Pedestrian detection terminal, method, system, and electronic device

Abstract

The application discloses a pedestrian detection terminal, method, system and electronic equipment, and relates to the technical field of pedestrian detection. The pedestrian detection terminal comprises a mounting beam body, a mounting support, a multi-angle visual collection component and a first adjusting mechanism. The mounting support is arranged at the middle part of the mounting beam body and is used for mounting the pedestrian detection terminal on a mounting carrier corresponding to a to-be-detected channel. The multi-angle visual collection component comprises at least two visual collectors, one of which is used for collecting pedestrian images in the entry direction of the to-be-detected channel, and the other is used for collecting pedestrian images in the exit direction of the to-be-detected channel. The first adjusting mechanism is connected with the visual collectors and is used for adjusting the positions and / or angles of the visual collectors. Through the above structure, pedestrian images can be obtained from the entry direction and the exit direction of the channel, and the positions or angles of the visual collectors can be adjusted according to the mounting environment, so that the channel image collection integrity and the equipment mounting adaptability are improved.

Description

Technical Field

[0001] This application relates to the field of pedestrian detection technology, and in particular to a pedestrian detection terminal, method, system, and electronic device. Background Technology

[0002] The existing anti-tailgating devices in the self-service pedestrian inspection channels at ports are prone to misjudgment or omission in scenarios such as carrying children, dragging luggage, walking closely together, or multiple people hugging, due to their single collection angle and insufficient ability to distinguish complex postures and personal belongings. In addition, the devices do not recover in time after image acquisition or communication failures, resulting in poor stability of continuous operation of the turnstiles. Summary of the Invention

[0003] The main purpose of this application is to provide a pedestrian detection terminal, method, system and electronic device, which aims to solve the technical problems of existing pedestrian detection equipment having a single acquisition perspective, difficulty in adjusting the position or angle after installation, resulting in incomplete channel image coverage and poor on-site adaptability.

[0004] To achieve the above objectives, this application proposes a pedestrian detection terminal, comprising: Install the beam body; The mounting bracket is located in the middle of the mounting beam body and is used to install the pedestrian detection terminal on the mounting carrier corresponding to the channel to be detected. A multi-angle visual acquisition component includes at least two visual acquisition units, one for acquiring pedestrian images in the direction of entry of the channel to be detected, and the other for acquiring pedestrian images in the direction of exit of the channel to be detected. A first adjustment mechanism, connected to the vision acquisition device, is used to adjust the position and / or angle of the vision acquisition device.

[0005] In one embodiment, the first adjustment mechanism includes a sliding guide rail, and an end mounting shell is provided at the end of the sliding guide rail.

[0006] In one embodiment, the vision acquisition device is disposed in the end mounting housing.

[0007] In one embodiment, a network interface is also included, disposed on the surface of the mounting beam body, for transmitting the pedestrian image and for receiving control signals.

[0008] In one embodiment, the mounting bracket includes a second adjustment mechanism for adjusting the height of the mounting beam body.

[0009] Furthermore, to achieve the above objectives, this application also proposes a pedestrian detection method applied to the aforementioned pedestrian detection terminal, the method comprising: The first pedestrian image and the second pedestrian image extracted from the synchronous frame images obtained from the first visual collector and the second visual collector are fused and mapped to form a fused detection image; After performing human feature detection to obtain human body contours and head count, and item detection to identify strollers and / or large luggage, the fused detection image is fused according to a preset weighting rule to form a behavior detection result, and the corresponding gate control signal is output based on the behavior detection result.

[0010] In one embodiment, the synchronized frame image is obtained by timestamp alignment or frame number matching of the video streams acquired by the first visual acquisition device and the second visual acquisition device.

[0011] In one embodiment, the step of fusing and mapping the first pedestrian image and the second pedestrian image extracted from the synchronized frame images acquired by the first visual collector and the second visual collector to form a fused detection image includes: Human detection is performed on the synchronization frame images corresponding to the first and second visual acquisition devices, respectively, to obtain a first pedestrian image and a second pedestrian image; and Based on the first pedestrian image and the second pedestrian image, the corresponding image regions are cropped to form the first pedestrian image and the second pedestrian image respectively.

[0012] In one embodiment, the human feature detection employs a two-level inference method, including: Human contour detection is performed on the fused detection image to obtain human candidate regions; and Head detection is performed based on the candidate human body regions to obtain head detection results.

[0013] In one embodiment, the fusion determination according to a preset weighting rule includes: The number of valid personnel in the channel is determined based on the head detection results, and the behavior detection results are generated based on the number of valid personnel. The head detection results include the number of heads and head location information.

[0014] In one embodiment, the fusion mapping includes: The target category, target bounding box location, and / or detection confidence in the first pedestrian image and the second pedestrian image are matched to determine the valid target.

[0015] In one embodiment, the step of performing fusion judgment according to preset weight rules to form a behavior detection result includes: The results are weighted and fused based on the confidence levels of the human contour detection results, head detection results, object detection results, and the corresponding detection results of the first and second visual acquisition devices.

[0016] In one embodiment, the method further includes: Track the trajectory of targets within the turnstile channel to obtain the target's passage direction, passage sequence, and / or target spacing.

[0017] In one embodiment, the step of forming a behavior detection result and outputting a corresponding gate control signal based on the behavior detection result includes: Based on the number of valid personnel and the trajectory tracking results, the behavior detection result is determined, and the corresponding gate control signal is output according to the behavior detection result and the preset access mapping table. The behavior detection results include at least one of the following behaviors: tailing behavior, passing through with a child, passing through with a stroller, passing through with large luggage, and unauthorized entry. The preset access mapping table includes a first type of behavior that allows the gate to pass and a second type of behavior that prevents the gate from passing.

[0018] In one embodiment, the method further includes: When a gate control signal is output that causes the gate to block passage, a warning signal is output. In response to a reset signal, the warning state is lifted and a gate control signal is output to restore passage to the gate.

[0019] In addition, to achieve the above objectives, this application also proposes an electronic device, characterized in that the device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method.

[0020] Furthermore, to achieve the above objectives, this application also proposes a pedestrian detection system, characterized in that the system comprises: Pedestrian detection terminals and / or electronic devices; The turnstile is communicatively connected to the electronic equipment and / or the pedestrian detection terminal.

[0021] One or more technical solutions proposed in this application have at least the following technical effects: This application forms an overall installation structure suitable for the channel to be inspected by installing the beam body and the mounting bracket set in the middle of the beam body, so that the pedestrian detection terminal can be stably installed on the mounting carrier corresponding to the channel and ensure that there is a relatively stable installation position relationship between multiple vision acquisition devices.

[0022] Meanwhile, this application sets up at least two vision acquisition devices through a multi-angle vision acquisition component, with one acquiring pedestrian images in the direction of entry of the channel to be detected and the other acquiring pedestrian images in the direction of exit of the channel to be detected. This enables the terminal to acquire pedestrian images from different directions, reducing the problem of incomplete acquisition caused by occlusion, insufficient field of view or missing target features under a single viewpoint.

[0023] Furthermore, this application adjusts the position and / or angle of the visual acquisition device through the first adjustment mechanism, enabling the visual acquisition device to be adaptively adjusted according to different channel widths, installation heights, and site layouts, thereby improving the device's adaptability to different installation environments and providing a more stable and complete image foundation for subsequent pedestrian detection, behavior recognition, or channel control. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A first-view structural diagram of the pedestrian detection terminal provided in this application; Figure 2 A schematic diagram of the second-view structure of the pedestrian detection terminal provided in this application; Figure 3 This is a schematic diagram of the camera angle of the pedestrian detection terminal in this application; Figure 4 A flowchart illustrating the pedestrian detection method provided in this application; Figure 5 This is a schematic diagram of the functional architecture of the pedestrian detection system provided in this application.

[0027] Label: Explanation of icon numbers: 1. Turnstile; 2. Electronic equipment; 3. Pedestrian detection terminal; 10. Install beam body; 20. Install bracket; 30. First adjustment mechanism; 40. Network interface; 201. Second regulating mechanism; 301. End mounting housing; 302. Vision acquisition unit; 303. Sliding guide rail.

[0028] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0029] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0030] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0031] Most existing pedestrian detection equipment is built around a single camera or a fixed-viewpoint visual terminal. Such devices typically consist of a fixed image acquisition unit placed near the passageway, capturing and identifying people within the passageway from a single direction. In simple passageway environments with relatively regular traffic flow, this approach can meet basic detection needs. However, once entering turnstiles, border inspection channels, or other scenarios requiring more precise judgment of pedestrian movement, the limitations of this single-viewpoint, fixed-installation solution become apparent. Especially when people pass through the passageway from different directions, multiple people obstruct each other's view, people overlap with their belongings, or there are significant differences in height and width at the installation site, existing equipment often struggles to reliably acquire complete and effective pedestrian images, leading to a decrease in the accuracy of subsequent detection.

[0032] More specifically, a prominent problem with existing technologies is the lack of targeted design for image acquisition structures in passageway scenarios. Many devices simply fix the camera in a certain position and let the backend algorithm adapt to the image from that perspective, but the quality of the front-end acquisition itself is not guaranteed at the hardware level. Since people in the passageway do not always pass through in a standard posture or along a standard route, they may enter from one side and leave from the other; sometimes, their bodies may be partially obscured, people may walk close together, people may overlap with luggage, or even strollers, large bags, and other objects may occupy the main part of the frame. In such cases, if the image comes from only one direction, the device often only sees a part of the target's outline, or can only obtain local features, lacking image information that corroborates from the other direction. As a result, although the system "captures an image," it does not "capture a sufficiently usable image," making subsequent recognition prone to misjudgment or missed judgment.

[0033] Another significant drawback of existing technologies lies in their insufficient adaptability to the installation structure and the on-site environment. Real-world inspection channels are not standardized and uniform. Different projects may vary in channel width, gate height, installation location, overhead space conditions, and obstructions around the channel. Many existing devices use integrated fixed brackets or allow only very limited posture adjustments. Even slight deviations in installation position can cause the field of view to deviate from the center of the channel, resulting in the device only capturing a portion of the channel, or capturing the channel but not the critical target area within the optimal field of view. This is especially problematic for scenarios requiring image acquisition from both the inbound and outbound directions. Existing single-bracket, unidirectional fixed installation solutions often struggle to meet the acquisition needs of both directions. In other words, existing technologies are not entirely uninstallable, but rather difficult to fine-tune based on on-site conditions after installation. This makes the actual effectiveness of the equipment highly dependent on installation experience, resulting in poor stability and versatility.

[0034] Furthermore, existing technologies generally suffer from a flawed approach: overemphasizing backend recognition algorithms while neglecting the decisive impact of the frontend image acquisition architecture on recognition performance. In many solutions, problems are often attributed to "insufficient recognition model strength" or "the detection algorithm needs optimization." However, many misjudgments are not due to the algorithm's inability to distinguish, but rather to inherent deficiencies in the acquired image itself. For example, a unidirectional acquisition angle can distort or obscure the target outline; an inappropriate installation height may result in missing sections in the lower or upper parts of the channel; and a single shooting direction can make it difficult to clearly see the relative relationships between people and objects. In such cases, even with further stacking of recognition algorithms, the overall performance improvement will be significantly limited. Therefore, from an engineering perspective, the problem with existing technologies lies not only in the recognition methods themselves, but also in the fact that the frontend hardware acquisition structure fails to provide a sufficient, stable, and adjustable image foundation for accurate detection.

[0035] To address the aforementioned issues, this application's improvement approach does not simply involve adding algorithmic steps. Instead, it starts with the hardware architecture of the device, first resolving the fundamental problem of "how to stably acquire multi-angle pedestrian images suitable for passageway scenarios." This application establishes an integrated installation structure that corresponds to the arrangement of the passageway to be detected by setting up a mounting beam and mounting brackets. The mounting beam, serving as the load-bearing foundation of the entire device, is no longer a simple, scattered support but a main structure capable of spanning or corresponding to the entire passageway area. The mounting brackets, located in the middle of the mounting beam, are used to install the entire pedestrian detection terminal onto the mounting carrier corresponding to the passageway to be detected. In this way, the overall stability of the terminal after installation can be ensured from a mechanical structural perspective, maintaining a stable relative positional relationship between multiple acquisition units, laying the foundation for subsequent multi-angle acquisition.

[0036] Building upon this foundation, this application further incorporates a multi-angle visual acquisition component, ensuring the terminal has at least two visual acquisition units. These units are not arbitrarily arranged, but specifically designed so that one unit is used to acquire pedestrian images from the direction of entry into the detection channel, and the other from the direction of exit. The essence of this design is to establish a hardware acquisition architecture oriented towards bidirectional passage scenarios. Compared to existing unidirectional acquisition methods, this application no longer relies on a single fixed direction to "see as clearly as possible," but instead assigns two visual acquisition units with different orientations to handle image acquisition tasks from different directions. Thus, when there is occlusion, partial loss, or incomplete target contours in one direction, the image acquired from the other direction can serve as a supplement and verification. Through this multi-angle acquisition method, the device can obtain more complete target appearance information and channel occupancy information, improving image effectiveness at the hardware level and providing a more reliable source of raw data for subsequent detection and behavior determination.

[0037] This application further enhances the overall adaptability of the device to the field through a first adjustment mechanism. The first adjustment mechanism is connected to the vision acquisition unit and is used to adjust the position and / or angle of the vision acquisition unit. This improvement is not simply adding a movable part, but a structural solution to the problem of existing equipment being difficult to adapt to the field after installation. With this adjustment mechanism, the vision acquisition unit can still be positioned or its angle optimized according to the actual conditions of the channel to be inspected after installation. For example, when the channel is wide, the orientation of the acquisition unit can be adjusted to expand lateral coverage; when the installation height is high or low, the angle can be adjusted to refocus the line of sight on the key detection area; when the installation carrier position does not perfectly correspond to the center of the channel, the adjustment mechanism can also compensate for installation deviations. In other words, this application no longer relies on "accurate installation on the first attempt" for the device's acquisition effect, but rather transforms the device into an adaptable and optimizable end product through an adjustable structure. This improvement significantly enhances the applicability of the device in different channel environments and also improves engineering deployment efficiency.

[0038] Furthermore, the core idea of ​​this application lies in combining "multi-angle acquisition" with "adjustable installation," rather than introducing one feature in isolation. Having only multiple acquisition units without adjustment capabilities may lead to unreasonable coverage relationships between multiple viewpoints, still resulting in insufficient acquisition in local areas of the channel; conversely, having only adjustable position and angle without a multi-view layout cannot fundamentally overcome the occlusion and information loss problems caused by single-direction acquisition. This application combines the two to form a hardware architecture more suitable for channel detection scenarios: on the one hand, pedestrian images are acquired from different directions through at least two visual acquisition units, establishing multi-directional image sources; on the other hand, the position and angle of the visual acquisition units are adjusted through a first adjustment mechanism, enabling the fields of view of different acquisition units to form a complementary relationship according to the actual channel environment. In this way, the terminal not only possesses multi-view capabilities, but these capabilities are configurable and optimizable, rather than static and rigid. Therefore, this application can obtain a more complete image coverage in complex channel environments, improving the visibility and distinguishability of pedestrian targets, thus laying a more reliable hardware foundation for subsequent detection and control.

[0039] Furthermore, this application includes a network interface in the terminal for transmitting pedestrian images and receiving control signals. This demonstrates that this application is not merely a mechanical installation, but a hardware terminal designed for a practical detection system. While some existing solutions possess image acquisition capabilities, they are often simply independent camera devices, lacking effective collaboration with backend detection systems, gate control systems, or upper-level management systems. This application, through its network interface, connects the front-end's multi-angle image acquisition capabilities with subsequent data processing, behavior analysis, and control execution capabilities. This allows the terminal to not only acquire images but also serve as a front-end device in the entire detection and control chain, connecting to the actual system. The significance of this design lies in the fact that this application does not protect an abstract visual acquisition concept, but rather a pedestrian detection terminal with a tangible product form, capable of practical deployment and collaborative operation with external systems.

[0040] In summary, the main problems with existing technologies lie in their limited acquisition perspective, poor installation adaptability, and the inability of the front-end hardware structure to provide sufficient image data for complex passageway scenarios, resulting in insufficient detection reliability in complex traffic environments. The improvement proposed in this application involves constructing a multi-angle, adjustable pedestrian detection terminal suitable for the passageway to be detected, using a mounting beam, mounting bracket, multi-angle visual acquisition components, and a first adjustment mechanism. This allows the device to simultaneously acquire images from both the inbound and outbound directions of the passageway, and optimizes the position and angle of the acquisition unit according to the on-site installation conditions. This improves the integrity of the passageway image coverage and adaptability to on-site deployment, providing more stable and effective hardware support for subsequent pedestrian detection, behavior analysis, and passageway control.

[0041] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or terminal capable of performing the above functions. The following description uses an electronic device as an example to illustrate this embodiment and the subsequent embodiments.

[0042] Based on this, embodiments of this application provide a pedestrian detection terminal, such as... Figure 1 , Figure 2 as well as Figure 3 As shown, the pedestrian detection terminal 3 is suitable for installation above or at the corresponding location of the gate 1 channel, port inspection channel, or other channel areas that require detection of pedestrian passage status, in order to acquire image information of pedestrians in the channel.

[0043] The pedestrian detection terminal 3 includes a mounting beam body 10, a mounting bracket 20, a multi-angle visual acquisition component, and a first adjustment mechanism 30.

[0044] The mounting beam body 10 is a structural component that extends laterally and is used to span or be arranged above the channel to be tested. It can be a metal beam structure, a profile structure, or other load-bearing structure with sufficient strength and rigidity to support the various functional components and ensure the overall structural stability.

[0045] The mounting bracket 20 is located in the middle of the mounting beam body 10 and is used to fix the entire pedestrian detection terminal 3 on the mounting carrier corresponding to the channel to be detected, such as a ceiling structure, a beam structure, or other fixed support structure. By setting the mounting bracket 20 in the middle of the mounting beam body 10, the mounting beam body 10 can form a stable suspended or supported state above the channel, thereby ensuring the relative position stability of the visual acquisition units 302 on both sides.

[0046] It should be noted that the "installation carrier" can be understood as a fixed structure used to support the pedestrian detection terminal 3. It can be a roof slab, beam, support frame or other installation base in a building structure, and is not limited to a specific form.

[0047] The multi-angle visual acquisition component is mounted on the mounting beam body 10 and includes at least two visual acquisition units 302. One visual acquisition unit 302 is used to acquire pedestrian images in the direction of entry into the channel to be detected, and the other visual acquisition unit 302 is used to acquire pedestrian images in the direction of exit from the channel to be detected. By arranging the different visual acquisition units 302 facing the entry and exit directions of the channel respectively, the passage status of pedestrians in the channel can be acquired from different directions, thereby improving the integrity of image information.

[0048] It should be noted that the "visual acquisition device 302" can be a camera, an image acquisition module, or other visual sensing device capable of acquiring image data. It may include an image sensor, a lens assembly, and related circuits, and is not limited to a specific model or structural form.

[0049] It should be noted that the "entry direction" and "exit direction" are relative directions defined with reference to the passage direction of the channel to be detected. The entry direction is the direction on which pedestrians enter the channel, and the exit direction is the direction on which pedestrians leave the channel. The division of these directions can be defined according to the actual channel settings.

[0050] In this embodiment, two vision acquisition units 302 can be installed at both ends of the mounting beam body 10, arranged at intervals along the length of the mounting beam body 10, thereby forming a laterally distributed acquisition structure. Of course, in other embodiments, the vision acquisition units 302 can also be set at different positions on the mounting beam body 10, such as near the center or at an off-center position, as long as image acquisition in different directions can be achieved.

[0051] The first adjustment mechanism 30 is connected to the vision acquisition device 302 and is used to adjust the position and / or angle of the vision acquisition device 302. Through the first adjustment mechanism 30, the installation position or shooting angle of the vision acquisition device 302 can be adjusted after installation to adapt to differences in channel width, height and installation environment.

[0052] It should be noted that the "position adjustment" may include the translation adjustment of the vision acquisition device 302 on the mounting beam body 10, the height adjustment along the support structure, or the spatial position change within a certain range; the "angle adjustment" may include the pitch angle adjustment, horizontal rotation angle adjustment, or other orientation adjustment of the vision acquisition device 302.

[0053] In one specific implementation, the first adjustment mechanism 30 may include an adjustable connector, a sliding structure, a rotating structure, or a connection structure with positioning holes. For example, by setting multiple positioning holes on the support arm or mounting piece, the vision acquisition device 302 can be installed in different positions, thereby achieving position adjustment; or by setting a hinge structure or a rotating connector, the shooting angle of the vision acquisition device 302 can be adjusted.

[0054] It should be noted that the "multi-angle visual acquisition component" is not limited to including only two visual acquisition units 302. In other embodiments, it may further include a third visual acquisition unit 302, such as a visual acquisition unit 302 located above the middle of the mounting beam body 10 or extended to a higher position through a support arm, for acquiring pedestrian images from a top-down or oblique angle, thereby further enriching the acquisition perspective.

[0055] Based on the above structure, the pedestrian detection terminal 3 of this embodiment forms an integral load-bearing structure through the mounting beam body 10, and is fixedly connected to the mounting carrier through the mounting bracket 20. It acquires images by facing the in-direction and out-direction of the channel with at least two vision acquisition devices 302 respectively, and adjusts the position and angle of the vision acquisition devices 302 through the first adjustment mechanism 30, so as to obtain a more reasonable multi-angle image acquisition effect in different installation environments.

[0056] Through the above structural design, the pedestrian detection terminal 3 can acquire images of pedestrians from multiple directions and angles in the detection channel, effectively reducing the occlusion problem caused by a single viewpoint, improving the coverage integrity of pedestrian images in the channel, and providing a reliable image foundation for subsequent pedestrian detection, behavior recognition or channel control.

[0057] Optionally, in one embodiment, the first adjustment mechanism 30 adopts a sliding guide rail 303 structure. The sliding guide rail 303 is disposed on the mounting beam body 10 and arranged along the extension direction of the mounting beam body 10. Through the sliding guide rail 303, the mounting component connected to the vision acquisition device 302 can be moved in a predetermined direction, thereby realizing the position adjustment of the vision acquisition device 302 on the mounting beam body 10 to adapt to the detection channel of different widths or different layouts.

[0058] It should be noted that the "sliding guide rail 303" refers to a guide structure used to guide the component to slide in a certain direction. It may include the guide rail body and the sliding parts that cooperate with it, such as sliders, grooves, etc. The sliding guide rail 303 can be a linear guide rail or other structural forms with guiding functions, as long as it can realize the sliding adjustment of the vision acquisition device 302 relative to the mounting beam body 10.

[0059] In this embodiment, an end mounting shell 301 is provided at the end of the sliding guide rail 303. The end mounting shell 301 can be fixedly disposed at the end of the sliding guide rail 303, or it can be connected to a sliding member of the sliding guide rail 303, allowing the end mounting shell 301 to move along the guide rail with the sliding member. By providing an end mounting shell 301 at the end of the sliding guide rail 303, a stable mounting carrier can be provided structurally for the vision acquisition device 302.

[0060] It should be noted that the "end mounting shell 301" can be understood as a shell structure for mounting and protecting the vision acquisition device 302. It can have a closed or semi-closed structural form to accommodate the vision acquisition device 302 and its connecting components, and to provide it with mechanical protection and environmental isolation. The end mounting shell 301 can be made of metal or engineering plastic, and its specific structural form is not limited.

[0061] Furthermore, the visual acquisition device 302 is disposed within the end mounting housing 301. That is, the visual acquisition device 302 is fixed inside the end mounting housing 301 by a mounting structure and is supported as a whole by the end mounting housing 301. The end mounting housing 301 may have an opening or transparent window at a position corresponding to the visual acquisition direction, enabling the visual acquisition device 302 to acquire images externally.

[0062] It should be noted that the phrase "set in the end mounting shell 301" can mean that the vision acquisition device 302 is completely located inside the end mounting shell 301, or that the vision acquisition device 302 is partially extended or embedded in the end mounting shell 301. As long as the vision acquisition device 302 is installed and supported by the end mounting shell 301, this limitation can be satisfied.

[0063] In one specific implementation, the vision acquisition device 302 can be installed inside the end mounting housing 301 via a threaded connection, snap-fit ​​connection, or other fixing method, and forms an indirect connection with the sliding guide rail 303 through the end mounting housing 301. When it is necessary to adjust the position of the vision acquisition device 302, the sliding member can be driven to move on the sliding guide rail 303, causing the end mounting housing 301 and the vision acquisition device 302 inside it to move along the guide rail direction, thereby changing the position of the vision acquisition device 302 on the mounting beam body 10.

[0064] Through the above structural design, the vision acquisition unit 302 can not only achieve position adjustment through the sliding guide rail 303, but also achieve stable installation and effective protection through the end mounting shell 301, thereby improving the adaptability of the equipment in different channel environments while ensuring structural reliability.

[0065] Optionally, in one embodiment, the network interface 40 is disposed on the surface of the mounting beam body 10, for example, it can be disposed on the side wall, bottom or end of the mounting beam body 10. By directly disposing of the network interface 40 on the surface of the mounting beam body 10, the network interface 40 can be easily connected for wiring without significantly increasing structural complexity, and it also facilitates later maintenance and repair.

[0066] It should be noted that the "network interface 40" can be an interface structure for implementing data communication. It can be an Ethernet interface, an industrial communication interface, or other forms of wired communication interface, used to establish communication connections with external devices. The network interface 40 can support the transmission of image data and the reception of control signals, and its specific communication protocol or interface form is not limited.

[0067] In this embodiment, the network interface 40 is electrically connected to the visual acquisition unit 302 and is used to transmit pedestrian image data acquired by the visual acquisition unit 302 to an external processing device, such as a host computer, server, or gate 1 control system. Simultaneously, the network interface 40 is also used to receive control signals from external devices, such as control commands for controlling the terminal's operating status or control signals for controlling the operation of the access control equipment.

[0068] In one specific implementation, the network interface 40 can be located on the outer surface of the mounting beam body 10 and connected to the internal circuit via wires. To ensure the stability and protection performance of the overall structure, the network interface 40 can be equipped with a protective structure, such as a protective shell or seal, to prevent external environmental factors such as dust and moisture from affecting the interface.

[0069] In addition, in some embodiments, the network interface 40 can also serve as a power supply function, for example, by providing operating power to the vision acquisition unit 302 and related circuits through the network interface 40, thereby reducing the need for additional power supply lines and improving the overall integration of the device.

[0070] Through the above structural configuration, the pedestrian detection terminal 3 can not only acquire pedestrian images, but also interact with external systems through the network interface 40 for data and control. This makes it easy to integrate the terminal into actual pedestrian detection systems or channel control systems, thereby improving the practicality and scalability of the equipment.

[0071] Optionally, embodiments of this application also provide a pedestrian detection method, referring to... Figure 4 This method is applied to the aforementioned pedestrian detection terminal 3. It identifies the pedestrian passage status within the detection channel and outputs corresponding control results. It includes steps S100 to S200.

[0072] Step S100: The first pedestrian image and the second pedestrian image extracted from the synchronous frame images obtained from the first visual collector and the second visual collector are fused and mapped to form a fused detection image; Step S200: After performing human feature detection to obtain human body contours and head count, and item detection to identify strollers and / or large luggage on the fused detection image, a fusion judgment is performed according to a preset weight rule to form a behavior detection result, and a corresponding gate control signal is output based on the behavior detection result.

[0073] Specifically, the first visual acquisition unit 302 and the second visual acquisition unit 302 first acquire images of the channel to be detected. The first visual acquisition unit 302 acquires pedestrian images in the direction of entry into the channel, and the second visual acquisition unit 302 acquires pedestrian images in the direction of exit from the channel. The video streams acquired by the two visual acquisition units 302 are then synchronized to form a synchronized frame image.

[0074] It should be noted that the "synchronized frame image" refers to the frame image corresponding to the images from different visual acquisition devices 302 in the time dimension. It can be obtained through timestamp alignment, frame number matching or other synchronization mechanisms, so as to ensure that the images from different perspectives can reflect the channel status at the same time or adjacent times.

[0075] After obtaining the synchronized frame image, a first pedestrian image is extracted from the synchronized frame image corresponding to the first visual collector 302, and a second pedestrian image is extracted from the synchronized frame image corresponding to the second visual collector 302. This extraction process can be achieved by detecting pedestrian targets in the image and cropping the detected target regions to obtain image data containing only pedestrian regions.

[0076] It should be noted that the "first pedestrian image" and "second pedestrian image" refer to the images containing pedestrian target areas collected by the first visual collector 302 and the second visual collector 302, respectively. They can be local areas in the complete image or pedestrian area images obtained after preprocessing.

[0077] Subsequently, the first and second pedestrian images are fused and mapped to form a fused detection image. The fusion mapping process may include spatial alignment, target correspondence, and feature fusion of pedestrian images from different viewpoints, so that information from different viewpoints can be expressed in the same image or a unified data structure.

[0078] It should be noted that the "fusion mapping" can be understood as a process of corresponding and integrating images or target information from different perspectives. It can be based on target location, target category or feature information to generate a fused detection image containing multi-view information.

[0079] After obtaining the fused detection image, human feature detection and object detection are performed on the fused detection image. Among them, human feature detection is used to obtain human contour and head number information, and object detection is used to identify whether there are strollers and / or large luggage in the channel.

[0080] It should be noted that the "human feature detection" refers to the process of identifying human-related features in an image, which may include human contour detection, key part detection or head detection, etc., to extract feature information representing the presence and quantity of human bodies.

[0081] It should be noted that the "item detection" refers to the process of identifying non-human targets in an image. It is used to identify items related to human passage, such as strollers and large luggage, thereby assisting in judging the pedestrian passage status.

[0082] After obtaining the human feature detection results and object detection results, the above results are fused and judged according to preset weight rules to form a behavior detection result. Specifically, different weights can be assigned to different detection results based on the importance of human contour detection results, number of heads, and object detection results, and comprehensive analysis can be performed to determine the behavior type in the current channel.

[0083] It should be noted that the "preset weighting rule" refers to the rule used to weight different detection results during the fusion judgment process. It can be set according to actual application needs. For example, different weights can be determined according to the confidence level of the detection results, target type or scene characteristics.

[0084] It should be noted that the "behavior detection result" refers to the result obtained by fusing and analyzing the human feature detection result and the object detection result. It can be used to characterize the current passage status, such as whether there is tailgating behavior, whether there is child-carrying behavior, whether there is stroller-pushing behavior or large luggage-carrying behavior, etc.

[0085] Finally, based on the behavior detection results, a corresponding gate control signal is output to control the passage status of gate 1. For example, when abnormal passage behavior is detected, a control signal prohibiting passage can be output; when normal passage behavior is detected, a control signal allowing passage can be output.

[0086] It should be noted that the "gate control signal" is a control command used to control the gate to perform the opening, closing or holding state. It can be sent to the gate control system through the communication interface to realize the control of the passage status.

[0087] Using the above method, the behavior of pedestrians in the channel can be more accurately identified based on the image information obtained from multi-angle visual acquisition, and corresponding gate control signals can be output accordingly, thereby improving the accuracy of pedestrian detection and the reliability of channel control.

[0088] Optionally, in this embodiment, after acquiring the video streams from the first visual acquisition device 302 and the second visual acquisition device 302, to ensure the temporal consistency of images from different viewpoints, the video streams are synchronized to obtain synchronized frame images. The synchronization process can be achieved by time alignment or frame number matching of the video frames output by the two visual acquisition devices 302.

[0089] It should be noted that the "timestamp alignment" refers to matching image frames from different visual acquisition devices 302 according to time by using the time information corresponding to each frame of the video stream, so that the selected image frames correspond to the same or adjacent time as much as possible in time.

[0090] It should be noted that the "frame number matching" refers to matching images according to the frame sequence number in the video stream, so that image frames from different visual acquisition devices 302 are consistent in the frame sequence, thereby achieving synchronous processing.

[0091] The above method can obtain synchronized frame images from different perspectives that correspond to each other in the time dimension, thus providing a foundation for subsequent image fusion and behavior detection.

[0092] After obtaining the synchronized frame images, human detection processing is performed on the synchronized frame images corresponding to the first visual acquisition unit 302 and the synchronized frame images corresponding to the second visual acquisition unit 302, respectively. The human detection processing is used to identify pedestrian targets from the images and output the detection results containing pedestrian regions.

[0093] It should be noted that "human detection" refers to the process of identifying human targets in an image through image processing or machine learning models, which can output the target's bounding box, location coordinates, or other information used to characterize the human body region.

[0094] After obtaining the human detection results, the original image can be cropped based on the detected human body regions to form a first pedestrian image and a second pedestrian image. Specifically, the corresponding region in the image can be cropped based on the boundary region corresponding to the human detection results to obtain an image region containing only the pedestrian target.

[0095] It should be noted that "image cropping" refers to extracting a corresponding local area from the original image according to a predetermined area range in order to remove irrelevant background information and thus obtain image data that is more focused on the target.

[0096] After obtaining the first pedestrian image and the second pedestrian image, they are fused and mapped. This fusion and mapping process may include corresponding matching of pedestrian images from different viewpoints and integration at the spatial or feature level, so that pedestrian information from different visual acquisition devices 302 can be analyzed in a unified representation.

[0097] Through the above processing flow, pedestrian images can be extracted from different perspectives and fused while ensuring time synchronization, thereby improving the integrity and consistency of pedestrian image information and providing a more reliable data foundation for subsequent human feature detection and behavior determination.

[0098] Optionally, in one embodiment, after obtaining the fused detection image, human feature detection is performed on the fused detection image. In this embodiment, human feature detection is performed using a two-level inference method.

[0099] It should be noted that the "two-stage reasoning method" refers to dividing the human feature detection process into at least two sequential processing stages. The subsequent stage is performed based on the output of the previous stage, thereby gradually refining the detection results and improving detection accuracy.

[0100] In the first stage, human contour detection is performed on the fused detection image to obtain human candidate regions. Human contour detection is used to identify regions in the fused detection image that may contain human bodies and output them as bounding boxes or regions.

[0101] It should be noted that the "human body contour detection" refers to the process of recognizing the overall shape of the human body in an image. Its purpose is to determine the approximate location range of the human body in the image, thereby filtering out candidate regions containing the human body.

[0102] It should be noted that the "human candidate region" refers to an image region that may contain human targets, obtained through human contour detection. This region is used as input for subsequent more refined detection.

[0103] In the second stage, head detection is performed based on the human candidate regions to obtain head detection results. Specifically, the image data corresponding to the human candidate regions can be used as input to detect head targets within them, thereby obtaining the number and location information of heads.

[0104] It should be noted that the "head detection" refers to the process of identifying the head region in a human image. It can output the number of head targets and their corresponding spatial location information to represent the actual number of people in the channel.

[0105] It should be noted that the "head detection result" includes the number of heads and head location information, where the number of heads is used to indicate the number of heads detected in the current image, and the head location information is used to indicate the positional distribution of each head in the image.

[0106] By using the above two-level reasoning method, human body candidate regions can be screened out first through human body contour detection, and then head detection can be performed within the candidate regions, thereby reducing the interference of irrelevant regions on the detection results and improving the accuracy of head detection.

[0107] After obtaining the head detection results, a fusion judgment is performed according to a preset weighting rule to form a behavior detection result. In this embodiment, the fusion judgment process includes determining the number of valid personnel in the channel based on the head detection results, and further generating a behavior detection result based on the number of valid personnel.

[0108] It should be noted that the "effective number of people" refers to the number of people identified as genuinely passing through the current passage. This number can be determined based on head detection results. For example, the number of detected heads can be used as the initial number of people, and then filtered or corrected in conjunction with head position information to obtain a number of people that better reflects the actual passage situation.

[0109] In one implementation, the spatial relationship between multiple heads can be determined based on head position information, such as the distance between heads, their distribution area, or whether there is overlap, thereby correcting the anomaly detection results to obtain a more accurate number of valid personnel.

[0110] After determining the number of valid personnel, behavior detection results can be generated based on the actual situation within the passage. For example, if the number of valid personnel does not match the preset passage rules, it can be determined as abnormal passage behavior; if the number of valid personnel meets expectations, it can be determined as normal passage behavior.

[0111] By using the above-mentioned processing method to determine the number of valid people and generate behavior detection results based on head detection results, the accuracy of identifying the number of people in the passage can be improved by utilizing the relatively stable human feature of the head, thereby providing a reliable basis for subsequent judgment of passage behavior.

[0112] Optionally, in one embodiment, after obtaining the first pedestrian image and the second pedestrian image, a fusion mapping process is performed on the two. In this embodiment, the fusion mapping process includes not only image-level integration but also corresponding matching of detection results from different viewpoints.

[0113] Specifically, target detection processing can be performed on the first pedestrian image and the second pedestrian image separately to obtain their respective target information, such as target category, target bounding box location, and detection confidence. Subsequently, the target information from the two viewpoints is matched to determine the same target corresponding to both viewpoints.

[0114] It should be noted that the "target category" refers to the category information obtained after classifying the target in the image through the detection model, such as human body, stroller or large luggage, which is used to distinguish different types of detection objects.

[0115] It should be noted that the "target bounding box position" refers to information used to represent the range of the target's position in the image. It can usually be represented by the coordinates of a rectangle and is used to determine the area where the target is located.

[0116] It should be noted that the "detection confidence" refers to the measure of the credibility of the detection model for the recognition result. It can be a numerical value used to represent the probability that the detection result belongs to a certain category.

[0117] During the matching process, information such as the consistency of target categories, the positional relationship of bounding boxes, and detection confidence in the first and second pedestrian images can be used to determine whether targets in different viewpoints correspond to the same actual object. For example, the correspondence can be determined by comparing the relative position, size, and category label of the targets in their respective images.

[0118] Through the matching process described above, targets that are identified and have high consistency in both perspectives can be selected, thus determining valid targets. Valid targets can be understood as targets that are confirmed to exist under multi-perspective information, and are used for subsequent behavior analysis.

[0119] After obtaining the human contour detection results, head detection results, and object detection results, they are further fused and judged according to preset weight rules to form behavior detection results.

[0120] In this embodiment, the fusion determination process can be based on a comprehensive analysis of multiple detection results, including human contour detection results, head detection results, object detection results, and the confidence levels of the corresponding detection results from the first visual acquisition device 302 and the second visual acquisition device 302. Specifically, different weights can be assigned to the detection results from the above different sources, and weighted fusion processing can be performed.

[0121] It should be noted that the "weighted fusion" refers to a processing method that combines multiple detection results according to preset weights to obtain a comprehensive judgment result. The weights of different detection results can be set according to their reliability, stability, or importance in the current scenario.

[0122] For example, when determining the number of people, head detection results can be given higher weight, while when judging whether there are special passage conditions, object detection results can be given higher weight. At the same time, for detection results from different vision acquisition devices 302, weights can be assigned according to their detection confidence, for example, giving higher weight to detection results with higher confidence.

[0123] By employing the weighted fusion process described above, information from different detection results can be comprehensively utilized, improving the accuracy and robustness of behavior detection. Ultimately, based on the fused result, the corresponding behavior detection result is obtained, such as determining whether there are abnormal passage behaviors or special passage situations within the current channel.

[0124] Through the above-mentioned fusion mapping and weighted fusion judgment process, results from different perspectives and different detection modules can be effectively integrated, thereby improving the overall detection effect.

[0125] Optionally, in one embodiment, after obtaining the human contour detection results, head detection results, and object detection results, the multi-source detection results are fused and determined. In this embodiment, the fusion determination process adopts a weighted fusion method based on preset weight rules.

[0126] Specifically, the confidence levels of the human contour detection results, head detection results, object detection results, and the corresponding detection results of the first visual acquisition device 302 and the second visual acquisition device 302 can be used as input parameters. Different weights are assigned to different detection results, and comprehensive calculations are performed to obtain the final behavior detection results.

[0127] It should be noted that the "human body contour detection result" refers to the result obtained by performing human body contour detection on the fused detection image, which is used to represent the position and distribution of the overall human body region in the image.

[0128] It should be noted that the "item detection result" refers to the result obtained after detecting non-human targets in the image. It is used to indicate whether there are targets such as strollers and large luggage in the image and their location distribution.

[0129] In practical implementation, corresponding weights can be set according to the importance of different detection results in the current scene. For example, in judging the number of people, head detection results can be given higher weight; in judging whether there is special passage behavior, object detection results can be given higher weight; at the same time, for detection results from different vision acquisition devices 302, the weights can be dynamically adjusted according to their detection confidence, so that detection results with higher confidence will occupy a larger proportion in the fusion process.

[0130] By using the weighted fusion method described above, the advantages of different detection results can be combined, the impact of errors in a single detection result on the overall judgment can be reduced, thereby improving the accuracy and stability of behavior detection results.

[0131] Based on this, in this embodiment, the method also includes a process of trajectory tracking of the target within the gate 1 channel.

[0132] Specifically, correlation analysis can be performed on targets in consecutive frames of images. By tracking the positional changes of the same target in different time frames, the target's trajectory can be obtained. Through trajectory tracking, information such as the target's direction of travel, travel order, and target spacing can be further obtained.

[0133] It should be noted that the "trajectory tracking" refers to the process of associating and recording the positional changes of a target in consecutive image frames. It can be achieved through target matching, feature tracking or other methods, and is used to describe the target's motion path in the time dimension.

[0134] It should be noted that the "direction of travel" refers to the direction of movement of the target within the channel, such as moving from the direction of entry to the direction of exit or moving in the opposite direction.

[0135] It should be noted that the "passage order" refers to the order in which multiple targets pass through the channel, and is used to characterize the time sequence of the targets' passage.

[0136] It should be noted that the "target spacing" refers to the relative distance between multiple targets in space, such as the interval between different targets at the same time.

[0137] By acquiring the aforementioned trajectory information, a more comprehensive analysis of pedestrian behavior can be conducted. For example, the presence of tailgating can be determined based on the order of passage and the distance between targets; whether targets are traveling in the wrong direction can be determined based on the direction of passage; and the presence of abnormal clustering or abnormal passage within the passage can be determined by combining the number of targets and their changes.

[0138] By combining the weighted fusion judgment results with the trajectory tracking results, the accuracy and reliability of the behavior detection results can be further improved, thus providing a more sufficient basis for the subsequent control of gate 1.

[0139] Optionally, in one embodiment, after obtaining the effective number of people and the trajectory tracking results, the passage behavior within the channel can be comprehensively analyzed based on the above information to determine the behavior detection result. This behavior detection result considers not only the number of people currently in the channel but also the target's movement trajectory information over time to improve the accuracy of the judgment.

[0140] It should be noted that the "preset access mapping table" refers to a set of rules used to associate different types of behavior detection results with the corresponding control strategies of gate 1. This mapping table can be pre-set according to actual application requirements. It can include multiple behavior categories and their corresponding control methods to guide the response actions of gate 1 under different access states.

[0141] In this embodiment, the behavior detection results can include various behavior types, such as tailing, passing through with children, passing through with strollers, passing through with large luggage, and unauthorized entry. For different behavior types, a judgment can be made by combining the number of valid personnel and trajectory tracking results.

[0142] For example, when judging tailgating behavior, the analysis can be combined with the number of valid personnel and the passage sequence information. When multiple targets are detected closely following each other through a passage within the same time period, and the distance between the targets is less than a preset threshold, it can be judged as tailgating behavior; when judging the behavior of carrying children, the relationship between head detection results and human contour detection results can be used to determine whether there are smaller auxiliary targets and adult targets passing through the passage at the same time; when judging the behavior of pushing strollers or carrying large luggage, the object detection results and trajectory information can be combined to determine whether the target is moving with specific objects; when judging unauthorized entry, the direction and sequence of passage can be used to identify whether there is reverse passage or unauthorized entry.

[0143] After determining the behavior detection results, the behavior types are classified according to a preset access mapping table, and the corresponding gate control signals are output. Specifically, the preset access mapping table can include two types of behavior: a first type of behavior and a second type of behavior. The first type of behavior is the behavior type that is allowed to pass, and the corresponding control strategy is to allow gate 1 to pass; the second type of behavior is the behavior type that is prohibited to pass, and the corresponding control strategy is to prevent gate 1 from passing.

[0144] For example, when the behavior detection result is normal single-person passage or passage with a child in accordance with regulations, it can be classified as Category I behavior, and a gate control signal allowing passage will be output; when the behavior detection result is tailgating or illegal entry, it can be classified as Category II behavior, and a gate control signal prohibiting passage will be output.

[0145] By using the above method, the behavior detection results obtained based on multi-source detection results and trajectory information are associated with the preset passage mapping table, so that different types of passage behaviors can correspond to different gate 1 control strategies, thereby realizing intelligent control of the passage status and improving the accuracy and security of passage management.

[0146] Optionally, in one embodiment, during the process of completing behavior detection and outputting the corresponding gate control signal, when the detected behavior belongs to the situation of prohibiting passage, in addition to outputting the control signal for prohibiting passage of gate 1, a warning signal is also output simultaneously to indicate that there is abnormal passage behavior on site.

[0147] It should be noted that the "warning signal" can be a signal used to indicate an abnormal state, which can be an audible and visual alarm signal, a display prompt signal, or other forms of alarm information. For example, it can be a sound prompt issued by a buzzer, a flashing signal issued by an indicator light, or an alarm prompt message output by a display terminal; the specific form is not limited.

[0148] In this embodiment, when the behavior detection result indicates the presence of tailgating, illegal intrusion, or other behaviors that do not comply with the passage rules, the system outputs a gate control signal that prohibits passage, causing the gate 1 to be closed or prohibited from allowing passage. At the same time, a warning signal is output to remind on-site personnel or the management system to take action.

[0149] After the abnormal state is handled, the current warning state can be cleared by inputting a reset signal, and the normal passage function of gate 1 can be restored.

[0150] It should be noted that the "reset signal" refers to the control signal used to trigger the system to recover from an abnormal state to a normal state. It can be triggered manually, such as by button input or management terminal operation, or it can be automatically sent by the host system according to preset conditions.

[0151] Upon receiving a reset signal, the system removes the current warning state, such as stopping alarm output or clearing abnormal markers, and simultaneously outputs a control signal to restore passage for gate 1, causing gate 1 to re-enter the passage-allowed state.

[0152] By employing the above-mentioned processing methods, when abnormal passage behavior is detected, not only can passage be stopped in a timely manner, but warning signals can also be issued, and the system state can be restored through a reset signal after the abnormality is resolved, thereby improving the security and controllability of channel management.

[0153] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the pedestrian detection method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0154] This application also provides an electronic device 2, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the pedestrian detection method in the first embodiment described above.

[0155] The electronic device 2 in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (such as vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers.

[0156] The electronic device 2 provided in this application, employing the pedestrian detection method in the above embodiments, can solve the technical problem of pedestrian detection. Compared with the prior art, the beneficial effects of the electronic device 2 provided in this application are the same as those of the pedestrian detection method provided in the above embodiments, and other technical features of the electronic device 2 are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0157] This application also provides a pedestrian detection system, referring to... Figure 5 This system is used to detect the pedestrian traffic status within the detection channel and control the passage status of the gate 1. The system includes a pedestrian detection terminal 3 and the gate 1, and may also include electronic equipment 2.

[0158] The pedestrian detection terminal 3 can adopt the structural form described in the previous embodiment, used to acquire images of pedestrians in the detection channel and output corresponding image data. The gate 1 is set at the detection channel to physically control pedestrian passage, such as allowing or prohibiting passage.

[0159] The turnstile 1 is communicatively connected to the pedestrian detection terminal 3 and / or electronic device 2 to receive control signals and execute corresponding actions. Specifically, the turnstile 1 can interact with the pedestrian detection terminal 3 or electronic device 2 via wired or wireless communication.

[0160] It should be noted that the "electronic device 2" can be a processing device for executing the pedestrian detection method. It can be a server, industrial control computer, embedded processing unit or other device with data processing capabilities, used to process the image data collected by the pedestrian detection terminal 3 and generate corresponding gate control signals.

[0161] In this embodiment, the pedestrian detection method can be implemented in various ways. One way is that the pedestrian detection terminal 3 itself performs the task, that is, the pedestrian detection terminal 3 integrates a processing unit to process the collected image data and directly generate gate control signals, which are then sent to the gate 1, thereby achieving localized processing and control.

[0162] Another approach is to use an external electronic device 2, where the pedestrian detection terminal 3 acts only as an image acquisition device, transmitting the acquired image data to the electronic device 2 via a communication interface. The electronic device 2 then executes each step of the pedestrian detection method and generates a gate control signal based on the processing results, which is then sent to the gate 1.

[0163] Alternatively, a hybrid approach can be adopted, where some processing steps are completed by the pedestrian detection terminal 3, such as image preprocessing or preliminary detection, while the remaining processing steps are completed by the electronic device 2, such as fusion judgment or behavior recognition, thereby allocating computing tasks between the terminal and the electronic device 2 according to system requirements.

[0164] It should be noted that the "communication connection" refers to the data interaction relationship established between devices through a communication link, which can be a network connection, a bus connection or other communication methods, used to realize the interaction of data transmission and control signals.

[0165] Through the above system structure, the execution subject of the pedestrian detection method can be flexibly selected according to actual application needs. This allows the system to achieve rapid response through local terminal processing or more complex calculations through centralized processing by external electronic device 2, thereby improving the system's adaptability and scalability. Simultaneously, through communication with the gate 1, pedestrian detection results can be directly converted into passage control actions, thus achieving effective management of the passage status.

[0166] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the pedestrian detection method in the above embodiments.

[0167] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0168] The aforementioned computer-readable storage medium may be included in an electronic device or may exist independently without being assembled into an electronic device.

[0169] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by an electronic device, cause the electronic device to perform the pedestrian detection method described above.

[0170] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0171] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0172] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0173] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the pedestrian detection method described above.

[0174] The computer program product provided in this application can solve the technical problems related to pedestrian detection methods. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the pedestrian detection methods provided in the above embodiments, and will not be repeated here.

[0175] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A pedestrian detection terminal, characterized in that, include: Install the beam body; The mounting bracket is located in the middle of the mounting beam body and is used to install the pedestrian detection terminal on the mounting carrier corresponding to the channel to be detected. A multi-angle visual acquisition component includes at least two visual acquisition units, one for acquiring pedestrian images in the direction of entry of the channel to be detected, and the other for acquiring pedestrian images in the direction of exit of the channel to be detected. A first adjustment mechanism, connected to the vision acquisition device, is used to adjust the position and / or angle of the vision acquisition device.

2. The pedestrian detection terminal as described in claim 1, characterized in that, The first adjustment mechanism includes a sliding guide rail, and an end mounting shell is provided at the end of the sliding guide rail.

3. The pedestrian detection terminal as described in claim 2, characterized in that, The vision acquisition device is disposed in the end mounting housing.

4. The pedestrian detection terminal as described in claim 1, characterized in that, It also includes a network interface, disposed on the surface of the mounting beam body, for transmitting the pedestrian image and for receiving control signals.

5. The pedestrian detection terminal as described in claim 1, characterized in that, The mounting bracket includes a second adjustment mechanism for adjusting the height of the mounting beam body.

6. A pedestrian detection method, characterized in that, The method, applied to a pedestrian detection terminal as described in any one of claims 1 to 5, comprises: The first pedestrian image and the second pedestrian image extracted from the synchronous frame images obtained from the first visual collector and the second visual collector are fused and mapped to form a fused detection image; After performing human feature detection to obtain human body contours and head count, and item detection to identify strollers and / or large luggage, the fused detection image is fused according to a preset weighting rule to form a behavior detection result, and the corresponding gate control signal is output based on the behavior detection result.

7. The method as described in claim 6, characterized in that, The synchronized frame image is obtained by aligning the timestamps or matching the frame numbers of the video streams acquired by the first and second visual acquisition devices.

8. The method as described in claim 6, characterized in that, The step of fusing and mapping the first pedestrian image and the second pedestrian image extracted from the synchronized frame images acquired by the first visual collector and the second visual collector to form a fused detection image includes: Human detection is performed on the synchronization frame images corresponding to the first and second visual acquisition devices, respectively, to obtain a first pedestrian image and a second pedestrian image; and Based on the first pedestrian image and the second pedestrian image, the corresponding image regions are cropped to form the first pedestrian image and the second pedestrian image respectively.

9. The method as described in claim 6, characterized in that, The human feature detection employs a two-level inference method, including: Human contour detection is performed on the fused detection image to obtain human candidate regions; and Head detection is performed based on the candidate human body regions to obtain head detection results.

10. The method as described in claim 6, characterized in that, The fusion determination according to the preset weighting rules includes: The number of valid personnel in the channel is determined based on the head detection results, and the behavior detection results are generated based on the number of valid personnel. The head detection results include the number of heads and head location information.

11. The method as described in claim 6, characterized in that, The fusion mapping includes: The target category, target bounding box location, and / or detection confidence in the first pedestrian image and the second pedestrian image are matched to determine the valid target.

12. The method according to any one of claims 9 to 11, characterized in that, The process of fusing and judging according to preset weight rules to form behavior detection results includes: The results are weighted and fused based on the confidence levels of the human contour detection results, head detection results, object detection results, and the corresponding detection results of the first and second visual acquisition devices.

13. The method as described in claim 6, characterized in that, The method further includes: Track the trajectory of targets within the turnstile channel to obtain the target's passage direction, passage sequence, and / or target spacing.

14. The method according to any one of claims 10 to 13, characterized in that, The process of generating a behavior detection result and outputting a corresponding gate control signal based on the behavior detection result includes: Based on the number of valid personnel and the trajectory tracking results, the behavior detection result is determined, and the corresponding gate control signal is output according to the behavior detection result and the preset access mapping table. The behavior detection results include at least one of the following behaviors: tailing behavior, passing through with a child, passing through with a stroller, passing through with large luggage, and unauthorized entry. The preset access mapping table includes a first type of behavior that allows the gate to pass and a second type of behavior that prevents the gate from passing.

15. The method as described in claim 6, characterized in that, The method further includes: When a gate control signal is output that causes the gate to block passage, a warning signal is output. In response to a reset signal, the warning state is lifted and a gate control signal is output to restore passage to the gate.

16. An electronic device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method as claimed in any one of claims 6 to 15.

17. A pedestrian detection system, characterized in that, The system includes: The pedestrian detection terminal as described in any one of claims 1 to 5, and / or the electronic device as described in claim 16; The turnstile is communicatively connected to the electronic equipment and / or the pedestrian detection terminal.

Images