includes semantic decision: the image attached to the probe signal is given to the recognition unit to determine whether the table is in a dining state, and the judgment result is awaited within a preset waiting time; if no judgment result is obtained within the preset waiting time or the recognition unit is unavailable, the decision is confirmed.
[0026] Optionally, the storage path is an independent directory in the object storage, organized by meal. The image is directly transmitted from the acquisition device to the object storage. The independent directory is appended but not modified or deleted. The freezing includes: generating a snapshot of the independent directory containing the object key, content identifier, and size. The entire identification and analysis process is based on the snapshot. If the snapshot changes during the analysis, the analysis is deemed to have failed.
[0027] Optionally, during the continuous image uploading by the acquisition device, identification and analysis are submitted in batches. The input for each batch is an explicitly listed list of object keys. The server maintains a cursor pointing to the last successfully analyzed object key, and the next batch starts after the cursor. Only one batch is allowed in transit at any given time. If a batch fails, the cursor does not advance, and the objects of that batch are merged into the next batch. Each batch returns the complete state of the entire meal up to the present, and the server completely overwrites the state of the previous batch with it. The input for the whole meal analysis submitted at the end of the meal includes the sequence number of the last successful batch and the list of object keys after the cursor that have not yet been overwritten by any batch.
[0028] Optionally, the identity feature map and process map generated by the identification analysis are uploaded to a path separate from the independent directory; when the identification analysis fails, the current meal is marked and the reason for the failure is recorded, the images in the independent directory remain unchanged and are not automatically retried, and the analysis is performed again by a new analysis explicitly initiated and an independent record is generated.
[0029] The present invention also proposes a dining image acquisition system, including an acquisition device and a server installed above a table; the acquisition device includes a camera module, a control unit and a wireless communication module, the control unit is configured to perform the above-described method; the server is configured to perform the above-described method.
[0030] The technical solution of this invention has the following advantages: First, the device is fixed to the mounting surface above the desktop and draws power from the mounting surface. The lens is located within the area surrounded by the light-emitting element and faces the desktop. The light source and the lens are coaxial, and the shadow of the photographed object is evenly distributed. The light-blocking structure cuts off the straight light path between the light-emitting element and the lens, so the light from the light-emitting element cannot directly enter the lens. Second, the acquisition device only answers the question "Is there anyone at the table?" locally, without performing identity recognition or content recognition. The edge computing power requirement is extremely low, and the local criteria can be replaced as computing power evolves without affecting the upper-level logic. Third, perception and decision-making are separated: the edge only generates a probe signal, and the server decides whether it is valid based on the context. The transmission gate only opens or closes for confirmed signals. Therefore, the device does not have the ability to automatically start long-term acquisition. During the gate closure period, the meal images are not encoded or uploaded, and there is no need to continuously transmit images to the server. The server's structural layer adjudication prevents the same device from repeatedly opening and receiving meals. When semantic adjudication is enabled and a rejection result is returned, the corresponding probe will not open the upload gate. After the gate is closed, it must undergo a complete continuous no-person judgment before it is allowed to send a start probe signal again to avoid the same group of people being retried immediately after they finish. Transmission failure and service rejection are handled separately. Expired probes are not resent, and service rejections are not retried. The device and server will not generate a second meal due to repeated retrying. Fourth, detection and uploading are two independent paths. Edge detection uses downsampled small images, while uploading is the original full-frame image. Edge computing power constraints will not be transmitted as the upper limit of analysis quality. When the whole image is missed, the overlapping sub-region is used to re-push and retrieve small targets, which does not require a larger model or higher input resolution. Fifth, each meal has a separate, append-only storage directory; the sequence number is persisted and monotonically incremented before sending; and a snapshot containing object keys and content identifiers is frozen before identification, ensuring that each analysis corresponds to a traceable and definitive input. Batch analysis uses an explicit object key list with a cursor for continuation; failed batches are merged into the next batch without loss; each batch returns the complete state of the entire table rather than an increment; when collecting meals, the accumulated state is reused, and only the last segment is recalculated; analysis failures do not implicitly retry, and each analysis is an independent and traceable event; image cleanup by meal does not require tracking each image individually. Attached Figure Description
[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.
[0032] Figure 1 This is a schematic diagram of the overall structure of the overhead imaging illumination device according to Embodiment 1 of the present invention; Figure 2 This is an exploded view of the overhead imaging illumination device according to Embodiment 1 of the present invention; Figure 3 This is another exploded structural diagram of the overhead imaging illumination device according to Embodiment 1 of the present invention; Figure 4 This is another exploded structural diagram of the overhead imaging illumination device according to Embodiment 1 of the present invention; Figure 5 This is another exploded structural diagram of the overhead imaging illumination device according to Embodiment 1 of the present invention; Figure 6 This is a partially exploded structural diagram of the overhead imaging illumination device according to Embodiment 1 of the present invention; Figure 7 This is a partially exploded structural diagram of the overhead imaging illumination device according to Embodiment 2 of the present invention; Figure 8 This is a flowchart illustrating the meal judgment state machine and upload gate in Embodiment 4 of the present invention; Figure 9 This is a schematic diagram of the meal storage catalog and batch cursor in Embodiment 4 of the present invention; Figure 10 This is a cross-sectional schematic diagram along the optical axis of the lens of the overhead imaging illumination device according to Embodiment 1 of the present invention. Figure 11 for Figure 10 A partially enlarged schematic diagram of the mating parts of the lamp plate, heat sink, and lens housing.
[0033] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0035] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of the components in a specific posture (as shown in the attached figure). If the specific posture changes, the directional indication will also change accordingly. In this document, "up" refers to the side closer to the mounting part, and "down" refers to the side away from the mounting part; after the device is screwed into the top lamp socket, "down" refers to the direction towards the ground and the tabletop.
[0036] Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.
[0037] This invention proposes a top-down imaging lighting device, a method and system for acquiring and controlling dining images. Example 1
[0038] In this embodiment, the mounting part is a lamp holder, and the overall device is shaped like a light bulb. Figures 1 to 7 , Figure 10 as well as Figure 11 As shown, in this embodiment, the bulb-type camera lighting device includes an upper housing 101, a lamp holder 102, a heat sink 103, a lamp board 104, LED beads 105, a lens housing 106, a camera module 107, a lamp board bracket 108, a functional motherboard 109, a motherboard bracket 110, and a lamp cover 111. The lamp holder 102 is located at the upper end of the upper housing 101. The lamp holder 102 is a standard E27 or E26 screw-in lamp head, used to screw into an existing lamp socket to draw power, and to suspend and fix the entire device to the lamp socket. The heat sink 103 is a heat-conducting plate that is attached to the lamp plate 104. It is detachably connected to the lower end of the upper housing 101 by screws and is located below the upper housing 101. The lamp plate 104 is detachably attached to the lower end wall of the heat sink 103. A plurality of lamp beads 105 are disposed on the lower end wall of the lamp plate 104 and distributed around the center of the lamp plate 104 to form the light source assembly of the present invention. The lamp beads 105 are arranged in a ring, leaving an area in the center for mounting the lens housing 106.
[0039] The lamp panel 104 has a central through hole 1041. The lens housing 106 is located on the side of the lamp panel 104 away from the lamp holder 102 and is positioned opposite to the central through hole 1041. Its end near the lamp holder 102 is fixed to the lamp panel 104. The central through hole 1041 is distinct from the offset positioning hole through which the positioning post passes. The camera module 107 is fixed to the camera circuit board, which is fixed inside the lens housing 106 with screws and located below the light-emitting surface of the LED beads 105. The lens of the camera module 107 is located at the end of the lens housing 106 away from the lamp holder 102. Its optical axis coincides with the central axis of the lamp holder 102 and faces away from the lamp holder 102; that is, after the device is screwed into the top lamp opening, the lens is vertically downward and aligned directly below the tabletop. The optical axis direction of the lens is not adjustable relative to the upper housing 101, and the device always maintains a directly downward viewing angle after installation. Multiple LED beads 105 are distributed around the lens, making the light source coaxial with the lens. In this embodiment, the light-emitting end face of the LED 105 faces downward and is distributed in concentric inner and outer rings. The LEDs 105 in both the inner and outer rings are evenly distributed along the circumference. The lens is a wide-angle lens.
[0040] The lens housing 106 serves a light-shielding function. One end of it, facing away from the lamp holder 102, protrudes along the optical axis from the light-emitting surface of the lamp bead 105, forming a light-shielding section 1063 surrounding the lens. The term "lens incident surface" as used herein refers to the optical surface at the front of the lens into which imaging light enters, and is not replaced by the mechanical aperture of the lens housing or the outer diameter of the lens assembly. In this embodiment, the light-emitting end face of the lamp bead 105 is located on the outer surface of the bottom surface of the lamp cover 111, near the lamp holder 102. The end of the light-shielding section 1063 is flush with the outer surface, and the spherical cap at the front of the lens protrudes from the outer surface in a direction away from the lamp holder 102. The light-shielding section 1063 forms a solid partition between the light-emitting element and the lens incident surface. Its inner and outer contours and axial length are mutually matched, cutting off the straight optical path between the light-emitting surface of the light-emitting element and the lens incident surface, preventing the light from the light-emitting element from directly entering the lens along this straight optical path. The specific dimensions of the light-shielding section are set in conjunction with the actual boundary of the lens incident surface and the position of the light-emitting element.
[0041] The lens housing 106 has a light-shielding function and leaves a gap between it and the heat sink 103.
[0042] The lampshade 111 is detachably connected to the lower end of the heat sink 103 and covers the underside of the lamp bead 105, serving to even out the light for the lamp bead 105. The bottom wall of the lampshade 111 has a through hole 1111. In this embodiment, the through hole 1111 is a tapered hole with an inner diameter larger than the outer diameter. The end of the light-shielding section 1063 extends into the through hole 1111 and is flush with the outer surface of the bottom of the lampshade 111. The imaging light path of the lens does not pass through the light-transmitting wall of the lampshade 111 at all, thus avoiding the scattering of light from the lamp bead 105 by the light-transmitting wall and the influence of stains on the light-transmitting wall. At the same time, the cooperation between the through hole 1111 and the light-shielding section 1063 also serves to radially position the lampshade 111. Therefore, the lens housing 106 has three functions: its light-shielding section 1063 blocks the light path of the lamp bead 105 directly hitting the lens; its positioning cooperation with the heat sink 103 assists in positioning the lens housing 106 relative to the central axis of the lamp holder 102; and its end cooperates with the lamp cover through hole 1111 to position the lamp cover 111.
[0043] The lamp board bracket 108 is detachably mounted on the side of the heat sink 103 near the lamp holder 102. The functional main board 109 is detachably mounted on the side of the lamp board bracket 108 near the lamp holder 102. The main board bracket 110 is detachably mounted on the side of the functional main board 109 near the lamp holder 102. The functional main board 109 is equipped with a control unit and a wireless communication module. The control unit includes a processor and a memory. The memory stores a lightweight target detection program for detecting whether there is a person in front of the local detection table. The wireless communication module is a Wi-Fi module used to receive and upload images captured by the camera module 107. A gap is maintained between the functional main board 109 and the heat sink 103 by the lamp board bracket 108, and a heat insulation layer 119 is provided in the gap. In this embodiment, the heat insulation layer 119 includes a hollow annular heat insulation component located on the outer periphery and sandwiched between the heat sink 103 and the lamp plate bracket 108, and two semi-circular heat insulation components located in the center and overlapping to cover the area above the lens housing 106. The heat insulation layer 119 covers the gap between the heat sink 103 and the functional mainboard 109, blocking airflow between the heat sink 103 and the functional mainboard 109 and slowing down heat conduction to the functional mainboard 109. Gaps are left between the heat sink 103 and the lens housing 106, and between the central heat insulation component and the lens housing 106, so they do not directly contact each other.
[0044] In this embodiment, the heat generated by the LED chip 105 is transferred to the heat sink 103 via the lamp board 104. The heat sink 103 is a heat-conducting plate that is attached to the lamp board 104, and its function is to control the temperature of the lamp board 104 within the allowable range of the LED chip 105. Thus, a light source area with the lamp board 104 and the heat sink 103 as its core and an electronic area with the functional main board 109 as its core are formed in the device. The two are separated by a heat insulation layer 119, and the electronic area is independently cooled by the convection channel between the opening slot 1013 and the heat dissipation slot 1011.
[0045] The upper end of the upper housing 101 has multiple heat dissipation grooves 1011 arranged circumferentially around the lamp holder 102; the outer peripheral wall of the upper housing 101 has multiple opening slots 1013 arranged circumferentially, the opening slots 1013 being located on the side of the heat insulation layer 119 near the lamp holder 102 and communicating with the space where the functional main board 109 is located. Both the heat dissipation grooves 1011 and the opening slots 1013 communicate with the inner cavity of the upper housing 101 and the outside, and an airflow channel is formed between them through the inner cavity above the heat insulation layer 119; the heat insulation layer 119 separates this airflow channel from the space where the heat sink 103 and the lamp board 104 are located. After the device is screwed into the top lamp socket, the lamp holder 102 is on top. Outside air enters the electronic area where the functional motherboard 109 is located through the opening slot 1013. After being heated, it rises along the inner cavity of the upper housing 101 and is discharged through the heat dissipation slot 1011 near the lamp holder 102, forming natural convection and continuously dissipating the heat from the electronic area. The heat from the light source area is received by the heat sink 103, and the heat insulation layer 119 restricts its transfer to the electronic area. The lamp holder 102, the lamp board 104, and the camera module 107 are all electrically connected to the functional motherboard 109.
[0046] Specifically, it also includes ventilation fins 112, which are mesh-shaped and are respectively embedded in the opening slots 1013. While ensuring ventilation, they also serve to prevent dust, prevent foreign objects from entering, and provide decoration.
[0047] Specifically, a first positioning post 1081 is protruding from the lower end wall of the lamp panel bracket 108, and a first positioning hole 1031 corresponding to the first positioning post 1081 is provided at the upper end of the heat sink 103. The first positioning post 1081 is detachably embedded in the first positioning hole 1031, so that the assembly positioning between the lamp panel bracket and the heat sink is more accurate.
[0048] Specifically, the lens housing 106 has two radially opposing positioning protrusions 1061 at one end near the lamp holder 102, offset from the lens optical axis. The lamp plate 104 has two positioning grooves 1042 corresponding to the positioning protrusions 1061. The heat sink 103 has a central hole 1032 coaxial with the central through hole 1041. In this embodiment, the diameter of the central hole 1032 is larger than the diameter of the central through hole 1041. Figure 10 and Figure 11As shown, the positioning protrusion 1061 passes through the positioning groove 1042 and extends into the central hole 1032, with a fitting clearance between its outer surface and the hole wall of the central hole 1032; the positioning groove 1042 defines the circumferential position of the lens housing 106 relative to the lamp plate 104, and the hole wall of the central hole 1032 restricts the radial offset of the positioning protrusion 1061. During assembly, this positioning fit assists in positioning the lens housing 106 relative to the central axis of the lamp holder 102; the coincidence accuracy of the lens optical axis and the central axis of the lamp holder 102 is determined by the above-mentioned positioning fit, the screw fixing of the lens housing 106 and the lamp plate 104, and the machining and assembly tolerances of each part. The cylindrical body of the lens housing 106 is located on the side of the lamp plate 104 away from the lamp holder 102. Its end face near the lamp holder 102 abuts against the surface of the lamp plate 104 away from the lamp holder 102. The lens housing 106 is fixed to the lamp plate 104 by screws. Except for the positioning protrusion 1061, the lens housing 106 does not extend into the central through hole 1041, nor does it contact the heat sink 103.
[0049] Specifically, it also includes a first screw (not shown), and the lamp plate bracket 108, the heat sink 103 and the lamp plate 104 are respectively provided with a plurality of first screw holes 113 in the circumferential direction. The upper end of the lens housing 106 is provided with a first threaded hole 1062 corresponding to the first screw hole 113. The first screw passes through the first screw hole 113 and is screwed into the first threaded hole 1062.
[0050] Specifically, a third positioning post 1082 protrudes from the upper end of the lamp panel bracket 108, and the functional main board 109 is provided with a third positioning hole 1091 corresponding to the third positioning post 1082. The third positioning post 1082 is detachably embedded in the third positioning hole 1091, making the assembly positioning between the lamp panel bracket and the functional main board more precise.
[0051] Specifically, the lower end of the motherboard bracket 110 is provided with a fourth positioning post 1101, and the upper end of the functional motherboard 109 is provided with a fourth positioning hole 1092 corresponding to the fourth positioning post 1101. The fourth positioning post 1101 is detachably embedded in the fourth positioning hole 1092, so that the assembly positioning between the motherboard bracket and the functional motherboard is more accurate.
[0052] Specifically, it also includes a second screw (not shown). The main board bracket 110 and the functional main board 109 are provided with a plurality of second screw holes 114 in opposite directions along the circumference. The upper end of the lamp board bracket 108 is provided with a second threaded hole 1083 corresponding to the second screw holes 114. The second screws pass through the second screw holes 114 and are screwed into the second threaded holes 1083.
[0053] Specifically, it also includes a third screw (not shown). The radiator 103 is provided with a plurality of third screw holes 1033 along the circumferential direction. The lower end of the upper housing 101 is provided with a third threaded hole 1012 corresponding to the third screw holes 1033. The third screws are respectively inserted into the third screw holes 1033 and screwed into the third threaded holes 1012.
[0054] Specifically, the upper end of the lampshade 111 is provided with a plurality of first elastic buckle arms 1112 along the circumferential direction, and the lower end of the heat sink 103 is provided with a first buckle groove 1034 along the circumferential direction. The first elastic buckle arms 1112 are detachably and fixedly connected to the first buckle groove 1034, making the installation and removal of the lampshade and the heat sink more convenient and quick.
[0055] Specifically, it also includes a circuit board 115. The inner sidewall of the upper end of the upper housing 101 is provided with a slot 116 on each side. The two sides of the circuit board 115 are respectively embedded in the slot 116. The lamp holder 102, the lamp bead 105, the camera module 107 and the functional motherboard 109 are all electrically connected to the circuit board 115. The circuit board 115 converts the mains power obtained from the lamp holder 102 into the working voltage required by the lamp bead 105, the camera module 107 and the functional motherboard 109, making the assembly between the circuit board and the upper housing more convenient and quick, and also playing a limiting role for the circuit board.
[0056] Specifically, the camera module 107 has a wide-angle lens, and at the common height of the light fixture from the desktop, its field of view can cover the entire desktop area directly below. Example 2
[0057] The difference between this embodiment and Embodiment 1 is that: like Figure 7 As shown, a second elastic latching arm 117 protrudes from both sides of the motherboard bracket 110, and a second latching groove 118 is provided at the upper end of the functional motherboard 109. The second elastic latching arms 117 are detachably embedded in the second latching groove 118, making the installation and removal between the motherboard bracket and the functional motherboard more convenient and quick. Example 3
[0058] The difference between this embodiment and Embodiment 1 lies in the form of the mounting part. In this embodiment, the upper end of the upper housing 101 does not have a lamp holder 102, but instead has a ceiling mount, which is fixed to the ceiling with screws and connected to the mains power via a power cord; or, the upper end of the upper housing 101 has a mounting spring for an embedded downlight, and the device is embedded in an opening in the ceiling; or, the upper end of the upper housing 101 has a hanging structure, and the device is suspended above the dining table via a hanging line and powered by a power cord inside the hanging line. In any of the above forms, the lens of the camera module 107 is still located in the area surrounded by the LED beads 105 and faces the tabletop; the arrangement of the light source assembly, lens housing 106 and its light-shielding section 1063, lampshade 111 and its through hole 1111, and thermal partition structure are all the same as in Embodiment 1. In other embodiments, the optical axis of the lens can also be parallel to but not coincident with the central axis of the device, or at a small angle to it, as long as the lens is located in the area surrounded by the light-emitting element and the light-shielding section cuts off the straight light path between the light-emitting element and the lens. Example 4
[0059] This embodiment describes a dining image acquisition and control method and system using the above-described device, such as... Figure 8 and Figure 9 As shown. The core idea of this method is to separate perception from decision-making: the acquisition device only generates a probing signal indicating "may have started" or "may have ended," and the server decides whether it is valid. The image upload channel is only opened or closed for signals that have been approved.
[0060] Step S1, Local Detection. The control unit continuously acquires the downsampled images output by the camera module 107 and runs a lightweight target detection program locally on the device. This program only answers the question "Is there anyone at the table?" without performing identity verification or image content recognition. Camera output and local detection are uninterrupted during gate opening and closing; local detection is the only processing that runs when the gate is closed. The downsampled images are processed only in the control unit's memory, without being written to disk or uploaded. Only the encoding and uploading of the entire image frame are truly controlled by the gate. The local criteria can be replaced as the edge computing power evolves without affecting subsequent steps.
[0061] Step S2, Decision-Making State Machine. The control unit maintains four state variables: whether it is in the "occupied" state and has issued a start probe; the number of consecutive occupancy checks; whether it is in the "waiting for an empty table" state; and whether there are probes pending retransmission due to transmission failure. A decision is performed each time a result is obtained. When someone is detected, the occupancy count is reset to zero; if there is an pending end probe, it is cleared; if there is a pending start probe and it is not in the waiting for an empty table state, it is retransmitted; if it is currently in a meal or in the waiting for an empty table state, no probe is sent; otherwise, it is set to "in a meal" and a start probe signal is issued. When occupancy is determined, the occupancy count is incremented by one; no probe is sent if the threshold is not reached; if the threshold is reached, the count is reset to zero and the waiting for an empty table state is lifted; if there is an pending start probe, it is cleared; if there is an pending end probe, it is retransmitted; otherwise, if it is currently in a meal, it is set to "not in a meal" and an end probe signal is issued. Therefore, the newly generated start and end probe events strictly alternate; probe signals resent due to transmission failure correspond to the same undelivered event and do not constitute a new event. When the gate is closed, the state is reset to "waiting for empty table": after the gate is closed, a complete continuous unattended check must be performed before the next start probe is allowed to prevent the same group of people from triggering it again immediately after it ends.
[0062] Step S3: Sending the probe signal and server decision-making. The probe signal is reported to the server in the form of a request via the wireless communication module. The request includes the signal type (start or end), the meal number at the end, and an optional occurrence time. The request body can include a frame of on-site image at that time, or it can be empty. The server's decision-making is divided into two layers. The structural layer is always in effect: When a start probe signal is received, if the device already has an ongoing meal, it directly returns the same meal number without creating a new one; if there is a record to be confirmed, it reuses it. When an end probe signal is received, if the meal number does not exist, it is rejected; if the meal has ended, it only returns confirmation without repeating the end. A device is only allowed to have one meal record to be confirmed at any given time. A retransmitted start probe is considered a second inquiry rather than opening a second table. The semantic layer is optional: When enabled, the probe signal along with the on-site image is sent to the recognition unit to determine whether the table is in a dining state, and waits for the result within a preset waiting time. If the timeout occurs or the recognition unit is unavailable, it is processed as confirmation to prevent a meal from being lost due to slow decision-making. The meal judgment strategy is centralized on the server and can be evolved to the point where no device firmware updates are required.
[0063] Step S4, Gate Control and Recovery. The control unit opens the image upload gate only upon receiving confirmation of the start probe signal and closes the gate only upon receiving confirmation of the end probe signal. The gate is considered closed only after all uploaded images have been transmitted. Unconfirmed probe signals do not change the gate state. During gate closure, full-frame images used for meal analysis are not encoded or uploaded, and continuous outflow from the camera and local detection are unaffected by the gate. Two types of failures are handled separately: For transmission failures (request not delivered, request delivered but response timed out or lost, unsuccessful response, response cannot be parsed), the type of the probe is recorded, and it is immediately resent upon receiving the next same-direction detection result. If the state reverses during this period (the person has left while waiting for the resentment to begin, or the person returns after waiting for the resentment to end), the probe to be resent is cleared and not resent, because the facts it describes no longer hold true. Service rejection (receiving an explicit non-acknowledgment reply) is considered a successful communication, not a failure, and is not retried. Since the probe is already marked as "in progress" when it is sent, a person sitting there will not send another start probe; a new start probe will only be sent after a round of empty tables. If the gate opening fails after the start is confirmed (e.g., signature failure or camera not ready), the meal number returned by the server is temporarily stored. The next probe will directly retry opening the gate with this number without sending a new start probe, thus preventing a situation where one person sits down and two tables are set up. The device side uses synchronous requests, with one request corresponding to one response; timeout resending uses the same probing event, and the server makes idempotent decisions according to the above-mentioned structural layer rules; request timeout and single image upload timeout can be set separately.
[0064] Step S5, Upload. Upon confirming the start of the probe signal, the server issues a dedicated upload path for this meal session. This path is an independent directory in object storage named after the device number and meal number (e.g., capture_session / <device number> / meal-<timestamp>-<random segment> / raw / ). During the gate opening period, camera module 107 captures full-frame images at fixed intervals. Images are not cropped or scaled, and are directly uploaded to object storage without going through the server. Only one image is uploaded at a time, and the filename consists of a sequence number and the acquisition time. The sequence number is persisted before sending, monotonically increasing and non-repeating. Lost frame skip numbers are not re-uploaded, and timestamps are not allowed to be reversed. Uploads only append to this directory; objects already written are never modified or deleted. After upload and verification are completed, the device immediately cleans up local temporary files. The downsampled images used for local detection and the uploaded full-frame images come from two independent paths, and the edge computing power constraint will not be transmitted as an upper limit to the analysis quality.
[0065] Small-scale re-detection. When the device is suspended directly above the table, the person occupies a very small portion of the entire image, making it easy to miss detections during a single frame-by-frame inference. If no person is detected in the entire downsampled image, the control unit divides the image into several overlapping sub-regions and re-infers them separately, then merges them with the results of the entire frame to remove duplicates. The overlapping of sub-regions is necessary; otherwise, if a person happens to fall on the dividing line, they will be truncated from both sides simultaneously. The fact that the judgment is not sensitive to latency is a prerequisite for this approach to work.
[0066] Step S6, Freeze and Analyze. The whole-meal analysis freezes the entire raw / directory's object list: when the recognition unit starts execution, it generates a snapshot of the directory containing object keys, content identifiers (e.g., ETag), and sizes. The entire analysis is based on this snapshot. If the snapshot changes during execution, the entire analysis fails and the changed input is not used. After submitting the whole-meal analysis, the server does not add, delete, or overwrite any files in the directory. The batch analysis freezes not the directory but an explicit list of object keys: each batch's input directly lists the files to be analyzed in that batch. The recognition unit only looks at the objects in the list and does not create its own directory. The server maintains a cursor pointing to the last object key that has been submitted and successfully analyzed. The next batch starts listing from after the cursor. During analysis, the following situations are handled: new files uploaded by the device are included in the next batch; there are no re-uploads, modifications, or deletions; if a batch fails, the cursor does not advance, and the files in that batch are merged into the next batch for re-analysis; if a batch has not yet returned, the second batch is not submitted concurrently; only one batch is in transit at any given time. The identification unit maintains its own cumulative state, while the server is only responsible for providing the new files in the current batch and where the previous batch left off. Each batch returns the complete state of the entire meal so far, not an increment; the server completely overwrites the previous batch. When collecting meals, a complete meal analysis is submitted, with two additional items in the input: the sequence number of the last successful batch and a list of files after the cursor that have not been overwritten by any batch. The identification unit uses this information to reuse the cumulative state and complete the end segment. If no batch has been successful, the analysis of the entire directory is restarted from the beginning.
[0067] The identity feature map and process diagram generated by the identification analysis are uploaded to a path separate from the raw / directory (e.g., identity_crop / and process_image / under observe / run_<analysis number> / ). References to the process diagram are only retained in the audit log. When the whole meal analysis fails, the current meal is marked and the reason is recorded. Images in the directory remain unchanged and are not automatically retried; to retry, a new analysis must be explicitly initiated. Each analysis has an independently verifiable record.
[0068] Data is stored in separate domains according to its purpose: original images are stored in a single directory for each meal; identified products are stored in a separate path; local temporary files are cleaned up immediately; and running status, stage trajectory, de-identified context, verification, and retry records are archived for traceability. Because there is a single directory for each meal, cleaning can be processed as a whole for each meal, without the need for tracking each image individually; the retention period for original images and the cleaning trigger conditions are set according to product and compliance requirements.
[0069] The system in this embodiment includes the aforementioned data acquisition device and server. The server and the wireless communication module of the data acquisition device are connected via the Internet. The system may include one or more data acquisition devices within the same household, and the same server serving multiple households. Specific methods for identification and analysis are not within the scope of this invention. Other implementation methods
[0070] In other embodiments, the lamp holder 102 can be a B22 bayonet lamp holder, a GU10 lamp holder, or other standard lamp holders, as long as it can draw power from the lamp port and fix the device to the lamp port.
[0071] In other embodiments, the light-shielding structure may be an annular light-shielding ring disposed on the lamp plate 104 and surrounding the lens, or an annular flange protruding inward around the through hole 1111 on the lamp cover 111, as long as it is located between the lens and the lamp bead 105 and protrudes from the light-emitting surface along the optical axis.
[0072] In other embodiments, the upper housing 101 may also be provided with a working status indicator light to indicate to the user that the camera module 107 is in the acquisition state.
[0073] In other implementations, the local detection in step S1 can also be coarsely judged based on changes in screen brightness or moving areas, and then confirmed by target detection; the judgment rules or models used for semantic adjudication in step S3, as well as the presentation methods of the device and the user side after rejection, can be determined according to application requirements.
[0074] In other embodiments, the light-emitting element may also include an invisible light-emitting element for providing supplemental lighting for shooting when the illumination is off.
[0075] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.
Claims
1. A top-down imaging illumination device, characterized in that, The device includes a mounting section, a housing, a light source assembly, a camera module, a control unit, and a wireless communication module. The mounting section is located at one end of the housing and is used to fix the device to a mounting surface above a desktop and to draw power from it. The light source assembly is located on the side of the housing opposite to the mounting section and includes a lamp panel and multiple light-emitting elements located on the side of the lamp panel opposite to the mounting section. The camera module is located at the end of the housing away from the mounting portion. The lens of the camera module is located in the area surrounded by the plurality of light-emitting elements, and the optical axis of the lens faces away from the mounting portion. A light-shielding structure is provided between the lens and the light-emitting elements. The light-shielding structure protrudes from the light-emitting surface of the light-emitting element along the optical axis and surrounds the lens, so that the straight line between the light-emitting surface of any light-emitting element and the light-incident surface of the lens is interrupted by the light-shielding structure. The control unit and the wireless communication module are located inside the housing and electrically connected to the camera module. The control unit is configured to: detect whether there is a person in front of the table locally on the device based on the downsampled image output by the camera module, generate a start test signal or end test signal accordingly, and report it to the server via the wireless communication module. The control unit will only open or close the upload channel of the image captured by the camera module after receiving confirmation of the corresponding test signal from the server.
2. The apparatus according to claim 1, characterized in that, The mounting part is a lamp holder, which is used to screw into or insert into the lamp socket to draw power and to fix the device to the lamp socket.
3. The apparatus according to claim 1, characterized in that, The mounting part is a ceiling mount, an embedded downlight mounting structure, or a hanging structure, and the device is powered by a power cord or an external power adapter.
4. The apparatus according to claim 1, characterized in that, The optical axis of the lens coincides with the central axis of the device, and the direction of the optical axis of the lens is not adjustable relative to the housing.
5. The apparatus according to claim 1, characterized in that, It also includes a heat sink and a lens housing. The heat sink is located on the side of the housing opposite to the mounting part and is connected to the housing. The lamp plate is attached to the side of the heat sink opposite to the mounting part. The lamp plate has a central through hole in its center. The lens housing is located on the side of the lamp plate opposite to the mounting part and is opposite to the central through hole. The lens housing is fixed relative to the lamp plate. The camera module is fixed inside the lens housing. The light-shielding structure is a light-shielding section of the lens housing that protrudes from the light-emitting surface.
6. The apparatus according to claim 5, characterized in that, It also includes a lampshade, which is disposed on the side of the light-emitting element away from the mounting part. The bottom wall of the lampshade is provided with a through hole, and the end of the light-shielding section extends into the through hole, so that the imaging light path of the lens does not pass through the light-transmitting wall of the lampshade.
7. The apparatus according to claim 5, characterized in that, The control unit and the wireless communication module are mounted on a functional motherboard. The functional motherboard is located on the side of the heat sink near the mounting part and has a gap between it and the heat sink. A heat insulation layer is provided in the gap. The outer peripheral wall of the housing has multiple lateral openings on the side of the heat insulation layer near the mounting part. The lateral openings communicate with the space where the functional motherboard is located. The end of the housing near the mounting part has multiple heat dissipation grooves. An airflow channel is formed between the lateral openings and the heat dissipation grooves through the inner cavity of the heat insulation layer near the mounting part. The heat insulation layer separates the airflow channel from the space where the heat sink and the lamp panel are located.
8. The apparatus according to claim 1, characterized in that, The control unit is further configured to: when no person is detected in the detection of the entire downsampled image, divide the downsampled image into multiple overlapping sub-regions, detect each sub-region separately, and merge the detection results of each sub-region with the detection results of the entire image to remove duplicates.
9. A top-down imaging illumination device, characterized in that, It includes a mounting part, a housing, a lamp panel, multiple light-emitting elements, a lens housing, a camera module, and a lampshade; the mounting part is located at one end of the housing and is used to fix the device to the mounting surface above the desktop and to draw power from it; The lamp panel is located on the side of the housing opposite to the mounting portion. Multiple light-emitting elements are located on the side of the lamp panel opposite to the mounting portion and distributed around the center of the lamp panel. A central through-hole is provided at the center of the lamp panel. The lens housing is located on the side of the lamp panel opposite to the mounting portion and is positioned opposite the central through-hole. The lens housing is fixed relative to the lamp panel. The camera module is fixed inside the lens housing. The lens of the camera module is located within the area surrounded by the multiple light-emitting elements, and the optical axis of the lens faces away from the mounting portion. The lens housing is an opaque component and has a light-shielding section protruding from the light-emitting surface of the light-emitting element along the optical axis. The light-shielding section surrounds the lens, interrupting the straight line between the light-emitting surface of any light-emitting element and the light-incident surface of the lens. A lampshade is located on the side of the light-emitting element opposite to the mounting portion. The bottom wall of the lampshade has a through-hole, and the end of the light-shielding section extends into the through-hole, preventing the imaging light path of the lens from passing through the light-transmitting wall of the lampshade.
10. The apparatus according to claim 9, characterized in that, The optical axis of the lens coincides with the central axis of the device and is not adjustable relative to the housing; the mounting part is a lamp holder, a ceiling mount, an embedded downlight mounting structure, or a hanging structure.
11. A method for controlling the acquisition of dining images, executed by an acquisition device installed above a tabletop, the acquisition device comprising a camera module, a control unit, and a wireless communication module, characterized in that, The method includes: continuously acquiring downsampled images output by the camera module and locally detecting whether there is someone in front of the table on the acquisition device, wherein the continuous acquisition and local detection are not affected by the state of the image upload gate; generating a start test event when someone is detected, and generating an end test event when the number of consecutive determinations of no one reaches a preset threshold; reporting the test signal corresponding to the test event to the server via the wireless communication module; opening the image upload gate only when the server confirms the start test signal, and closing the image upload gate only when the server confirms the end test signal, and the unconfirmed test signal does not change the state of the image upload gate; and uploading full-frame images for meal analysis one by one to the storage path specified by the server for this meal only when the image upload gate is open, and not uploading full-frame images for meal analysis when the image upload gate is closed.
12. The method according to claim 11, characterized in that, The newly generated start and end trial events are strictly alternated; after the image upload gate is closed, a process must be performed to determine that no one is present for a certain number of consecutive times before a start trial event can be generated again; the encoding of the whole frame image used for meal analysis is only performed during the period when the image upload gate is open.
13. The method according to claim 11, characterized in that, The probe signal may include at most one frame of the image from the camera module at that time for the server to make a decision; the downsampled image used for local detection is not stored or uploaded.
14. The method according to claim 11, characterized in that, When a probe signal transmission fails, the type of the probe event is recorded, and the probe signal is resent when a corresponding detection result is obtained again. The transmission failure includes the request not being delivered to the server, the request being delivered but no response being received within a preset time, and the received response being unparseable. If the detection result is reversed before resentment, the probe event to be resent is cleared and not resent. When the server explicitly refuses to start a probe signal, no retry is made until the number of consecutive times that no one is detected reaches the preset threshold.
15. The method according to claim 11, characterized in that, After receiving confirmation of the start probe signal from the server, if opening the image upload gate fails, the meal number returned by the server is saved, and the image upload gate is retried using the meal number in the next probe event, without reporting a new start probe signal.
16. The method according to claim 11, characterized in that, The image-by-image upload includes: assigning a monotonically increasing and non-repeating sequence number to each image, which is persisted before transmission; the sequence numbers of images that are not successfully uploaded are no longer used and are not re-uploaded; and immediately clearing the local temporary files on the acquisition device after the upload and verification are completed. The local detection also includes: when no person is detected in the detection of the entire downsampled image, dividing the downsampled image into multiple overlapping sub-regions for separate detection and merging to remove duplicates.
17. A method for controlling the acquisition of dining images, executed by a server, characterized in that, include: Receive start or end probe signals from the acquisition device installed above the desktop, the probe signals being generated by the acquisition device based on local detection of whether there is someone in front of the table; The received probe signals are adjudicated, and the adjudication includes at least the following: for a start probe signal, if the acquisition device already has an ongoing meal, the meal number is returned without creating a new meal; if the acquisition device already has a meal record to be confirmed, the record is reused; for an end probe signal, if the meal it points to does not exist, it is rejected; if the meal it points to has ended, only confirmation is returned without repeating the end; only one meal record to be confirmed is allowed to exist on the same acquisition device at any given time; for probe signals that are deemed valid, confirmation is returned to the acquisition device, enabling the acquisition device to open or close the image upload gate accordingly; when a start probe signal is confirmed, a dedicated storage path is issued for this meal and provided to the acquisition device; images uploaded by the acquisition device to the storage path are received; before starting the recognition analysis of the images for this meal, the image list under the storage path is frozen, so that one analysis corresponds to one specific input.
18. The method according to claim 17, characterized in that, The decision also includes semantic decision: the image attached to the probe signal is given to the recognition unit to determine whether the table is in a dining state, and the judgment result is waited for within a preset waiting time; if no judgment result is obtained within the preset waiting time or the recognition unit is unavailable, the decision is confirmed.
19. The method according to claim 17, characterized in that, The storage path is an independent directory in the object storage, organized by meal. The images are directly transmitted from the acquisition device to the object storage. The independent directory is only appended to, not modified or deleted. The freezing includes: generating a snapshot of the independent directory containing object keys, content identifiers and size. The entire identification and analysis process is based on the snapshot. If the snapshot changes during the analysis, the analysis is deemed to have failed.
20. The method according to claim 17, characterized in that, During the continuous uploading of images by the acquisition device, identification and analysis are submitted in batches. The input for each batch is an explicitly listed list of object keys. The server maintains a cursor pointing to the last successfully analyzed object key, and the next batch starts after the cursor. Only one batch is allowed to be en route at any given time; When a batch fails, the cursor does not advance, and the objects of that batch are merged into the next batch; each batch returns the complete state of the entire meal up to the present, and the server completely overwrites the state of the previous batch with it; the input for the whole meal analysis submitted at the end of the meal includes the sequence number of the last successful batch and a list of object keys that have not been overwritten by any batch after the cursor.
21. The method according to claim 19, characterized in that, The identity feature map and process map generated by the identification analysis are uploaded to a path separate from the independent directory; when the identification analysis fails, the current meal is marked and the reason for the failure is recorded. The images in the independent directory remain unchanged and are not automatically retried. The analysis is then performed again by a new analysis explicitly initiated and an independent record is generated.
22. A dining image acquisition system, characterized in that, The system includes a data acquisition device and a server mounted on a desktop; the data acquisition device includes a camera module, a control unit, and a wireless communication module, the control unit being configured to perform the method of any one of claims 11 to 16; and the server being configured to perform the method of any one of claims 17 to 21.
23. The system according to claim 22, characterized in that, The acquisition device is the overhead imaging illumination device according to any one of claims 1 to 10.
Citation Information
Patent Citations
Device having built-in digital data device for LED bulb
US8562158B2
Internet protocol security camera connected light bulb / system
US9172917B1
Bulb with sensing function and camera
US9228731B2
Camera light
US9497421B2
Inbuilt camera lens with lamps
WO2009129664A1
