Digital elevator control method and system oriented to action and image recognition

Abstract

This invention discloses a digital elevator control method and system oriented towards action and image recognition, belonging to the field of intelligent elevator control technology under multiple recognition. By using "multi-view synchronous monitoring images" and "multi-target parallel feature recognition", it utilizes action and image recognition technology to achieve panoramic real-time perception of the waiting / riding environment and deep understanding of individual passenger intentions. High-dimensional, real-time visual semantic information (actions, postures, group states) is used as the core input of the elevator control system. Then, through the sequential application of "multi-intention conflict analysis rules" (load balancing → intention conflict analysis → conditional triggering), an intelligent path from "perception" to "decision" is established. It can not only predict the trend of group passenger flow, but also analyze the intention conflicts between different passengers in real time. Under the premise of ensuring overall operating efficiency, it can conditionally trigger personalized scheduling services, realizing the unity of global optimization and individual care.

Description

Technical Field

[0001] This invention relates to the field of intelligent elevator control technology under multiple recognition, and more specifically, to a digital elevator control method and system oriented towards motion and image recognition. Background Technology

[0002] With the acceleration of urbanization and the increasing number of high-rise buildings, elevators, as the core tool of vertical transportation, are of paramount importance in terms of operational efficiency and user experience. Traditional elevator control systems rely on physical buttons (call buttons, floor buttons) as trigger signals for passive and single-mode interaction, which poses risks of contact infection, has low interaction efficiency, and can only respond after passengers press the button. It is unable to visually predict upcoming peak passenger flow (such as the end of a meeting or the crowds pouring out of the subway station) and passenger intentions, thus lacking the ability to conduct forward-looking scheduling.

[0003] A search revealed that Chinese patent document CN108639880B discloses an elevator group control system and method based on image recognition. This system is a typical example of using image recognition to collect passenger flow data, and then using the images inside the elevator car to identify the number of people and their entry and exit, followed by post-event pattern analysis to optimize macro-scheduling parameters.

[0004] Elevator dispatching based on passenger count still employs a uniform service strategy (such as fixed door opening wait times and uniform stopping logic) for all passengers and scenarios. It fails to understand and classify individual passenger behaviors and intentions, and cannot provide differentiated and personalized services. For example, the system cannot identify and provide differentiated door opening and closing times and stopping strategies for passengers carrying heavy objects, wheelchair users, or those rushing to catch the elevator. When both upward and downward passenger flows call for elevators on intermediate floors at the same time, the system can only dispatch elevators based on a predetermined algorithm (such as minimum waiting time), and cannot identify which side has more people or more urgent needs. It cannot make dynamic decisions that optimize efficiency for the group, nor does it have the ability to coordinate conflicts. As a result, elevator dispatching is not optimized, and there is still room for improvement in operational efficiency and user experience.

[0005] To address the practical technical shortcomings, a digital elevator control method and system based on motion and image recognition is proposed. Summary of the Invention

[0006] The purpose of this invention is to address practical technical deficiencies. It provides a digital elevator control method and system oriented towards action and image recognition. By using visual image feature recognition processing technology, it simultaneously identifies "individual action intention features of multiple passenger targets" and "group scene features". Based on this, it performs "multi-intention conflict analysis" and "comprehensive scheduling decision", achieving more accurate elevator demand prediction, dynamic scheduling and safety control, and improving efficiency and user experience.

[0007] The objective of this invention can be achieved through the following technical solution: a digital elevator control system oriented towards motion and image recognition, comprising: The multi-view image acquisition and processing module is used to acquire raw video data containing passengers waiting for the elevator and passengers inside the elevator car. After preprocessing, it generates monitoring images of the multi-view monitoring area at the same timestamp. The parallel feature recognition and analysis module is used to perform multi-target parallel feature recognition on continuous frame monitoring images to obtain individual action intention features of multiple passengers and group scene features of the waiting hall / car, which constitute the feature recognition results. The intent fusion decision module is used to receive feature recognition results, perform fusion reasoning and optimized scheduling decisions based on preset multi-intent conflict analysis rules, and generate the optimal elevator scheduling instruction set to cope with different real-time scenarios. The execution control module is used to execute the optimal elevator scheduling instruction set and drive the corresponding elevator scheduling operation.

[0008] Furthermore, individual action intention characteristics include at least one of the following: raising hand to call for a ladder (intention to call for a ladder going up or down), carrying heavy objects, running fast, and slowing down. Among these, carrying heavy objects, running fast, and slowing down are defined as specific action intention characteristics. The characteristics of the group scene include at least one of the following: the number of people waiting in each waiting area, the number of people riding in the elevator car, and the direction vector of the passenger group's movement.

[0009] Furthermore, the process of identifying individual action intent features includes: identifying passenger targets in the surveillance image and tracking them frame by frame; analyzing the action model to independently classify the individual action features of each passenger target and outputting the individual action intent features of each passenger target. The process of identifying group scene features includes: performing image recognition and segmentation on continuous frame monitoring images, using target detection algorithms to identify and count facial features in the monitoring images, obtaining the number of people waiting in each waiting area and the number of people riding in the elevator car, and using multi-target tracking algorithms to analyze the positional changes of passenger targets in continuous frame monitoring images and extract the motion direction vector.

[0010] Furthermore, the multi-intent conflict analysis rules include call intent priority rules, dynamic load balancing rules, and specific intent spatiotemporal priority rules; Dynamic load balancing rules are used to proactively pre-allocate load when dispatching elevators based on the number of passengers in the group scenario, avoiding full car load; elevator call intention priority rules are used to allocate elevator service priority based on the ratio of waiting passengers with same-direction and opposite-direction call intentions; specific intention spatiotemporal priority rules are used to dynamically adjust elevator door opening and closing times or stopping priorities for passengers with specific action intentions, provided that dynamic load balancing rules are met.

[0011] Furthermore, the optimal elevator scheduling instruction generation process includes: Receive the feature recognition results, evaluate the carrying capacity of each car, determine the set of candidate elevators that can be scheduled to serve the floor demand of the newly called elevator based on the dynamic load balancing rules, and generate car resource allocation instructions; Apply elevator call intent priority rules to resolve directional intent conflicts, determine the primary and secondary order of uplink and downlink services, and generate uplink and downlink elevator dispatch instructions. Apply specific spatiotemporal priorities to passenger targets identified as having specific action intent characteristics to generate personalized service instructions.

[0012] The present invention also proposes a digital elevator control method oriented towards motion and image recognition, comprising the following steps: S1. Multi-view image acquisition and preprocessing: Through image acquisition equipment deployed in the elevator lobby and inside the elevator car, raw video data is acquired synchronously, and preprocessed by timestamp alignment, image denoising and format standardization to generate standardized multi-view continuous frame monitoring images under the same timestamp. S2. Parallel Feature Recognition and Analysis: Multi-target parallel processing is performed on continuous frame surveillance images. Through multi-target detection, tracking and recognition algorithms, individual action intention features and group scene features are output to form feature recognition results. S3. Multi-intent fusion and optimization decision: Receive feature recognition results, perform hierarchical and sequential fusion reasoning based on the multi-intent conflict analysis rule set, and generate the optimal elevator scheduling instruction set; S4. Scheduling Instruction Execution and Elevator Control: Analyze the optimal elevator scheduling instruction set and control elevator operation according to the logical timing between instructions.

[0013] Furthermore, the reasoning and decision-making process in S3 is executed in the following order of sub-steps: S31: Dynamic load balancing assessment: Based on the number of passengers in the group scenario, assess the capacity of each car, apply dynamic load balancing rules, compare the number of passengers with the preset safety redundancy threshold, select elevators with the number of passengers within the preset safety redundancy threshold range, form a candidate elevator set that can be used to respond to the demand of new elevator calls, and generate car resource allocation instructions. S32: Group Intent Conflict Analysis: Based on the characteristics of individual action intent, the number of people waiting for elevators with different calling intents in the same monitoring area is counted. Then, the elevator calling intent priority rule is applied to compare the passenger scale of the upward and downward intents. From the candidate elevator set, appropriate elevator resources are allocated to the intent groups in different directions to form a preliminary elevator dispatching scheme and generate upward and downward elevator dispatching instructions. S33: Conditional Triggering of Personalized Services: Based on the initial elevator dispatching scheme, the specific intent spatiotemporal priority rules are applied for fine-tuning. For passenger targets whose specific action intent characteristics are identified, conditional judgment is made to determine whether to generate personalized service instructions. S34: Generate a structured optimal scheduling instruction set: Output a structured optimal elevator scheduling instruction set that includes car resource allocation instructions, up and down elevator dispatch instructions, and personalized service instructions.

[0014] Furthermore, the process of deciding whether to generate a personalized service instruction includes: checking whether the expected passenger load of the candidate elevator serving the passenger target is lower than a preset safety rating threshold. If so, a personalized service instruction is generated for the passenger target; otherwise, the generation of the personalized service instruction is suppressed, the original scheduling is maintained, and a guidance prompt instruction may be added.

[0015] Compared with the prior art, the advantages of this invention are: 1. This invention utilizes "multi-view synchronous monitoring images" and "multi-target parallel feature recognition" to achieve panoramic real-time perception of the elevator waiting / riding environment and deep understanding of individual passenger intentions through action and image recognition technologies. It uses high-dimensional, real-time visual semantic information (actions, postures, group states) as the core input of the elevator control system. Then, through the sequential application of "multi-intention conflict analysis rules" (load balancing → intention conflict analysis → conditional triggering), an intelligent path from "perception" to "decision-making" is established. This not only predicts the trend of group passenger flow but also analyzes the intention conflicts between different passengers in real time. Under the premise of ensuring overall operational efficiency, it conditionally triggers personalized scheduling services, achieving a unity of global optimization and individual care.

[0016] 2. This invention also collects original video data of people, objects, and environment from multi-view monitoring areas. Using computer vision and pattern recognition algorithms, it extracts multi-level and multi-dimensional feature information from the original video data. The "multi-passenger intent features" and "group scene features" obtained by image recognition technology are used as a new type of high-dimensional input information, enabling the elevator system to "see" the environment and people's behavior and "understand" the intent behind the behavior, thereby making decisions that are far more intelligent, efficient, and humane than following fixed procedures. Attached Figure Description

[0017] Figure 1This is a system principle block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] 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 some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Example 1: This invention discloses a digital elevator control system oriented towards motion and image recognition. Please refer to [link / reference]. Figures 1-2 It includes a multi-view image acquisition and processing module, a parallel feature recognition and analysis module, an intent fusion decision module, and an execution control module.

[0020] The multi-view image acquisition and processing module is deployed in the elevator lobby and inside the elevator car to collect raw video data containing passengers waiting for the elevator and passengers inside the car. After preprocessing, it generates monitoring images of the multi-view monitoring area at the same timestamp. The parallel feature recognition and analysis module communicates with the multi-view image acquisition and processing module to perform multi-target parallel feature recognition on continuous frame monitoring images, obtain individual action intention features of multiple passengers and group scene features of the waiting hall / car, and constitute the feature recognition results. Individual action intention characteristics include at least one of the following: raising hand to call for a ladder (intention to call for a ladder going up or down), carrying heavy objects, running fast, and slowing down. Among these, carrying heavy objects, running fast, and slowing down are defined as specific action intention characteristics. The characteristics of the group scene include at least one of the following: the number of people waiting in each waiting area, the number of people riding in the elevator car, and the direction vector of the passenger group's movement; The process of identifying individual action intent features includes: identifying passenger targets in the surveillance images and tracking them frame by frame; analyzing the action model and classifying the individual action features of each passenger target independently; and outputting the individual action intent features of each passenger target. Specifically, the following are the characteristics of the movement of raising a hand to call an elevator: detecting the rapid movement and brief pause of key points of the hand towards the elevator call panel area; the characteristics of the movement of carrying large objects: detecting the bending of the human body posture and the shift of the center of gravity, combined with object contour recognition; the characteristics of the movement of running fast: detecting the amplitude and frequency characteristics of limb movements; and the characteristics of the movement of slow movement: detecting the contour characteristics of slow gait and reliance on support (such as canes or wheelchairs). The process of recognizing group scene features includes: performing image recognition and segmentation on continuous frame monitoring images, using target detection algorithms to identify and count facial features in the monitoring images, obtaining the number of people waiting in each waiting area and the number of people riding in the elevator car, and using multi-target tracking algorithms to analyze the positional changes of passenger targets in continuous frame monitoring images, determine whether passengers are walking towards or leaving the elevator, and extract the motion direction vector.

[0021] This module uses multi-view images from inside the elevator car and in the waiting hall to identify multiple features such as "actions", "postures", "target objects", "number of people", "flow direction", and "abnormalities" to obtain real-time, multi-passenger individual intent sets and group scene status.

[0022] The intent fusion decision module communicates with the parallel feature recognition and analysis module to receive feature recognition results, perform fusion reasoning and optimized scheduling decisions based on preset multi-intent conflict analysis rules, resolve group intent conflicts, and generate the optimal elevator scheduling instruction set to cope with different real-time scenarios. The preset multi-intent conflict analysis rules include call intent priority rules, dynamic load balancing rules, and specific intent spatiotemporal priority rules; Dynamic load balancing rules are used to proactively pre-allocate load when dispatching elevators based on the number of passengers in the group scenario, avoiding full car load and optimizing global efficiency. The elevator call intent priority rule is used to allocate elevator service priority based on the ratio of people waiting for elevators with call intents in the same direction and opposite directions, and is used to handle directional conflicts in elevator calls. The specific intent spatiotemporal priority rule is used to dynamically adjust the elevator door opening and closing time or stopping priority for passengers with specific action intent characteristics, under the premise of satisfying the dynamic load balancing rule, so as to provide personalized care services. The generation of the optimal elevator scheduling instruction set follows a sequential logic of "resource assessment - conflict analysis - conditional execution," and the specific generation process includes: The system receives the feature recognition results, evaluates the capacity of each car, and determines the set of candidate elevators that can be dispatched to serve the new call floor based on dynamic load balancing rules. It generates a car resource allocation instruction and continuously monitors the real-time number of passengers in all cars. When a new call request is generated (the hand-raising call action feature is recognized), the dispatch algorithm will prioritize assigning the request to cars with a current number of passengers below the preset safety redundancy threshold (low load rate) and reasonable path as candidate elevators. This proactively avoids some elevators from becoming fully loaded too early, achieves capacity balancing, and generates a car resource allocation instruction to pre-assign passengers to cars with lower load. The system applies elevator call intent priority rules to resolve directional intent conflicts, determines the priority order of up and down services, and generates up and down elevator dispatch instructions. For example, when there are passenger groups with both up and down elevator call intents on a certain floor, the system compares the proportion of people in the two groups in real time. If the number of people in one group is significantly greater (e.g., more than twice the number of people in the other group), the system prioritizes dispatching elevators to serve the main direction and provides alternative solutions for the minority of passengers (e.g., prompting them to wait for the next suitable elevator). Under the premise of satisfying the first two rules, a specific spatiotemporal priority is applied to passenger targets identified with specific action intent characteristics to generate personalized service instructions. The execution of these instructions is based on the aforementioned dynamic load balancing rule, which stipulates that the system can only trigger personalized service instructions for passenger targets (passengers with special needs) whose specific action intent characteristics are identified when the total number of passengers in the target candidate elevator car and the current number of people waiting for the elevator is lower than a preset safety threshold (e.g., 80% of the rated passenger capacity). Specifically: Triggering conditions: Passenger targets with specific action intention characteristics such as carrying heavy objects, running fast, or moving slowly are identified, and the restrictions are not exceeded; Action execution: For passengers who are "carrying" or "moving slowly", extend the elevator door opening / closing time to the destination floor (e.g., extend it from the standard 3 seconds to 8 seconds), and generate instructions to extend the door opening / closing time. For passengers who are "running fast", dynamically extend the closing elevator door or reopen the door, and generate instructions to dynamically extend the closing elevator door time or reopen the door.

[0023] Limitations: If the number of people in the target elevator car exceeds the safety limit, the system will suppress the triggering of the time and space priority rules for safety and efficiency reasons. It may also cancel the stop, not extend the door opening time, or reschedule another less busy elevator to respond. This ensures that humanized service does not sacrifice the overall operating efficiency and safety of the system. The optimal elevator dispatch instruction set consists of car resource allocation instructions, up and down elevator dispatch instructions, and personalized service instructions.

[0024] The execution control module communicates with the intent fusion decision module to execute the optimal elevator scheduling instruction set and drive the corresponding elevator scheduling operation.

[0025] Multi-level and multi-dimensional feature information is extracted from the raw video data. The "multi-passenger intent features" and "group scene features" obtained by image recognition technology are used as a new type of high-dimensional input information, enabling the elevator system to "see" the environment and people's behavior, and "understand" the intentions behind the behavior. By using high-dimensional, real-time visual semantic information (actions, postures, and group states) as the core input of the elevator control system, and then applying the "multi-intent conflict analysis rules" in sequence (load balancing → intent conflict analysis → conditional triggering), an intelligent path from "perception" to "decision-making" is established. This not only predicts the trend of group passenger flow, but also analyzes the intent conflicts between different passengers in real time. Under the premise of ensuring overall operational efficiency, personalized scheduling services can be conditionally triggered, achieving a unity of global optimization and individual care.

[0026] Example 2: This invention also proposes a digital elevator control method oriented towards motion and image recognition. Please refer to [link / reference]. Figure 2 It includes the following steps: S1. Multi-view image acquisition and preprocessing: By deploying image acquisition equipment in the elevator lobby and inside the elevator car, raw video data is collected synchronously. The raw video data is then preprocessed with timestamp alignment, image denoising, and format standardization to generate standardized multi-view continuous frame monitoring images under the same timestamp. S2. Parallel Feature Recognition and Analysis: Multi-target parallel processing is performed on continuous frame surveillance images. Through multi-target detection, tracking and recognition algorithms, individual action intent features and group scene features are output simultaneously to form feature recognition results. S3. Multi-intent fusion and optimized decision-making: The feature recognition results output by S2 are received, and hierarchical and sequential fusion reasoning is performed based on the preset multi-intent conflict analysis rule set to generate a structured optimal elevator scheduling instruction set. S4. Dispatch instruction execution and elevator control: The optimal elevator dispatch instruction set generated by S3 is sent to the execution control module. This module parses the instruction set and drives the corresponding elevator to perform dispatching, door opening and closing control, and operation mode switching operations according to the logical timing between the instructions. It also synchronously controls the guidance and instruction system in the building to complete this intelligent dispatch cycle.

[0027] The reasoning and decision-making process in S3 is executed in the following order of sub-steps: S31: Dynamic load balancing assessment: Based on the number of passengers in the group scenario, assess the capacity of each car, apply dynamic load balancing rules, compare the number of passengers with the preset safety redundancy threshold, select elevators with the number of passengers within the preset safety redundancy threshold range, form a candidate elevator set that can be used to respond to the demand of new elevator calls, and generate car resource allocation instructions. S32: Group Intent Conflict Analysis: Based on the characteristics of individual action intent, the number of people waiting for elevators with different calling intents in the same monitoring area (such as the waiting hall on a certain floor) is counted. The elevator calling intent priority rule is applied to compare the passenger scale of the upward and downward intents. If the number of people in one direction is significantly greater (for example, more than twice that of the other), then that direction is determined to be the current mainstream service direction. From the candidate elevator set, appropriate elevator resources are allocated to the intent groups in different directions to form a preliminary elevator dispatching scheme and generate upward and downward elevator dispatching instructions. S33: Conditional Triggering of Personalized Services: Based on the initial elevator dispatching scheme, the specific intent spatiotemporal priority rules are applied for fine-tuning. For passenger targets whose specific action intent characteristics are identified, conditional judgment is made to determine whether to generate personalized service instructions. S34: Generate a structured optimal scheduling instruction set: Integrate the reasoning results from S31 to S33, and output a structured optimal elevator scheduling instruction set that includes car resource allocation instructions, up and down elevator dispatch instructions, and personalized service instructions.

[0028] The process of deciding whether to generate personalized service instructions includes: Check whether the expected passenger load (the total number of passengers in the target elevator car and the current number of people waiting for the elevator) of the candidate elevator serving the passenger target is lower than the preset safety rating threshold. If the safety threshold is lower than the preset safety rating threshold, a personalized service instruction is generated for the passenger target. For example, an extended door opening instruction of T seconds is configured for the target elevator on a specific floor, or a reopen instruction is generated for the closing elevator door. Conversely, the generation of personalized service instructions is suppressed, the original scheduling is maintained, and additional guidance prompts can be generated (such as a voice prompt "The elevator is full, please wait for the next one").

[0029] It should be added that the article involves comparisons of various thresholds. Thresholds, preset values, preset ranges, etc., are set for result comparison and analysis to determine good or bad. The magnitude of these thresholds is determined by a combination of large-scale model analysis of sample data and human experience. They can also be appropriately adjusted based on seasonal or common-sense influences.

[0030] In summary, this invention involves acquiring and preprocessing multi-view monitoring images, simultaneously identifying the individual action intentions of multiple passengers and group scene features, and, based on preset multi-intention conflict analysis rules, fusing and reasoning the feature recognition results to optimize scheduling decisions, generating the optimal elevator scheduling command, and executing the command to control elevator operation. By directly understanding passengers' real-time behavioral intentions through visual image recognition processing technology and dynamically coordinating group needs using conflict analysis rules, this invention overcomes the shortcomings of existing technologies, such as reliance on physical buttons, inability to predict intentions, and the use of uniform scheduling strategies. It achieves a transformation from passive response to proactive service, from uniform scheduling to personalized care, and from fixed-time adjustments to real-time dynamic optimization, significantly improving the operating efficiency, safety, and user experience of the elevator system.

[0031] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto; any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and its improved concept, should be covered within the scope of protection of the present invention.

Claims

1. A digital elevator control system oriented towards motion and image recognition, characterized in that: include: The multi-view image acquisition and processing module is used to acquire raw video data containing passengers waiting for the elevator and passengers inside the elevator car. After preprocessing, it generates monitoring images of the multi-view monitoring area at the same timestamp. The parallel feature recognition and analysis module is used to perform multi-target parallel feature recognition on continuous frame monitoring images to obtain individual action intention features of multiple passengers and group scene features of the waiting hall / car, which constitute the feature recognition results. The intent fusion decision module is used to receive feature recognition results, perform fusion reasoning and optimized scheduling decisions based on preset multi-intent conflict analysis rules, and generate the optimal elevator scheduling instruction set to cope with different real-time scenarios. The execution control module is used to execute the optimal elevator scheduling instruction set and drive the corresponding elevator scheduling operation.

2. The digital elevator control system for motion and image recognition according to claim 1, characterized in that: Individual action intention characteristics include at least one of the following: raising hand to call for a ladder (intention to call for a ladder going up or down), carrying heavy objects, running fast, and slowing down. Among these, carrying heavy objects, running fast, and slowing down are defined as specific action intention characteristics. The characteristics of the group scene include at least one of the following: the number of people waiting in each waiting area, the number of people riding in the elevator car, and the direction vector of the passenger group's movement.

3. A digital elevator control system for motion and image recognition according to claim 2, characterized in that: The process of identifying individual action intent features includes: identifying passenger targets in the surveillance images and tracking them frame by frame; analyzing the action model; classifying the individual action features of each passenger target independently; and outputting the individual action intent features of each passenger target. The process of identifying group scene features includes: performing image recognition and segmentation on continuous frame monitoring images, using target detection algorithms to identify and count facial features in the monitoring images, obtaining the number of people waiting in each waiting area and the number of people riding in the elevator car, and using multi-target tracking algorithms to analyze the positional changes of passenger targets in continuous frame monitoring images and extract the motion direction vector.

4. A digital elevator control system for motion and image recognition according to claim 3, characterized in that: Multi-intent conflict analysis rules include call intent priority rules, dynamic load balancing rules, and specific intent spatiotemporal priority rules; Dynamic load balancing rules are used to proactively pre-allocate load when dispatching elevators based on the number of passengers in the group scenario, avoiding full car load; elevator call intention priority rules are used to allocate elevator service priority based on the ratio of waiting passengers with same-direction and opposite-direction call intentions; specific intention spatiotemporal priority rules are used to dynamically adjust elevator door opening and closing times or stopping priorities for passengers with specific action intentions, provided that dynamic load balancing rules are met.

5. A digital elevator control system for motion and image recognition according to claim 4, characterized in that: The optimal elevator scheduling instruction generation process includes: Receive the feature recognition results, evaluate the carrying capacity of each car, determine the set of candidate elevators that can be scheduled to serve the floor demand of the newly called elevator based on the dynamic load balancing rules, and generate car resource allocation instructions; Apply elevator call intent priority rules to resolve directional intent conflicts, determine the primary and secondary order of uplink and downlink services, and generate uplink and downlink elevator dispatch instructions. Apply specific spatiotemporal priorities to passenger targets identified as having specific action intent characteristics to generate personalized service instructions.

6. A digital elevator control method oriented towards motion and image recognition, employing a digital elevator control system oriented towards motion and image recognition as described in any one of claims 1-5, characterized in that, Includes the following steps: S1. Multi-view image acquisition and preprocessing: Through image acquisition equipment deployed in the elevator lobby and inside the elevator car, raw video data is acquired synchronously, and preprocessed by timestamp alignment, image denoising and format standardization to generate standardized multi-view continuous frame monitoring images under the same timestamp. S2. Parallel Feature Recognition and Analysis: Multi-target parallel processing is performed on continuous frame surveillance images. Through multi-target detection, tracking and recognition algorithms, individual action intention features and group scene features are output to form feature recognition results. S3. Multi-intent fusion and optimization decision: Receive feature recognition results, perform hierarchical and sequential fusion reasoning based on the multi-intent conflict analysis rule set, and generate the optimal elevator scheduling instruction set; S4. Scheduling Instruction Execution and Elevator Control: Analyze the optimal elevator scheduling instruction set and control elevator operation according to the logical timing between instructions.

7. A digital elevator control method based on motion and image recognition according to claim 6, characterized in that: The reasoning and decision-making process in S3 is executed in the following order of sub-steps: S31: Dynamic load balancing assessment: Based on the number of passengers in the group scenario, assess the capacity of each car, apply dynamic load balancing rules, compare the number of passengers with the preset safety redundancy threshold, select elevators with the number of passengers within the preset safety redundancy threshold range, form a candidate elevator set that can be used to respond to the demand of new elevator calls, and generate car resource allocation instructions. S32: Group Intent Conflict Analysis: Based on the characteristics of individual action intent, the number of people waiting for elevators with different calling intents in the same monitoring area is counted. Then, the elevator calling intent priority rule is applied to compare the passenger scale of the upward and downward intents. From the candidate elevator set, appropriate elevator resources are allocated to the intent groups in different directions to form a preliminary elevator dispatching scheme and generate upward and downward elevator dispatching instructions. S33: Conditional Triggering of Personalized Services: Based on the initial elevator dispatching scheme, the specific intent spatiotemporal priority rules are applied for fine-tuning. For passenger targets whose specific action intent characteristics are identified, conditional judgment is made to determine whether to generate personalized service instructions. S34: Generate the optimal scheduling instruction set: Output an optimal elevator scheduling instruction set that includes car resource allocation instructions, up and down elevator dispatch instructions, and personalized service instructions.

8. A digital elevator control method based on motion and image recognition according to claim 7, characterized in that: Check whether the expected passenger load of the candidate elevator serving the passenger target is lower than the preset safety rating threshold. If so, generate a personalized service instruction for the passenger target; otherwise, suppress the generation of the personalized service instruction and add a guidance prompt instruction.

Images