Obstacle image recognition system of distribution robot

Abstract

The invention discloses an obstacle image recognition system for a distribution robot, and relates to the technical field of image recognition, and the system comprises a data collection module which is used for collecting the corresponding image information of the distribution robot in the driving process according to a preset path through a camera device arranged in the distribution robot; the image analysis module is used for analyzing and processing corresponding image information in the running process of the distribution robot according to the preset path, performing distribution safety analysis on the distribution robot based on an analysis and processing result, and making an early warning strategy according to a distribution safety analysis result; the early warning module is used for responding to the early warning strategy and outputting an early warning signal to the distribution robot; and the communication module is used for outputting notification information to a control terminal corresponding to the distribution robot. The method has the effect of improving the distribution safety of the distribution robot.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to an obstacle image recognition system for a delivery robot. Background Technology

[0002] With the development of logistics automation and service robot technology, delivery robots are being used more and more widely in communities, shopping malls, office buildings and other scenarios. Their autonomous obstacle avoidance capability is the core of ensuring delivery efficiency and safety.

[0003] In related technologies, existing obstacle recognition systems for delivery robots are limited in scope, relying heavily on the shape or color features of images. They lack the ability to extract in-depth information such as the material, size, and movement trends of obstacles, which can easily lead to misjudgments. Furthermore, their risk assessment mechanisms are crude, lacking quantitative risk grading standards. They often apply a uniform obstacle avoidance strategy to all obstacles, resulting in either excessive braking that affects efficiency or insufficient response that causes collisions. In addition, they do not incorporate historical data to optimize recognition and lack targeted marking for recurring high-risk obstacles (such as frequently piled-up clutter), leading to low efficiency in repetitive processing and consequently affecting delivery safety and stability. There are areas for improvement. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this application provides an obstacle image recognition system for delivery robots.

[0005] In a first aspect, this application provides an obstacle image recognition system for a delivery robot, comprising:

[0006] The data acquisition module is used to collect image information of the delivery robot as it travels along a preset path using the built-in camera device.

[0007] The image analysis module is used to analyze and process the image information corresponding to the delivery robot traveling along the preset path, and to perform delivery safety analysis on the delivery robot based on the analysis results, and then formulate early warning strategies based on the delivery safety analysis results;

[0008] The early warning module is used to respond to the early warning strategy and output an early warning signal to the delivery robot;

[0009] The communication module is used to output notification information to the control terminal corresponding to the delivery robot.

[0010] Preferably, the image analysis module includes an image recognition unit, a data analysis unit, and an obstacle marking unit;

[0011] The image recognition unit is used to identify and obtain the obstacle type information, obstacle location information, obstacle basic information and marking information corresponding to the delivery robot traveling along the preset path based on the image information during the delivery robot's journey along the preset path;

[0012] The data analysis unit is used to comprehensively analyze the obstacle type information, obstacle location information, basic obstacle information and marking information corresponding to the delivery robot traveling along the preset path;

[0013] The obstacle marking unit is used to mark the obstacles that the delivery robot encounters while traveling along a preset path.

[0014] Preferably, a comprehensive analysis is performed on the obstacle type information, obstacle location information, basic obstacle information, and marking information corresponding to the delivery robot's movement along the preset path, specifically including:

[0015] The risk of safety hazards during the delivery process of the delivery robot is evaluated based on the obstacle type information, obstacle location information, and marking information encountered by the delivery robot as it travels along a preset path. The evaluation process is as follows:

[0016] A delivery risk assessment model is constructed to correspond to the delivery process. The obstacle type information, obstacle location information and marking information are input into the delivery risk assessment model, and then the risk assessment coefficient of the delivery robot traveling according to the preset path is determined.

[0017] The risk assessment coefficient is compared with a preset risk assessment threshold, wherein the preset risk assessment threshold includes a first risk assessment threshold and a second risk assessment threshold, and the second risk assessment threshold is greater than the first risk assessment threshold;

[0018] If the risk assessment coefficient is less than the first risk assessment threshold, then there is no need to perform a delivery safety analysis on the delivery robot.

[0019] If the risk assessment coefficient is between the first risk assessment threshold and the second risk assessment threshold, then a delivery safety analysis of the delivery robot is required.

[0020] If the risk assessment coefficient is greater than the second risk assessment threshold, emergency braking is triggered, and an emergency warning signal is output through the warning module.

[0021] Preferably, the steps for performing a delivery safety analysis on the delivery robot include:

[0022] Acquire basic information about obstacles encountered by the delivery robot as it travels along a preset path. This basic information includes obstacle size information, obstacle material information, and traffic compliance information.

[0023] The system compares and calculates the obstacle size, material, and traffic compliance information with the standard information stored in the cloud database to determine the first safety index of the delivery robot as it travels along the preset path.

[0024] Preferably, the steps for conducting delivery safety analysis on the delivery robot also include:

[0025] The preset path corresponding to the delivery robot is obtained, and the actual path corresponding to the delivery robot is recorded in real time through the camera device built into the delivery robot. Then, the first path deviation of the delivery robot is determined based on the preset path and the actual path.

[0026] The system acquires information on obstacle type, size, and material during the delivery robot's journey along a preset path. Based on these information, it assesses the obstacles and determines their impact coefficient on the delivery robot's delivery process. The system then compares this impact coefficient with a preset impact threshold. If the impact coefficient is higher than the preset impact threshold, it acquires the obstacle's location information and determines the second path deviation of the delivery robot based on the actual path and obstacle location information.

[0027] The second safety index is determined by weighting the first path deviation and the second path deviation of the delivery robot and calculating it according to the preset path.

[0028] The first and second safety indices corresponding to the delivery robot traveling along the preset path are substituted into the preset comprehensive evaluation function to determine the comprehensive safety index corresponding to the delivery robot traveling along the preset path.

[0029] The overall safety index of the delivery robot during its journey along the preset route is compared with the preset overall safety threshold.

[0030] If the overall safety index of the delivery robot exceeds the preset overall safety threshold while it is traveling along the preset path, then there is no need for a braking warning strategy.

[0031] If the overall safety index of the delivery robot does not exceed the preset overall safety threshold while it is traveling along the preset path, a braking warning strategy is required.

[0032] Preferably, the steps of the braking warning strategy specifically include:

[0033] Obtain the risk assessment coefficient and the first risk assessment threshold, and obtain the comprehensive safety index and the preset comprehensive safety threshold, thereby confirming the risk assessment deviation and the safety index deviation respectively;

[0034] Then, based on the risk assessment deviation and safety index deviation, a comprehensive reference value for potential safety hazards during the delivery process of the delivery robot is determined;

[0035] The comprehensive reference value is compared with the preset standard reference value;

[0036] If the overall reference value is lower than the preset standard reference value, the warning module will output a first warning signal.

[0037] If the comprehensive reference value exceeds the preset standard reference value, the warning module will output a second warning signal, and at the same time, the communication module will output a notification message to the control terminal corresponding to the delivery robot.

[0038] Preferably, the process of marking obstacles encountered by the delivery robot as it travels along a preset path specifically includes:

[0039] The system obtains historical interrupted delivery records corresponding to the delivery robot, and obtains obstacle type information affecting the delivery process based on the historical interrupted delivery records. Then, it marks the obstacles corresponding to the delivery robot during its journey along the preset path based on the obstacle type information, and then confirms the marking information.

[0040] Preferably, the process of obtaining the preset path of the delivery robot involves determining the preset path corresponding to the delivery robot based on the initial position and the target delivery position of the delivery robot.

[0041] Secondly, this application provides an obstacle image recognition method for a delivery robot, comprising the following steps:

[0042] The delivery robot's built-in camera captures image information as the robot travels along a preset path.

[0043] The system analyzes and processes the image information corresponding to the delivery robot's journey along the preset path, and performs delivery safety analysis on the delivery robot based on the analysis results. Then, it formulates early warning strategies based on the results of the delivery safety analysis.

[0044] In response to the warning strategy, a warning signal is output to the delivery robot;

[0045] Output notification information to the control terminal corresponding to the delivery robot.

[0046] Thirdly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform an obstacle image recognition system for a delivery robot as described in any of the above claims.

[0047] In summary, this application includes the following beneficial technical effects:

[0048] This application provides an obstacle image recognition system for a delivery robot. The system uses a built-in camera to collect image information of the delivery robot as it travels along a preset path. It then analyzes and processes this image information, performs a delivery safety analysis based on the results, and formulates an early warning strategy. Responding to the early warning strategy, the system outputs a warning signal to the delivery robot and notification information to its control terminal. This effectively reduces the occurrence of untimely obstacle avoidance due to insufficient recognition accuracy or delayed risk assessment, thereby significantly improving the safety and stability of delivery. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of an obstacle image recognition system for a delivery robot according to an embodiment of this application.

[0051] Figure 2 This is a flowchart of the obstacle image recognition method for a delivery robot according to an embodiment of this application. Detailed Implementation

[0052] The following is in conjunction with the appendix Figure 1-2 This application will be described in further detail.

[0053] Example 1

[0054] This application discloses an obstacle image recognition system for a delivery robot.

[0055] Reference Figure 1 An obstacle image recognition system for a delivery robot, comprising:

[0056] The data acquisition module is used to collect image information of the delivery robot as it travels along a preset path using the built-in camera device.

[0057] The image analysis module is used to analyze and process the image information corresponding to the delivery robot traveling along the preset path, and to perform delivery safety analysis on the delivery robot based on the analysis results, and then formulate early warning strategies based on the delivery safety analysis results;

[0058] The early warning module is used to respond to the early warning strategy and output an early warning signal to the delivery robot;

[0059] The communication module is used to output notification information to the control terminal corresponding to the delivery robot.

[0060] In this embodiment of the application, a loading module is also included before the data acquisition module. The loading module includes a loading unit and a weight detection module. The loading unit is set up for staff to load goods onto the delivery robot. The weight detection module is used to check the weight of the loading delivery robot and determine whether the delivery robot is overweight. If the delivery robot is overweight, an overweight warning signal is output in a timely manner.

[0061] Using the above technical solution, in this embodiment, the data acquisition module activates the built-in camera device of the delivery robot to collect real-time image information of the surrounding environment while the robot travels along a preset path. This includes various visual data such as road conditions, pedestrians, vehicles, and fixed facilities, ensuring the continuity and clarity of image acquisition and providing basic data support for subsequent analysis. Next, the image analysis module receives the image information transmitted by the data acquisition module, preprocesses the images (such as noise reduction, enhancement, distortion correction, etc.), extracts obstacle features through image recognition technology, and performs delivery safety analysis in conjunction with the robot's driving status to determine whether the obstacle poses a threat to delivery. Based on the degree of threat, corresponding early warning strategies are formulated, such as deceleration prompts for minor threats, path adjustment suggestions for moderate threats, and emergency braking commands for severe threats. Subsequently, the early warning module immediately outputs corresponding early warning signals to the delivery robot according to the early warning strategies formulated by the image analysis module, such as audible and visual alarms, screen prompts, or directly controls the robot's power system to perform deceleration, steering, and other operations. At the same time, the communication module transmits early warning information, the robot's current position, obstacle conditions, and other notification information to the corresponding control terminal in real time, enabling managers to remotely monitor and intervene.

[0062] The above solution automates the entire process of obstacle recognition and safety management during delivery. The data acquisition module ensures comprehensive capture of environmental information, avoiding missed judgments due to missing information. The intelligent processing capability of the image analysis module improves the accuracy of obstacle recognition and the efficiency of safety analysis, making it faster and more objective than manual judgment. The timely response of the early warning module effectively reduces the risk of collisions, ensuring the safety of the robot and its surrounding environment. The communication module enables remote monitoring and collaborative management, allowing managers to keep abreast of the delivery status in real time and take intervention measures in emergencies. Overall, it improves the autonomous operation capability and delivery safety of the delivery robot, making it particularly suitable for complex urban roads or indoor environments.

[0063] It should be noted that the image analysis module includes an image recognition unit, a data analysis unit, and an obstacle marking unit;

[0064] The image recognition unit is used to identify and obtain the obstacle type information, obstacle location information, obstacle basic information and marking information corresponding to the delivery robot traveling along the preset path based on the image information during the delivery robot's journey along the preset path;

[0065] The data analysis unit is used to comprehensively analyze the obstacle type information, obstacle location information, basic obstacle information and marking information corresponding to the delivery robot traveling along the preset path;

[0066] The obstacle marking unit is used to mark the obstacles that the delivery robot encounters while traveling along a preset path.

[0067] Specifically, the image recognition unit in the image analysis module first extracts features from the image information transmitted by the data acquisition module. It then uses deep learning algorithms (such as convolutional neural networks) to identify obstacles in the image, determining obstacle type information (such as pedestrians, bicycles, walls, stacked objects, etc.), location information (converted from image coordinates to robot relative or absolute coordinates), and basic information. Combined with historical data, it performs preliminary labeling of common obstacles, generating labeling information. After receiving the obstacle type, location, basic information, and labeling information output by the image recognition unit, the data analysis unit performs multi-dimensional comprehensive analysis. The obstacle labeling unit then accurately labels the obstacles based on the preliminary labeling by the image recognition unit and the comprehensive analysis results of the data analysis unit, and stores the labeling results in the cloud database for reference in subsequent recognition and analysis.

[0068] It should be noted that a comprehensive analysis is performed on the obstacle type information, obstacle location information, basic obstacle information, and marking information during the delivery robot's journey along the preset path. Specifically, this includes:

[0069] The risk of safety hazards during the delivery process of the delivery robot is evaluated based on the obstacle type information, obstacle location information, and marking information encountered by the delivery robot as it travels along a preset path. The evaluation process is as follows:

[0070] A delivery risk assessment model is constructed to correspond to the delivery process. The obstacle type information, obstacle location information and marking information are input into the delivery risk assessment model, and then the risk assessment coefficient of the delivery robot traveling according to the preset path is determined.

[0071] The risk assessment coefficient is compared with a preset risk assessment threshold, wherein the preset risk assessment threshold includes a first risk assessment threshold and a second risk assessment threshold, and the second risk assessment threshold is greater than the first risk assessment threshold;

[0072] If the risk assessment coefficient is less than the first risk assessment threshold, then there is no need to perform a delivery safety analysis on the delivery robot.

[0073] If the risk assessment coefficient is between the first risk assessment threshold and the second risk assessment threshold, then a delivery safety analysis of the delivery robot is required.

[0074] If the risk assessment coefficient is greater than the second risk assessment threshold, emergency braking is triggered, and an emergency warning signal is output through the warning module.

[0075] Specifically, when conducting a comprehensive analysis of obstacle information, the first step is to construct a delivery risk assessment model. This model can be obtained by fitting historical data or machine learning. The model takes obstacle type information (e.g., dynamic obstacles have higher weight than static obstacles), location information (the closer the obstacle, the higher its weight), and marking information (high-risk markings have higher weight) as input parameters. The weight values ​​of each parameter are calculated using the analytic hierarchy process (AHP) or neural network algorithms, and then a weighted sum is used to obtain the risk assessment coefficient. The larger the coefficient, the higher the safety hazard. Next, preset risk assessment thresholds are set, where the first threshold corresponds to a low-risk state (e.g., obstacles are far away and static), and the second threshold corresponds to a high-risk state. (e.g., the obstacle is close and is a fast-moving dynamic target), and the second threshold is greater than the first threshold; then, the calculated risk assessment coefficient is compared with the two thresholds: if the risk assessment coefficient is less than the first threshold, it means that the obstacle has minimal impact on delivery safety, no additional delivery safety analysis is required, and the robot can proceed normally along the original path; if the risk assessment coefficient is between the first and second thresholds, it indicates that there is a certain safety hazard, and a detailed delivery safety analysis needs to be initiated to further assess the risk; if the risk assessment coefficient is greater than the second threshold, it means that there is an emergency danger, and the robot's emergency braking system is immediately triggered, while the warning module outputs an emergency warning signal containing the obstacle's location and type to alert surrounding personnel.

[0076] By employing the aforementioned technical solution, the risk assessment mechanism achieves precise control over obstacle risks through quantitative analysis and tiered processing, avoiding a "one-size-fits-all" approach: low-risk situations do not interfere with normal delivery, ensuring delivery efficiency; medium-risk situations trigger detailed analysis, balancing safety and efficiency; and high-risk situations require emergency braking to minimize collision risk. The construction of the risk assessment model shifts risk judgment from experience-driven to data-driven, improving the scientific rigor and consistency of decision-making. The setting of different thresholds provides the system with clear action guidelines, reducing decision-making delays. Especially in dynamic and complex environments, it can quickly respond to sudden obstacles, significantly improving the safety performance of the delivery robot.

[0077] It should be noted that the steps for conducting a delivery safety analysis on the delivery robot specifically include:

[0078] Acquire basic information about obstacles encountered by the delivery robot as it travels along a preset path. This basic information includes obstacle size information, obstacle material information, and traffic compliance information.

[0079] The system compares and calculates the obstacle size, material, and traffic compliance information with the standard information stored in the cloud database to determine the first safety index of the delivery robot as it travels along the preset path.

[0080] Specifically, when the risk assessment coefficient falls between the first and second risk assessment thresholds, and a delivery safety analysis is required, the basic information of the obstacle is first obtained through an image recognition unit. This includes: obtaining obstacle size information (such as height, width, and volume) through image size measurement and proportional conversion; obtaining obstacle material information (such as metal, plastic, fabric, and living organisms) through image color and texture feature recognition; and determining traffic compliance information (such as whether the obstacle is in a prohibited area or violates traffic rules) by comparing it with a traffic rules database. Next, the above basic information is compared and calculated with standard information stored in a cloud database. For example, the obstacle size is compared with the robot's safe passage clearance standard to calculate the size matching degree; the material information is compared with obstacle hardness and fragility standard data to assess the severity of the collision consequences; and the traffic compliance information is compared with the violation risk level standard to determine the reasonable risk of the obstacle. Finally, the size matching degree, collision consequence assessment value, and compliance risk level are integrated using a preset algorithm (such as weighted average) to calculate the first safety index. The higher the index, the lower the safety risk posed by the obstacle's own characteristics.

[0081] By employing the above technical solution, the calculation of the first safety index focuses on the key characteristics of the obstacle itself. Through comparison with standard information in the cloud database, the risk assessment achieves standardization and objectivity, avoiding biases from subjective judgment: analysis of size information ensures the robot can determine whether there is sufficient space to avoid the obstacle; assessment of material information can predict potential collision damage; and consideration of traffic compliance information, combined with environmental rules, makes the safety analysis more aligned with real-world scenarios. Compared to assessment methods that rely solely on distance or type, these steps more comprehensively reflect the safety impact of obstacles, providing accurate foundational data for subsequent comprehensive safety assessments and enhancing the depth and reliability of the safety analysis.

[0082] It should be noted that the steps for conducting delivery safety analysis on delivery robots also include:

[0083] The preset path corresponding to the delivery robot is obtained, and the actual path corresponding to the delivery robot is recorded in real time through the camera device built into the delivery robot. Then, the first path deviation of the delivery robot is determined based on the preset path and the actual path.

[0084] The system acquires information on obstacle type, size, and material during the delivery robot's journey along a preset path. Based on these information, it assesses the obstacles and determines their impact coefficient on the delivery robot's delivery process. The system then compares this impact coefficient with a preset impact threshold. If the impact coefficient is higher than the preset impact threshold, it acquires the obstacle's location information and determines the second path deviation of the delivery robot based on the actual path and obstacle location information.

[0085] The second safety index is determined by weighting the first path deviation and the second path deviation of the delivery robot and calculating it according to the preset path.

[0086] The first and second safety indices corresponding to the delivery robot traveling along the preset path are substituted into the preset comprehensive evaluation function to determine the comprehensive safety index corresponding to the delivery robot traveling along the preset path.

[0087] The overall safety index of the delivery robot during its journey along the preset route is compared with the preset overall safety threshold.

[0088] If the overall safety index of the delivery robot exceeds the preset overall safety threshold while it is traveling along the preset path, then there is no need for a braking warning strategy.

[0089] If the overall safety index of the delivery robot does not exceed the preset overall safety threshold while it is traveling along the preset path, a braking warning strategy is required.

[0090] Specifically, in delivery safety analysis, in addition to the first safety index, a second safety index needs to be calculated. The specific steps include: First, obtaining the coordinate information of the preset path through the robot's positioning module, and simultaneously recording the coordinates of the actual driving path in real time through camera devices and motion sensors, calculating the positional deviation of the two paths within the same time period to obtain the first path deviation degree, which reflects the stability of the robot's own driving; Second, judging the impact of obstacles on delivery by combining the type, size, and material information of obstacles. For example, the impact of large rigid dynamic obstacles is higher than that of small flexible static obstacles. The impact is quantified into an impact coefficient by an assignment method. If the coefficient is higher than the preset impact threshold (i.e., the obstacle significantly affects the path), then the robot is calculated to avoid the obstacle based on the obstacle's position information and the actual path. The potential additional deviations caused by obstacles are used to obtain a second path deviation degree. Then, a weighted average is calculated based on the first path deviation degree (with lower weight, reflecting inherent deviation) and the second path deviation degree (with higher weight, reflecting deviation caused by obstacles) to obtain a second safety index. The higher the index, the lower the risk caused by path deviation. Next, the first and second safety indices are substituted into a preset comprehensive evaluation function (such as the weighted sum or product of the two). The preset comprehensive evaluation function can be obtained through a machine learning model to obtain a comprehensive safety index. Finally, the comprehensive safety index is compared with a preset comprehensive safety threshold: if the comprehensive safety index exceeds the preset comprehensive safety threshold, it means that the overall safety risk is within an acceptable range and no braking warning is required; if it does not exceed the threshold, a braking warning strategy needs to be activated.

[0091] By adopting the above technical solution and introducing path deviation analysis, the limitations of solely assessing the characteristics of obstacles are overcome. The solution comprehensively considers the interaction between the robot's driving state and obstacles: the first path deviation reflects the robot's control precision, and the second path deviation reflects the interference of obstacles on the driving trajectory. The combination of these two factors allows for a comprehensive assessment of path safety. The comprehensive safety index integrates the dual impact of obstacle characteristics and path deviation, making the safety assessment more systematic and comprehensive. By comparing the index with the comprehensive safety threshold, it is possible to accurately determine whether a warning is needed, avoiding overreaction to minor deviations and promptly identifying potential path conflict risks, significantly improving the accuracy and practicality of delivery safety analysis.

[0092] It should be noted that the steps of the braking warning strategy specifically include:

[0093] Obtain the risk assessment coefficient and the first risk assessment threshold, and obtain the comprehensive safety index and the preset comprehensive safety threshold, thereby confirming the risk assessment deviation and the safety index deviation respectively;

[0094] Then, based on the risk assessment deviation and safety index deviation, a comprehensive reference value for potential safety hazards during the delivery process of the delivery robot is determined;

[0095] The comprehensive reference value is compared with the preset standard reference value;

[0096] If the overall reference value is lower than the preset standard reference value, the warning module will output a first warning signal.

[0097] If the comprehensive reference value exceeds the preset standard reference value, the warning module will output a second warning signal, and at the same time, the communication module will output a notification message to the control terminal corresponding to the delivery robot.

[0098] Specifically, when the comprehensive safety index does not exceed the preset threshold and a braking warning strategy needs to be activated, the risk assessment deviation and safety index deviation are first calculated: the risk assessment deviation is the difference between the current risk assessment coefficient and the first risk assessment threshold, reflecting the degree to which the risk exceeds the low-risk range; the safety index deviation is the difference between the preset comprehensive safety threshold and the current comprehensive safety index, reflecting the degree to which the safety level is below the standard. Then, the two deviations are integrated using a preset formula (such as a weighted sum of the two) to obtain a comprehensive reference value. The larger this value, the more serious the safety hazard. Next, the comprehensive reference value is compared with the preset standard reference value. Value (set based on historical safety data) comparison: If the comprehensive reference value is lower than the standard reference value, it indicates that the safety hazard is minor. The early warning module outputs the first early warning signal, such as controlling the robot to slow down, turning on the warning lights, and displaying "Caution: Avoid obstacles" on the robot's display screen. If the comprehensive reference value exceeds the standard reference value, it indicates that the safety hazard is serious. The early warning module outputs the second early warning signal, such as controlling the robot to stop moving and issuing a buzzer alarm. At the same time, the communication module sends information such as obstacle details, robot location, and early warning level to the control terminal to notify management personnel to intervene in a timely manner. If necessary, the robot can be remotely controlled to bypass the obstacle.

[0099] By employing the above technical solution, a braking warning strategy is used to achieve graded early warnings through quantified deviation, making the response measures more targeted: the first warning signal is applicable to minor risks, minimizing the impact on delivery efficiency while ensuring safety; the second warning signal is applicable to higher risks, ensuring timely handling of risks through stricter braking measures and remote notifications. The calculation of the comprehensive reference value integrates both risk assessment and safety index information, making the judgment of the warning level more accurate and avoiding misjudgments that may be caused by a single indicator; the linkage between the communication module and the control terminal realizes human-machine collaboration, introducing human intervention when the robot's autonomous processing capabilities are insufficient, improving the risk response capability in complex scenarios, and improving the overall safety and reliability of the delivery system.

[0100] Furthermore, the process of marking obstacles encountered by the delivery robot as it travels along the preset path specifically includes:

[0101] The system obtains historical interrupted delivery records corresponding to the delivery robot, and obtains obstacle type information affecting the delivery process based on the historical interrupted delivery records. Then, it marks the obstacles corresponding to the delivery robot during its journey along the preset path based on the obstacle type information, and then confirms the marking information.

[0102] Specifically, when marking obstacles, the system first retrieves historical delivery interruption records of the delivery robot from the control terminal or cloud database via the communication module. These records include the interruption time, location, cause, and corresponding obstacle descriptions. Then, the historical records are cleaned and analyzed to extract key obstacle type information that caused the delivery interruption. For example, "low-lying stacks" that repeatedly cause robot jams and "suddenly rushing pedestrians" that trigger collisions are used to establish a high-risk obstacle type database. Next, during the robot's current movement, after the image recognition unit identifies an obstacle, its type information is compared with the high-risk obstacle type database. If a match is found, the obstacle is marked, with marking information including "historical high risk" and "interruption risk level." Simultaneously, the marking information is associated with and stored with the obstacle's location, size, and other information. In subsequent data analysis, the marking information is given higher weight, affecting the calculation of risk assessment coefficients and safety indices. Furthermore, the marking process is continuously updated; if a certain type of obstacle does not cause further interruptions in subsequent deliveries, its marking level gradually decreases, and vice versa.

[0103] By employing the above technical solution, the marking process based on historical interruption records enables obstacle recognition to possess "experience learning" capabilities, prioritizing obstacles that have previously caused problems and improving the targeting and accuracy of risk prediction. The introduction of historical data avoids the robot's "ignorance" of potentially high-risk obstacles in new environments, triggering early warning mechanisms through marking. Dynamic updates to the marking information ensure the system adapts to environmental changes, avoiding misjudgments caused by outdated markings. Assigning higher weight to the marking information in data analysis makes risk assessment more aligned with actual delivery experience, reducing the discrepancy between "theoretical assessment" and "actual risk." Compared to marking methods without historical data, this step significantly improves the robot's sensitivity to high-risk obstacles, reduces the probability of recurring similar delivery interruptions, and enhances the system's robustness.

[0104] Furthermore, the process of obtaining the preset path for the delivery robot involves confirming the preset path corresponding to the delivery robot based on its initial position and the target delivery position.

[0105] Specifically, the process of obtaining the preset path for a delivery robot begins with determining the initial location and the target delivery location. The initial location is usually the robot's starting point (such as a delivery station or warehouse exit), and its precise coordinates are obtained through the robot's GPS positioning module or indoor positioning system. The target delivery location is the delivery location of the goods (such as the user's doorstep or a designated pickup point), and its coordinate information is provided by the control terminal. Next, the initial location and the target delivery location are input into a path planning algorithm (such as the A* algorithm, Dijkstra's algorithm, or a deep learning path planning model), and combined with electronic map data (including road layout, traffic restrictions, etc.). The system calculates information such as obstacle distribution, prioritizing paths with the shortest distance, lowest difficulty of passage, and few historical obstacles. During path planning, it also considers the robot's driving capabilities (such as maximum climbing angle and minimum turning radius), delivery time requirements (such as prioritizing routes with fewer traffic lights), and real-time traffic conditions (such as temporary traffic control information obtained through a cloud platform) to optimize and adjust the initially planned path. The final generated preset path is stored in the robot's control system as a sequence of coordinate points, serving as a reference for driving navigation. At the same time, the path information is synchronized to the image analysis module for subsequent path deviation calculation.

[0106] Specifically, when the delivery robot arrives at the target delivery location, the receiving personnel need to open the door by scanning the corresponding identification code of the delivery robot and complete the unloading. If the receiving personnel have goods to return, they need to place the returned goods inside the delivery robot, and then the delivery robot will return according to the preset return path; the delivery robot is also equipped with an automatic parking unit.

[0107] By employing the above technical solution, pre-planned paths based on initial and target locations ensure the directionality and efficiency of delivery, preventing robots from getting lost or taking detours in complex environments. The planning process, combining map data and real-time information, makes the path more closely match actual traffic conditions, reducing delivery failures due to impassable roads. Optimization and adjustment considering the robot's own capabilities ensure path feasibility, avoiding the planning of routes the robot cannot execute (such as overly steep slopes). The pre-planned path serves as a benchmark, providing a reference framework for subsequent path deviation analysis and obstacle avoidance, enabling the robot to clearly determine the compatibility of its driving state with the environment. Compared to random driving or fixed routes, this path acquisition method significantly improves delivery efficiency, reduces unnecessary energy consumption and time waste, and provides a clear navigation benchmark for the entire obstacle recognition system, ensuring the targeted and effective nature of safety analysis.

[0108] Example 2

[0109] This application also discloses an obstacle image recognition method for delivery robots.

[0110] Reference Figure 2An obstacle image recognition method for a delivery robot includes the following steps:

[0111] The delivery robot's built-in camera captures image information as the robot travels along a preset path.

[0112] The system analyzes and processes the image information corresponding to the delivery robot's journey along the preset path, and performs delivery safety analysis on the delivery robot based on the analysis results. Then, it formulates early warning strategies based on the results of the delivery safety analysis.

[0113] In response to the warning strategy, a warning signal is output to the delivery robot;

[0114] Output notification information to the control terminal corresponding to the delivery robot.

[0115] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention, they should all fall within the protection scope of the present invention.

[0116] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0117] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. An obstacle image recognition system for a delivery robot, characterized in that, include: The data acquisition module is used to collect image information of the delivery robot as it travels along a preset path using the built-in camera device. The image analysis module is used to analyze and process the image information corresponding to the delivery robot traveling along the preset path, and to perform delivery safety analysis on the delivery robot based on the analysis results, and then formulate early warning strategies based on the delivery safety analysis results; The early warning module is used to respond to the early warning strategy and output an early warning signal to the delivery robot; The communication module is used to output notification information to the control terminal corresponding to the delivery robot.

2. The obstacle image recognition system for a delivery robot according to claim 1, characterized in that, The image analysis module includes an image recognition unit, a data analysis unit, and an obstacle marking unit; The image recognition unit is used to identify and obtain the obstacle type information, obstacle location information, obstacle basic information and marking information corresponding to the delivery robot traveling along the preset path based on the image information during the delivery robot's journey along the preset path; The data analysis unit is used to comprehensively analyze the obstacle type information, obstacle location information, basic obstacle information and marking information corresponding to the delivery robot traveling along the preset path; The obstacle marking unit is used to mark the obstacles that the delivery robot encounters while traveling along a preset path.

3. The obstacle image recognition system for a delivery robot according to claim 2, characterized in that, A comprehensive analysis is performed on the obstacle type information, obstacle location information, basic obstacle information, and marker information encountered by the delivery robot as it travels along the preset path. Specifically, this includes: The risk of safety hazards during the delivery process of the delivery robot is evaluated based on the obstacle type information, obstacle location information, and marking information encountered by the delivery robot as it travels along a preset path. The evaluation process is as follows: A delivery risk assessment model is constructed to correspond to the delivery process. The obstacle type information, obstacle location information and marking information are input into the delivery risk assessment model, and then the risk assessment coefficient of the delivery robot traveling according to the preset path is determined. The risk assessment coefficient is compared with a preset risk assessment threshold, wherein the preset risk assessment threshold includes a first risk assessment threshold and a second risk assessment threshold, and the second risk assessment threshold is greater than the first risk assessment threshold; If the risk assessment coefficient is less than the first risk assessment threshold, then there is no need to perform a delivery safety analysis on the delivery robot. If the risk assessment coefficient is between the first risk assessment threshold and the second risk assessment threshold, then a delivery safety analysis of the delivery robot is required. If the risk assessment coefficient is greater than the second risk assessment threshold, emergency braking is triggered, and an emergency warning signal is output through the warning module.

4. The obstacle image recognition system for a delivery robot according to claim 3, characterized in that, The steps for conducting a delivery safety analysis on delivery robots include: Acquire basic information about obstacles encountered by the delivery robot as it travels along a preset path. This basic information includes obstacle size information, obstacle material information, and traffic compliance information. The system compares and calculates the obstacle size, material, and traffic compliance information with the standard information stored in the cloud database to determine the first safety index of the delivery robot as it travels along the preset path.

5. The obstacle image recognition system for a delivery robot according to claim 4, characterized in that, The steps for conducting a delivery safety analysis on delivery robots also include: The preset path corresponding to the delivery robot is obtained, and the actual path corresponding to the delivery robot is recorded in real time through the camera device built into the delivery robot. Then, the first path deviation of the delivery robot is determined based on the preset path and the actual path. The system acquires information on obstacle type, size, and material during the delivery robot's journey along a preset path. Based on these information, it assesses the obstacles and determines their impact coefficient on the delivery robot's delivery process. The system then compares this impact coefficient with a preset impact threshold. If the impact coefficient is higher than the preset impact threshold, it acquires the obstacle's location information and determines the second path deviation of the delivery robot based on the actual path and obstacle location information. The second safety index is determined by weighting the first path deviation and the second path deviation of the delivery robot and calculating it according to the preset path. The first and second safety indices corresponding to the delivery robot traveling along the preset path are substituted into the preset comprehensive evaluation function to determine the comprehensive safety index corresponding to the delivery robot traveling along the preset path. The overall safety index of the delivery robot during its journey along the preset route is compared with the preset overall safety threshold. If the overall safety index of the delivery robot exceeds the preset overall safety threshold while it is traveling along the preset path, then there is no need for a braking warning strategy. If the overall safety index of the delivery robot does not exceed the preset overall safety threshold while it is traveling along the preset path, a braking warning strategy is required.

6. The obstacle image recognition system for a delivery robot according to claim 5, characterized in that, The steps of the braking warning strategy specifically include: Obtain the risk assessment coefficient and the first risk assessment threshold, and obtain the comprehensive safety index and the preset comprehensive safety threshold, thereby confirming the risk assessment deviation and the safety index deviation respectively; Then, based on the risk assessment deviation and safety index deviation, a comprehensive reference value for potential safety hazards during the delivery process of the delivery robot is determined; The comprehensive reference value is compared with the preset standard reference value; If the overall reference value is lower than the preset standard reference value, the warning module will output a first warning signal. If the comprehensive reference value exceeds the preset standard reference value, the warning module will output a second warning signal, and at the same time, the communication module will output a notification message to the control terminal corresponding to the delivery robot.

7. The obstacle image recognition system for a delivery robot according to claim 2, characterized in that, The process of marking obstacles encountered by the delivery robot as it travels along a preset path specifically includes: The system obtains historical interrupted delivery records corresponding to the delivery robot, and obtains obstacle type information affecting the delivery process based on the historical interrupted delivery records. Then, it marks the obstacles corresponding to the delivery robot during its journey along the preset path based on the obstacle type information, and then confirms the marking information.

8. The obstacle image recognition system for a delivery robot according to claim 1, characterized in that, The process of obtaining the preset path for the delivery robot involves confirming the preset path corresponding to the delivery robot based on its initial position and target delivery position.

9. An obstacle image recognition method for a delivery robot, applied to the obstacle image recognition system for a delivery robot as described in any one of claims 1-8, characterized in that, Includes the following steps: The delivery robot's built-in camera captures image information as the robot travels along a preset path. The system analyzes and processes the image information corresponding to the delivery robot's journey along the preset path, and performs delivery safety analysis on the delivery robot based on the analysis results. Then, it formulates early warning strategies based on the results of the delivery safety analysis. In response to the warning strategy, a warning signal is output to the delivery robot; Output notification information to the control terminal corresponding to the delivery robot.

10. A computer-readable storage medium, characterized in that: The system stores instructions that, when executed on a computer, cause the computer to perform an obstacle image recognition system for a delivery robot as described in any one of claims 1 to 8.

Images