A visual analysis method for household robots

Abstract

The present application relates to the field of visual analysis, and in particular to a visual analysis method for a household robot. The method includes: obtaining a real-time following image, and determining real-time behavior information and home environment information based on the real-time following image; obtaining following information, and determining abnormal following information based on the real-time behavior information, following information and home environment information; adjusting the following information based on the abnormal following information and the real-time following image, determining the real-time following information, and following family members based on the real-time following information. The present application obtains following images in real time, accurately identifies and analyzes the behavior information and environmental conditions of family members, and determines real-time following information. By comparing the real-time behavior information with the home environment, abnormal following situations are detected, thereby optimizing the following path and behavior. The adjusted following information ensures that the robot adapts to the activities of family members more intelligently, improves the home automation experience, and enhances the safety of family members.

Description

Technical Field

[0001] The present application relates to the field of visual analysis, and in particular to a visual analysis method for a household robot. Background Art

[0002] With the continuous improvement of people's living standards and economic and social development levels, the application of household robots has become more and more extensive and popular. When using household robots to monitor and follow home residents, it can improve their living standards at home while ensuring their safety.

[0003] However, due to the existence of many dynamic uncertain factors within the home, the monitoring and following status of the household robot is easily affected by the dynamic factors within the home, thereby reducing the monitoring level of the household robot. Summary of the invention

[0004] The present application provides a visual analysis method for a household robot to solve the above-mentioned problems.

[0005] In a first aspect, the present application provides a visual analysis method for a household robot, the method comprising:

[0006] Acquire a real-time following image, and determine real-time behavior information and home environment information according to the real-time following image;

[0007] Acquire follow-up information, and determine abnormal follow-up information according to the real-time behavior information, the follow-up information and the home environment information;

[0008] According to the abnormal following information and the real-time following image, the following information is adjusted to determine the real-time following information, and the family members are followed according to the real-time following information.

[0009] Through this solution, by capturing real-time following images, the real-time behavior information of family members and home environment information can be accurately identified. The following information of the household robot itself is combined with the real-time behavior information and home environment information for comprehensive analysis to identify any abnormal following behavior, such as misjudgment of family members' movements or sudden complex situations. Once an abnormality is detected, the following information is immediately adjusted intelligently based on the abnormal following information and real-time following images, and new real-time following information is generated and executed. This process ensures that the household robot can flexibly adapt to changes in the home environment and family members' behavior, achieve stable and accurate following of family members, and enhance the intelligence and convenience of family life.

[0010] Optionally, determining the real-time behavior information and home environment information according to the real-time following image includes:

[0011] Analyzing the real-time tracking image to determine the movement characteristics of key parts of the person being followed;

[0012] Determining the walking speed and walking direction of the person being followed according to the movement characteristics;

[0013] Incorporating the walking speed and the walking direction into the real-time behavior information;

[0014] Determine the home space structure and home item types based on the real-time tracking image;

[0015] Determining a following road width in the home according to the home space structure and the types of home items;

[0016] The following road width is incorporated into the home environment information.

[0017] Through this solution, the key parts of the person being followed are accurately identified and tracked through the analysis of real-time following images, and then their motion characteristics, such as gait, movement amplitude, etc., are extracted, and the walking speed and walking direction of the person being followed are further calculated. These motion data are integrated into real-time behavior information to reflect the dynamic situation of the person being followed in real time. In addition, by processing real-time following images, the home space structure and home item types are analyzed to provide support for the accurate modeling of environmental characteristics. On this basis, the width of the following road in the home is calculated to ensure that it conforms to the actual space layout. All of this information is incorporated into the home environment information, and a quick response is made according to the home environment and the behavioral changes of the person being followed, which helps to improve the intelligent following ability of the home robot in the home space.

[0018] Optionally, the real-time behavior information includes walking stride, the following information includes following speed, following direction and following distance, and determining abnormal following information according to the real-time behavior information, the following information and the home environment information includes:

[0019] determining whether the person being followed falls backwards according to the walking speed, the walking stride and the walking direction;

[0020] If the person falls backwards, the falling direction, falling speed and height of the person being followed are determined based on the real-time following image;

[0021] Predicting a backward backward distance of the person being followed according to the backward direction, the backward speed and the height of the person;

[0022] Determining a reverse trajectory of the person according to the following road width, the reverse direction, and the reverse speed;

[0023] Determining a following trajectory according to the following speed and the following direction;

[0024] According to the backward distance, determining whether the following trajectory and the backward trajectory of the person have an intersection;

[0025] If an intersection is generated, the intersection position is determined according to the following moving trajectory and the backward trajectory of the person, and the intersection position is determined as the collision position;

[0026] The collision position is incorporated into the abnormal follow-up information.

[0027] Through this solution, by accurately analyzing the walking speed, walking stride and walking direction of the person being followed, it is possible to determine in real time whether there is a risk of falling over. When it is detected that the person being followed has signs of falling over, the falling direction, falling speed and height of the person are further determined based on the real-time following image, thereby predicting the falling distance. Based on these data, the person's falling trajectory is calculated, and spatial calculations are performed in combination with the width of the following road to predict the person's possible falling path. At the same time, based on the following speed and direction, the following trajectory is inferred, and it is determined whether the trajectory intersects with the person's falling trajectory. If there is an intersection, the collision position is located and marked as the collision position, thereby generating abnormal following information. The implementation of this process effectively prevents and handles the risk of collision that may be caused by the falling over of the person being followed, improves the safety and responsiveness of the household robot, and ensures the safety of people in the home environment.

[0028] Optionally, adjusting the following information according to the abnormal following information and the real-time following image to determine the real-time following information includes:

[0029] Determining the volume and height of the person being followed according to the real-time following image;

[0030] Determine the volume of the space where the followed person falls backwards according to the collision position, the person's volume, the person's height and the person's falling trajectory;

[0031] Determine a real-time avoidance angle according to the volume of the backward space, the following speed, the volume of the person, the following distance and the height of the person;

[0032] The real-time avoidance angle is incorporated into the real-time following information.

[0033] Through this solution, the volume and height of the person being followed are accurately measured through real-time following images, providing basic data for subsequent collision prevention. After detecting the collision position, the volume of the person being followed when falling backwards is accurately calculated by combining the person's volume, height and falling trajectory. This data provides an important basis for judging the space required for a person to fall backwards. According to the falling space volume, following speed, person volume, following distance and person height, the real-time avoidance angle is calculated to ensure that the household robot responds in time during the person's falling process to avoid collisions. The real-time avoidance angle is incorporated into the real-time following information, enhancing the intelligent response capability of the household robot. The dynamic response speed of the household robot is greatly improved, and the probability of accidents is reduced while adjusting the following behavior in real time.

[0034] Optionally, determining the real-time avoidance angle according to the volume of the backward space, the following speed, the volume of the person, the following distance and the height of the person includes:

[0035] determining, according to the real-time following image, an inclination angle formed between the body of the followed person and the wall when the followed person falls backwards;

[0036] The real-time avoidance angle is determined according to the volume of the backward space, the following speed, the inclination angle, the volume of the person, the following distance and the height of the person, and is calculated according to the following formula:

[0037] ;

[0038] in, represents the real-time avoidance angle, represents the volume of the person, represents the following distance, represents the height of the person, represents the following speed, represents the tilt angle.

[0039] Through this solution, the tilt angle formed between the body and the wall of the person being followed when falling backwards is accurately determined through real-time following image analysis, which is used as the key parameter for subsequent avoidance calculation. On this basis, the real-time avoidance angle is calculated using a formula in combination with the volume of the falling space, the following speed, the volume of the person, the following distance and the height of the person. The various variables in the formula, such as the volume of the person, the following distance, the height of the person, the following speed and the tilt angle, jointly determine the avoidance angle. Through this calculation, the following angle is adjusted in time during the process of the person falling backwards, ensuring that the household robot can flexibly avoid according to the real-time situation and avoid potential collision risks. The introduction of the tilt angle in the formula further enhances the sensitivity of the household robot to the falling posture, making the calculation of the avoidance angle more accurate and better able to cope with complex spatial environments. Through the analysis of real-time data through precise mathematical formulas, efficient dynamic avoidance is achieved, which significantly improves the safety and response speed of the household robot in the home environment.

[0040] Optionally, after determining the real-time avoidance angle according to the volume of the backward space, the following speed, the volume of the person, the following distance and the height of the person, the method further includes:

[0041] Acquire a rear monitoring image, and determine whether a rear obstacle moves according to the rear monitoring image;

[0042] If movement occurs, determining the movement speed, movement direction and volume of the obstacle based on the rear monitoring image;

[0043] Determining a direction angle according to the moving direction and the real-time avoidance angle;

[0044] According to the movement speed, the movement direction, the direction angle, the obstacle volume and the real-time avoidance angle, the secondary avoidance angle of the robot when an obstacle appears behind is determined and calculated according to the following formula:

[0045] ;

[0046] in, represents the secondary avoidance angle, represents the real-time avoidance angle, represents the direction angle, represents the obstacle volume, Indicates the speed of the movement.

[0047] Through this solution, by acquiring the rear monitoring image, it is possible to monitor in real time whether there are obstacles in the rear and determine whether they are moving. If the obstacle is found to be moving, the movement speed, direction and volume of the obstacle are determined according to the monitoring image, thereby providing necessary information for adjusting the avoidance path. By calculating the directional angle between the movement direction of the obstacle and the real-time avoidance angle, the avoidance requirements can be accurately understood. Furthermore, the secondary avoidance angle is calculated using a formula, which takes into account factors such as the real-time avoidance angle, directional angle, obstacle volume, and obstacle movement speed. This calculation can help the household robot flexibly adjust the avoidance trajectory according to the real-time status of the rear obstacle, and effectively deal with the risks brought by dynamic obstacles. Through precise mathematical models and dynamic adjustment mechanisms, efficient avoidance in complex environments is achieved, which improves the intelligent response ability and safety of the household robot in the real-time following process.

[0048] Optionally, determining abnormal following information according to the real-time behavior information, the following information and the home environment information includes:

[0049] Determining whether the dropped object bounces according to the real-time tracking image;

[0050] If a bounce occurs, the volume of the object, the bounce height and the bounce direction of the object after it falls are determined based on the real-time tracking image;

[0051] According to the bouncing height and the bouncing direction, determining whether the walking speed and the walking direction affect the bouncing of the object;

[0052] If so, determining the degree of influence of the person's behavior on the bouncing of the object according to the walking speed, the walking direction and the volume of the object;

[0053] Determining a bouncing trajectory of the object according to the walking impact degree, the bouncing height, and the bouncing direction;

[0054] Determining a track contact position and a track contact height according to the bouncing track and the following traveling track;

[0055] The track contact position and the track contact height are incorporated into the abnormal following information.

[0056] Through this solution, through real-time following image analysis, it is determined whether the dropped object bounces. If the object bounces, the volume, bounce height and bounce direction of the object are further analyzed to accurately predict the movement trajectory of the object. Combined with the walking speed and walking direction, it is determined whether the behavior of the person being followed has an impact on the bounce of the object. If the walking behavior of the person being followed has an impact on the bounce of the object, the degree of the person's behavior on the bounce of the object will be evaluated based on the walking speed, walking direction and object volume. Through this process, the bounce trajectory of the object is accurately calculated and compared with the follow-up travel trajectory to determine the trajectory contact position and contact height. All relevant trajectory contact information is included in the abnormal following information, providing a basis for further processing and emergency response. Effectively predict and manage the safety hazards that may be caused by the bounce of objects, and enhance the adaptability of household robots in dynamic environments.

[0057] Optionally, adjusting the following information according to the abnormal following information and the real-time following image to determine the real-time following information includes:

[0058] Determining the relative distance between the robot and the bouncing object according to the real-time following image;

[0059] According to the following speed, the track contact height, the relative distance, the object volume and the bouncing direction, the moving distance of the robot when avoiding the bouncing object is determined and calculated according to the following formula:

[0060] ;

[0061] in, represents the moving distance, represents the relative distance, represents the following speed, Indicates the volume of the item, represents the track contact height, Indicates the bounce direction;

[0062] The moving distance is incorporated into the real-time following information.

[0063] Through this scheme, the relative distance between the household robot and the bouncing object is determined by real-time following images, which provides important spatial information for the household robot in the avoidance process. On this basis, the moving distance of the household robot when avoiding the bouncing object is calculated by combining the following speed, track contact height, relative distance, object volume and bouncing direction. The various variables in the formula jointly affect the avoidance behavior of the household robot. Through mathematical formulas, the moving distance of the household robot is calculated to ensure that it can maintain a safe distance and avoid collision with objects when avoiding bouncing objects. The moving distance is incorporated into the real-time following information to ensure that the household robot makes timely and accurate responses in a dynamic environment. It not only improves the intelligent avoidance ability of the household robot, but also greatly enhances the emergency handling ability of the household robot in complex environments, ensuring the efficiency and safety of the avoidance process.

[0064] Optionally, after determining the secondary avoidance angle of the robot when an obstacle appears behind the robot according to the movement speed, the movement direction, the direction angle, the obstacle volume and the real-time avoidance angle, the method further includes:

[0065] Determining the collision position of the personnel according to the real-time tracking image;

[0066] Determining the severity of the collision of the part of the person according to the reversing speed, the reversing direction and the collision part of the person;

[0067] Determining a collision risk index according to the collision severity, the height of the person, the reversing speed, and the volume of the person;

[0068] Acquiring a danger index level, and determining a danger alarm frequency according to the danger index level and the collision danger index;

[0069] Incorporate the hazard alarm frequency into the hazard alarm strategy.

[0070] Through this solution, the collision site of the person is accurately identified through real-time tracking images, providing key data for subsequent collision risk assessment. After judging the reversing speed, reversing direction and collision site, the severity of the collision is evaluated to deduce the potential harm of the collision. The collision danger index is further calculated by combining the severity of the collision, the height of the person, the reversing speed and the volume of the person. Based on this index, the danger index level is determined, so that the frequency of danger alarms can be flexibly adjusted. A high danger index level corresponds to a higher frequency of alarms, ensuring that the household robot can respond in time and warn of potential dangerous situations. Through this comprehensive strategy, the collision risk is dynamically monitored and evaluated, and the warning strategy is adjusted in real time, effectively enhancing the household robot's ability to respond to sudden collision risks while ensuring the safety of the person being followed.

[0071] Optionally, the collision risk index is determined according to the severity of the collision, the height of the person, the reversing speed and the volume of the person, and is calculated according to the following formula:

[0072] ;

[0073] in, represents the collision risk index, represents the height of the person, Indicates the severity of the collision, represents the volume of the person, Indicates the reverse speed.

[0074] Through this solution, by comprehensively evaluating the severity of the collision, the height of the person, the speed of the person turning back, and the volume of the person, a collision risk index is calculated using a formula, thereby providing a quantitative indicator for the collision risk. In the formula, the height of the person, the speed of the person turning back, the volume of the person, and the severity of the collision interact with each other and jointly affect the calculation result of the collision risk index. Specifically, the formula takes into account the inverse proportion of the impact of the volume of the person on the collision, the square of the speed of the person turning back, and the exponential growth of the severity of the collision, ensuring a high-precision assessment of potential dangers. In particular The calculation results help the household robot to grasp the danger level of the collision in real time and provide data support for subsequent early warning and risk avoidance strategies. This enables the household robot to accurately predict potential safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0076] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0077] Figure 2 A flowchart of a visual analysis method for a household robot provided in one embodiment of the present application. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0079] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0080] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0081] Since there are many dynamic uncertain factors in the home, the monitoring and following status of the household robot is easily affected by the dynamic factors in the home, thereby reducing the monitoring level of the household robot.

[0082] Based on this, the present application provides a visual analysis method for a household robot, which captures real-time following images to accurately identify the real-time behavior information and home environment information of family members, and obtain detailed scene cognition. The household robot's own following information is combined with the real-time behavior information and home environment information for comprehensive analysis to identify any abnormal following behavior, such as misjudgment of family members' movements or sudden complex situations. Once an abnormality is detected, the following information is immediately adjusted intelligently based on the abnormal following information and the real-time following image, and new real-time following information is generated and executed. This process ensures that the household robot can flexibly adapt to changes in the home environment and family members' behavior, achieve stable and accurate following of family members, and enhance the intelligence and convenience of family life.

[0083] Figure 1 A schematic diagram of an application scenario provided for the present application. When a household robot follows a family member, the method provided by the present application is applied to obtain and analyze real-time following images and following information, determine the real-time following information, and the household robot follows the family member based on the real-time following information.

[0084] Specifically, the method provided in the present application is applied to the built-in control chip of any household robot, and any household robot may include a built-in control chip, a robot management system, and a video monitoring device. The built-in control chip interacts with the robot management system and the video monitoring device, and the robot management system may be a system for managing and controlling the movement and monitoring of the household robot. The built-in control chip determines the real-time following information by acquiring and analyzing the real-time following image of the video monitoring device and the following information of the robot management system, and the built-in control chip performs following control on the household robot according to the real-time following information. By acquiring the following image in real time, the behavior information and environmental conditions of family members are accurately identified and analyzed to determine the real-time following information. By comparing the real-time behavior information with the home environment, abnormal following situations are detected, thereby optimizing the following path and behavior. The adjusted following information ensures that the household robot adapts to the activities of family members more intelligently, improves the home automation experience, and enhances the safety of family members.

[0085] For specific implementation methods, please refer to the following embodiments.

[0086] Figure 2 This is a flow chart of a visual analysis method for a household robot provided in one embodiment of the present application. The method of this embodiment can be applied to the built-in control chip in the above scenario. Figure 2 As shown, the method includes:

[0087] S201, acquiring a real-time following image, and determining real-time behavior information and home environment information according to the real-time following image.

[0088] The real-time following image may be an image obtained when the household robot uses a video surveillance device to monitor the person being followed while following the person being followed.

[0089] The real-time behavior information may be information corresponding to the behavior of the person being followed when following the person being followed. The real-time behavior information may include walking speed and walking direction.

[0090] The home environment information may be information corresponding to the environment constituted by various areas and objects in the home where the person being followed lives. The home environment information may include the smoothness of the home floor.

[0091] Specifically, real-time follow-up images are obtained from the video surveillance equipment, and the target detection algorithm is used to analyze the real-time follow-up images to obtain images of the person being followed and images of the home environment. The target detection algorithm is used to analyze the image of the person being followed to obtain the walking speed and walking direction of the person being followed. The walking speed and walking direction are incorporated into the real-time behavior information. The image of the home is analyzed using an image segmentation algorithm to obtain the smoothness of the home floor. The smoothness of the home floor is incorporated into the home environment information.

[0092] S202: Acquire following information, and determine abnormal following information based on the real-time behavior information, following information, and home environment information.

[0093] The following information may be basic information corresponding to the following behavior of the household robot itself when the household robot follows the person being followed. The following information may include the following speed and the following direction.

[0094] The abnormal following information may be information corresponding to the normal following state that affects the household robot.

[0095] Specifically, the following information is obtained from the robot management system. The following speed and following distance of the household robot when following are extracted from the following information. It is determined whether the following direction and the walking direction are opposite. If they are opposite, a mathematical analysis method is used to analyze the smoothness of the household floor, the following speed, the following direction and the walking direction obtained in the above steps to determine the braking distance of the household robot when the person being followed walks towards the household robot. The braking distance is compared with the following distance. If the braking distance is greater than the following distance, it is determined that the household robot is about to collide with the person being followed. At this time, the braking distance is included in the abnormal following information.

[0096] S203: adjusting the following information according to the abnormal following information and the real-time following image, determining the real-time following information, and following the family members according to the real-time following information.

[0097] The real-time following information may be information consisting of device parameters determined to enable the following behavior of the household robot to meet the actual following needs to the greatest extent possible when the household robot follows the person being followed.

[0098] Specifically, an image analysis method is used to analyze the real-time following image to determine the relative distance between the person being followed and the household robot when they are walking towards each other, and the relative distance is divided by the following speed to determine the collision time between the household robot and the person being followed, and a mathematical analysis method is used to analyze the collision time, relative distance, braking distance, following time and following direction obtained in the above steps to determine the turning avoidance angle of the household robot when the person being followed walks towards the household robot. The following angle is extracted from the following information, and the following angle is adjusted according to the turning avoidance angle, and the turning avoidance angle obtained in the above steps is included in the real-time following information, and the household robot follows the family member according to the real-time following information.

[0099] Through this solution, by capturing real-time following images, the real-time behavior information of family members and home environment information can be accurately identified. The following information of the household robot itself is combined with the real-time behavior information and home environment information for comprehensive analysis to identify any abnormal following behavior, such as misjudgment of family members' movements or sudden complex situations. Once an abnormality is detected, the following information is immediately adjusted intelligently based on the abnormal following information and real-time following images, and new real-time following information is generated and executed. This process ensures that the household robot can flexibly adapt to changes in the home environment and family members' behavior, achieve stable and accurate following of family members, and enhance the intelligence and convenience of family life.

[0100] In some embodiments, the real-time following images are analyzed to determine the motion characteristics of the key parts of the person being followed; based on the motion characteristics, the walking speed and walking direction of the person being followed are determined; the walking speed and walking direction are incorporated into the real-time behavior information; based on the real-time following images, the home space structure and the types of home items are determined; based on the home space structure and the types of home items, the width of the following road in the home is determined; and the following road width is incorporated into the home environment information.

[0101] The critical part may be a part of the body that is vulnerable to injury when the person being followed is walking or exercising.

[0102] The motion feature may be a feature of a key part of the person being followed when the key part is moving.

[0103] The home space structure can be the structure corresponding to the space formed by objects and areas in the home environment.

[0104] The household item type may be a type corresponding to an object in a household environment.

[0105] The following road width may be the width of the road on which the household robot is located when the household robot follows the person being followed.

[0106] Specifically, the feature recognition algorithm is used to analyze the real-time following image to determine the motion characteristics of the key parts of the person being followed. The computer vision technology is used to analyze the motion characteristics to determine the walking speed and walking direction of the person being followed, and the walking speed and walking direction are incorporated into the real-time behavior information. The target detection algorithm is used to analyze the real-time following image to determine the home space structure and the type of home items. The computer vision and image processing technology are used to analyze the home space structure and the type of home items, determine the width of the following road, and incorporate the width of the following road into the home environment information.

[0107] Through this solution, the key parts of the person being followed are accurately identified and tracked through the analysis of real-time following images, and then their motion characteristics, such as gait, movement amplitude, etc., are extracted, and the walking speed and walking direction of the person being followed are further calculated. These motion data are integrated into real-time behavior information to reflect the dynamic situation of the person being followed in real time. In addition, by processing real-time following images, the home space structure and home item types are analyzed to provide support for the accurate modeling of environmental characteristics. On this basis, the width of the following road in the home is calculated to ensure that it conforms to the actual space layout. All of this information is incorporated into the home environment information, and a quick response is made according to the home environment and the behavioral changes of the person being followed, which helps to improve the intelligent following ability of the home robot in the home space.

[0108] In some embodiments, whether the person being followed falls backward is determined based on the walking speed, walking stride and walking direction; if falling backward, the falling direction, falling speed and height of the person being followed are determined based on the real-time following image; the falling distance of the person being followed is predicted based on the falling direction, falling speed and height of the person; the falling trajectory of the person is determined based on the following road width, falling direction and falling speed; the following trajectory is determined based on the following speed and following direction; based on the falling distance, it is determined whether the following trajectory and the falling trajectory of the person intersect; if an intersection is generated, the intersection position is determined based on the following trajectory and the falling trajectory of the person, and the intersection position is determined as the collision position; the collision position is included in the abnormal following information.

[0109] The falling direction may be the direction in which the followed person falls backwards.

[0110] The falling speed may be the speed at which the person being followed falls backward.

[0111] The person height may be the normal body height of the person being followed.

[0112] The backward falling distance may be the distance in the horizontal direction between the feet of the person being followed and the head of the person being followed after the person being followed falls backward.

[0113] The person falling trajectory may be a trajectory corresponding to the body movement of the followed person when falling backwards.

[0114] The following travel trajectory may be a trajectory formed by the travel route when the household robot follows the person being followed.

[0115] The intersection position may be a position corresponding to the intersection generated between the backward trajectory of the person and the following travel trajectory.

[0116] The collision position may be the position where the household robot collides with the person being followed.

[0117] Specifically, when the household robot follows the person being followed, if the walking speed of the person being followed is zero and the walking stride is also zero, and the walking direction of the person being followed is backward at this time, it is determined that the person being followed falls backward. The real-time following image is analyzed using the target detection algorithm to determine the falling direction, falling speed and height of the person being followed. The falling direction, falling speed and height of the person are analyzed using a mathematical analysis method to determine the falling distance. The width of the following road, the falling direction and the falling speed are analyzed using computer vision and deep learning technology to determine the falling trajectory of the person. The following speed, following direction and following distance are extracted from the following information. The following speed and following direction are analyzed using computer vision and deep learning technology to determine the following trajectory. Based on the falling distance, it is determined whether the following trajectory and the falling trajectory of the person have an intersection. If an intersection is generated, the generated intersection is determined as the intersection position, the intersection position is determined as the collision position, and the collision position is included in the abnormal following information.

[0118] Through this solution, by accurately analyzing the walking speed, walking stride and walking direction of the person being followed, it is possible to determine in real time whether there is a risk of falling over. When it is detected that the person being followed has signs of falling over, the falling direction, falling speed and height of the person are further determined based on the real-time following image, thereby predicting the falling distance. Based on these data, the person's falling trajectory is calculated, and spatial calculations are performed in combination with the width of the following road to predict the person's possible falling path. At the same time, based on the following speed and direction, the following trajectory is inferred, and it is determined whether the trajectory intersects with the person's falling trajectory. If there is an intersection, the collision position is located and marked as the collision position, thereby generating abnormal following information. The implementation of this process effectively prevents and handles the risk of collision that may be caused by the falling over of the person being followed, improves the safety and responsiveness of the household robot, and ensures the safety of people in the home environment.

[0119] In some embodiments, the volume and height of the person being followed are determined based on the real-time following image; the volume of the falling space of the person being followed when falling backwards is determined based on the collision position, the volume of the person, the height of the person and the falling trajectory of the person; the real-time avoidance angle is determined based on the falling space volume, following speed, the volume of the person, the following distance and the height of the person; and the real-time avoidance angle is incorporated into the real-time following information.

[0120] The falling space volume may be the volume occupied by the body of the person being followed in the space when the person being followed falls backwards.

[0121] The real-time avoidance angle may be an angle formed by the following direction of the household robot and determined when avoiding a collision between the household robot and the person being followed.

[0122] Specifically, the real-time following image is analyzed using an image analysis method to determine the volume and height of the person being followed. The collision position, the volume, the height and the falling trajectory of the person are analyzed using a mathematical analysis method to determine the falling space volume of the person being followed when falling backward. The falling space volume, the following speed, the volume, the following distance and the height of the person are analyzed using a mathematical analysis method to determine the real-time avoidance angle. The real-time avoidance angle is incorporated into the real-time following information.

[0123] Through this solution, the volume and height of the person being followed are accurately measured through real-time following images, providing basic data for subsequent collision prevention. After detecting the collision position, the volume of the person being followed when falling backwards is accurately calculated by combining the person's volume, height and falling trajectory. This data provides an important basis for judging the space required for a person to fall backwards. According to the falling space volume, following speed, person volume, following distance and person height, the real-time avoidance angle is calculated to ensure that the household robot responds in time during the person's falling process to avoid collisions. The real-time avoidance angle is incorporated into the real-time following information, enhancing the intelligent response capability of the household robot. The dynamic response speed of the household robot is greatly improved, and the probability of accidents is reduced while adjusting the following behavior in real time.

[0124] In some embodiments, the tilt angle formed between the body of the followed person and the wall when the followed person falls backwards is determined based on the real-time following image; the real-time avoidance angle is determined based on the volume of the falling space, the following speed, the tilt angle, the volume of the person, the following distance and the height of the person according to formula (1):

[0125] (1)

[0126] in, Indicates the real-time avoidance angle. Indicates the volume of personnel, Indicates the following distance, Indicates the height of the person. Indicates the following speed, Indicates the tilt angle.

[0127] The tilt angle may be a value corresponding to an angle formed between the body of the person being followed and the wall of the home environment in which the person being followed is located when the person being followed falls backwards.

[0128] The following speed may be the speed at which the household robot follows the person being followed.

[0129] The following distance may be the distance between the household robot and the person being followed when the household robot follows the person being followed.

[0130] Specifically, an image analysis method is used to analyze the real-time following image to determine the inclination angle between the body of the person being followed and the wall when the person is falling backwards. A mathematical analysis method is used to analyze the falling space volume, following speed, inclination angle, person volume, following distance and person height, and the real-time avoidance angle is determined according to formula (1).

[0131] In formula (1), It reflects the relationship between the size of the space occupied by the person being followed and the following distance and speed of the household robot and the person being followed. The larger the space the person being followed has to fall back to, or the closer the distance between the household robot and the person being followed is, or the taller the person being followed is and the faster the household robot is, the more space and angle the household robot needs to avoid. The avoidance angle of the household robot is further adjusted by the tilt angle when the person being followed falls down. The tilt angle at which the person being followed falls Converting to a relative scale ensures that changes in tilt angle are appropriately quantified. Chinese: The purpose of adding 1 is to ensure that even if When it is 0 (i.e. the person being followed is completely upright and has not fallen down), the real-time avoidance angle will not be completely 0. It is ensured that the angle of the space occupied by the fallen person has a reasonable effect on the avoidance response of the household robot. When the person falls more tilted (the angle is larger), the angle that the household robot needs to avoid is also increased accordingly to avoid collision.

[0132] Through this solution, the tilt angle formed between the body and the wall of the person being followed when falling backwards is accurately determined through real-time following image analysis, which is used as the key parameter for subsequent avoidance calculation. On this basis, the real-time avoidance angle is calculated using a formula in combination with the volume of the falling space, the following speed, the volume of the person, the following distance and the height of the person. The various variables in the formula, such as the volume of the person, the following distance, the height of the person, the following speed and the tilt angle, jointly determine the avoidance angle. Through this calculation, the following angle is adjusted in time during the process of the person falling backwards, ensuring that the household robot can flexibly avoid according to the real-time situation and avoid potential collision risks. The introduction of the tilt angle in the formula further enhances the sensitivity of the household robot to the falling posture, making the calculation of the avoidance angle more accurate and better able to cope with complex spatial environments. Through the analysis of real-time data through precise mathematical formulas, efficient dynamic avoidance is achieved, which significantly improves the safety and response speed of the household robot in the home environment.

[0133] In some embodiments, a rear monitoring image is obtained, and it is determined whether the rear obstacle is moving according to the rear monitoring image; if it is moving, the movement speed, movement direction and volume of the obstacle are determined according to the rear monitoring image; the direction angle is determined according to the movement direction and the real-time avoidance angle; and the secondary avoidance angle of the robot when an obstacle appears behind the robot is determined according to formula (2) based on the movement speed, movement direction, direction angle, obstacle volume and real-time avoidance angle:

[0134] (2)

[0135] in, represents the secondary avoidance angle, Indicates the real-time avoidance angle. represents the direction angle, represents the obstacle volume, Indicates the speed of movement.

[0136] The rear monitoring image may be an image obtained by monitoring the rear of the household robot when the household robot follows the person being followed.

[0137] The movement speed may be the speed at which the obstacle behind the household robot moves.

[0138] The movement direction may be the direction in which the obstacle behind the household robot corresponds when it moves.

[0139] The direction angle may be the angle formed between the movement direction of the rear obstacle and the real-time avoidance angle of the household robot.

[0140] The secondary avoidance angle may be the travel angle determined when the household robot avoids the person being followed based on the real-time avoidance angle and when an obstacle appears behind the household robot and needs to avoid the obstacle.

[0141] The rear obstacle may be an object that appears behind the household robot and creates an obstacle to the normal avoidance of the household robot when the household robot is avoiding the followed person according to the real-time avoidance angle.

[0142] Specifically, when the household robot avoids a fallen person being followed, obstacles appearing behind the household robot will also affect the safe avoidance of the household robot, thereby affecting the normal working state of the household robot.

[0143] Therefore, the rear monitoring image is obtained from the video surveillance equipment, and the target detection algorithm is used to analyze the rear monitoring image to determine whether the rear obstacle is moving. If it is moving, the feature recognition algorithm is used to analyze the rear monitoring image to determine the movement speed, movement direction and obstacle volume of the obstacle. A spatial rectangular coordinate system is established with the household robot as the coordinate origin, and the movement direction and real-time avoidance angle are marked in the spatial rectangular coordinate system to obtain the movement direction vector and the real-time avoidance angle vector. The movement direction vector and the real-time avoidance angle vector are analyzed using mathematical analysis methods to determine the direction angle. The movement speed, movement direction, direction angle, obstacle volume and real-time avoidance angle are analyzed using mathematical analysis methods, and the secondary avoidance angle of the robot when an obstacle appears behind is determined according to formula (2).

[0144] In formula (2), It is the direction determined by the household robot based on the current situation, reflecting the initial posture of the household robot to avoid collision with the person being followed. , normalize the direction angle value to a range of [0,1], so that the influence of the direction angle on the secondary avoidance angle becomes smoother and more reasonable. Indicates the "movement intensity" of the obstacle - the larger the obstacle is, or the faster it moves, the greater the angle the household robot needs to avoid. Take the inverse tangent , a relatively smooth avoidance angle adjustment method is obtained, which reflects the potential threat of obstacles to the household robot. This makes the adjustment of the secondary avoidance angle more in line with actual needs and avoids changes that are too drastic or too smooth.

[0145] Through this solution, by acquiring the rear monitoring image, it is possible to monitor in real time whether there are obstacles in the rear and determine whether they are moving. If the obstacle is found to be moving, the movement speed, direction and volume of the obstacle are determined according to the monitoring image, thereby providing necessary information for adjusting the avoidance path. By calculating the directional angle between the movement direction of the obstacle and the real-time avoidance angle, the avoidance requirements can be accurately understood. Furthermore, the secondary avoidance angle is calculated using a formula, which takes into account factors such as the real-time avoidance angle, directional angle, obstacle volume, and obstacle movement speed. This calculation can help the household robot flexibly adjust the avoidance trajectory according to the real-time status of the rear obstacle, and effectively deal with the risks brought by dynamic obstacles. Through precise mathematical models and dynamic adjustment mechanisms, efficient avoidance in complex environments is achieved, which improves the intelligent response ability and safety of the household robot in the real-time following process.

[0146] In some embodiments, a determination is made based on real-time following images whether a dropped object bounces; if a bounce occurs, the volume of the object, the bounce height and the bounce direction of the object after it falls are determined based on the real-time following images; based on the bounce height and the bounce direction, a determination is made as to whether the walking speed and the walking direction affect the bounce of the object; if so, the degree of walking influence of the person's behavior on the bounce of the object is determined based on the walking speed, the walking direction and the volume of the object; the bounce trajectory of the object is determined based on the walking influence degree, the bounce height and the bounce direction; the trajectory contact position and the trajectory contact height are determined based on the bounce trajectory and the following travel trajectory; the trajectory contact position and the trajectory contact height are included in the abnormal following information.

[0147] The walking influence degree may be a value corresponding to the degree of influence of the behavior of the followed person on the movement characteristics of the object when it bounces.

[0148] The bounce trajectory may be the movement trajectory of the object when the bounce occurs.

[0149] Bounce height can be the vertical distance between an item and the ground when it bounces.

[0150] The bounce direction may be the direction in which the item corresponds when the bounce occurs.

[0151] The track contact position may be the position of the intersection point when the bouncing track and the following traveling track intersect.

[0152] The track contact height may be the height of the intersection point when the bouncing track and the following traveling track intersect.

[0153] Specifically, the target detection algorithm is used to analyze the real-time following image to determine whether the dropped object bounces. If it bounces, the feature recognition method is used to analyze the real-time following image to determine the object volume, bounce height and bounce direction. The walking speed, walking direction, person volume and person height are analyzed using a mathematical analysis method to determine the disturbance intensity of the air flow caused by the behavior of the followed person. The bounce height, bounce direction and object volume are analyzed using a mathematical analysis method to determine the object movement intensity. The disturbance intensity is compared with the movement intensity to determine whether the disturbance intensity is greater than the movement intensity. If it is greater, it is determined that the walking speed and walking direction affect the bounce of the object. The walking speed, walking direction and object volume are analyzed using a mathematical analysis method to determine the degree of walking influence. Based on the degree of walking influence, the bounce height and bounce direction are analyzed using computer vision analysis technology to determine the bounce trajectory. Mark the bouncing trajectory and the following trajectory in the spatial rectangular coordinate system determined in the above steps, and determine whether the bouncing trajectory and the following trajectory have an intersection. If an intersection is generated, the position of the intersection between the bouncing trajectory and the following trajectory is determined as the trajectory contact position, and the height of the intersection between the bouncing trajectory and the following trajectory is determined as the trajectory contact height. The trajectory contact position and trajectory contact height are included in the abnormal following information.

[0154] Through this solution, through real-time following image analysis, it is determined whether the dropped object bounces. If the object bounces, the volume, bounce height and bounce direction of the object are further analyzed to accurately predict the movement trajectory of the object. Combined with the walking speed and walking direction, it is determined whether the behavior of the person being followed has an impact on the bounce of the object. If the behavior of the person being followed has an impact on the bounce of the object, the degree of impact of the person's behavior on the bounce of the object will be evaluated based on the walking speed, walking direction and object volume. Through this process, the bounce trajectory of the object is accurately calculated and compared with the follow-up travel trajectory to determine the trajectory contact position and contact height. All relevant trajectory contact information is included in the abnormal following information, providing a basis for further processing and emergency response. Effectively predict and manage the safety hazards that may be caused by the bounce of objects, and enhance the adaptability of household robots in dynamic environments.

[0155] In some embodiments, the relative distance between the robot and the bouncing object is determined based on the real-time following image; the moving distance of the robot when avoiding the bouncing object is determined according to the following speed, track contact height, relative distance, object volume and bouncing direction according to formula (3); and the moving distance is included in the real-time following information:

[0156] (3)

[0157] in, Indicates the moving distance, Indicates the relative distance, Indicates the following speed, Indicates the volume of the item. represents the track contact height, Indicates the bounce direction.

[0158] The relative distance may be the distance between the household robot and the object that bounces when the object bounces.

[0159] The moving distance may be the distance determined when the household robot moves to avoid the bouncing object when the bouncing object affects the normal following state of the household robot.

[0160] Specifically, an image analysis algorithm is used to analyze the real-time following image to determine the relative distance between the household robot and the bouncing object. A mathematical analysis method is used to analyze the following speed, track contact height, relative distance, object volume, and bouncing direction. According to formula (3), the moving distance of the robot when avoiding the bouncing object is determined and the moving distance is included in the real-time following information.

[0161] In formula (3), It reflects the "response strength" of the household robot to the movement of the object. The square root operation is used to appropriately scale this response strength. The introduction of the square root operation can avoid excessive expansion of the calculation results due to the increase of the relative distance and the following speed. The absolute value of is taken as the cube root, taking into account the nonlinear effect of the object volume on the moving distance. By taking the fourth root of the absolute value of the track contact height, the impact on the household robot when the track contact height is too high is avoided, indicating that the impact of the track contact height on the avoidance action is limited, and the moving distance will not increase dramatically with the increase of the track contact height, but there is a smooth rising process. The introduction of the sine function indicates that the relationship between the moving distance of the household robot and the bouncing direction of the object is nonlinear, and there is a certain angular dependence between the bouncing direction of the object and the avoidance direction.

[0162] It takes into account multiple nonlinear factors and calculates an "avoidance strength" based on relative distance, following speed, object volume and track contact height. Through the square root operation, these factors are adjusted comprehensively without any single factor having an excessive impact on the moving distance. A smooth and constrained result is given to determine the reasonable moving distance of the household robot when avoiding bouncing objects.

[0163] Through this scheme, the relative distance between the household robot and the bouncing object is determined by real-time following images, which provides important spatial information for the household robot in the avoidance process. On this basis, the moving distance of the household robot when avoiding the bouncing object is calculated by combining the following speed, track contact height, relative distance, object volume and bouncing direction. The various variables in the formula jointly affect the avoidance behavior of the household robot. Through mathematical formulas, the moving distance of the household robot is calculated to ensure that it can maintain a safe distance and avoid collision with objects when avoiding bouncing objects. The moving distance is incorporated into the real-time following information to ensure that the household robot makes timely and accurate responses in a dynamic environment. It not only improves the intelligent avoidance ability of the household robot, but also greatly enhances the emergency handling ability of the household robot in complex environments, ensuring the efficiency and safety of the avoidance process.

[0164] In some embodiments, the collision site of the person is determined based on the real-time tracking image; the collision severity of the collision site of the person is determined based on the reversing speed, the reversing direction and the collision site of the person; the collision risk index is determined based on the collision severity, the height of the person, the reversing speed and the volume of the person; the risk index level is obtained, and the risk alarm frequency is determined based on the risk index level and the collision risk index; the risk alarm frequency is incorporated into the risk alarm strategy.

[0165] The person collision part may be a part where the body of the person being followed collides when the person being followed falls backward.

[0166] The severity of the collision may be a numerical value corresponding to the strength of the collision at the collision site of the person.

[0167] The collision risk index may be a numerical value corresponding to the degree of danger when a collision occurs at a collision site of a person.

[0168] The danger index level may be a level of danger of the followed person determined according to the collision risk index.

[0169] The danger alarm frequency may be the frequency at which the household robot issues an alarm when the person being followed encounters danger.

[0170] The danger alarm strategy may be a strategy determined by the household robot when making a danger alarm.

[0171] Specifically, an image analysis algorithm is used to analyze the real-time tracking image to determine the personnel collision site and the collision area of ​​the personnel collision site. An image analysis method is used to analyze the collision area, the reversing speed and the reversing direction to determine the severity of the collision. A mathematical analysis method is used to analyze the severity of the collision, the height of the personnel, the reversing speed and the volume of the personnel to determine the collision risk index. The risk index level is obtained from the robot management system, and the collision risk index is matched with the risk index level to obtain the risk index level corresponding to the collision risk index and the risk alarm frequency corresponding to the risk index level, and the risk alarm frequency is incorporated into the risk alarm strategy.

[0172] Through this solution, the collision site of the person is accurately identified through real-time tracking images, providing key data for subsequent collision risk assessment. After determining the reversing speed, reversing direction and collision site, the severity of the collision is evaluated to infer the potential hazards of the collision. The collision danger index is further calculated based on the severity of the collision, the height of the person, the reversing speed and the volume of the person. Based on this index, the danger index level is determined, so that the frequency of danger alarms can be flexibly adjusted. A high danger index level corresponds to a higher frequency of alarms, ensuring that the household robot can respond to potential dangerous situations in a timely manner. Through this comprehensive strategy, the collision risk is dynamically monitored and evaluated, and the early warning strategy is adjusted in real time, effectively enhancing the household robot's ability to respond to sudden collision risks while ensuring the safety of the person being followed.

[0173] In some embodiments, the collision risk index is determined according to the collision severity, the height of the person, the reversing speed and the volume of the person, according to formula (4):

[0174] (4)

[0175] in, Represents the collision risk index, Indicates the height of the person. Indicates the severity of the collision. Indicates the volume of personnel, Indicates reverse speed.

[0176] Specifically, a mathematical analysis method is used to analyze the severity of the collision, the height of the person, the speed of the reverse movement, and the volume of the person, and the collision risk index is determined according to formula (4).

[0177] In formula (4), The kinetic energy, body size, and severity of the collision of the person being followed are combined to quantify the total impact force and energy distribution after the fall. , the severity of the collision is further amplified, emphasizing the nonlinear increase of the danger index in severe collision situations. is a progressive growth function used to modify the final impact of collision severity on the hazard index. As the impact of the collision increases, this part of the impact will gradually increase, but the growth rate is slow, reflecting the gradual effect of the impact of the collision severity on the collision risk index. The existence of limits the excessive amplification effect of the collision severity on the collision risk index, even if If the value is very large, the growth of the collision risk index will not be explosive, but will increase gradually. It reflects the impact of collision severity on the collision risk index. As the collision severity increases, the growth of the collision risk index will accelerate, but it will not be amplified endlessly. The logarithmic function limits this growth and avoids excessive exaggeration of the calculation results due to extreme collision situations.

[0178] Through this solution, by comprehensively evaluating the severity of the collision, the height of the person, the speed of the person turning back, and the volume of the person, a collision risk index is calculated using a formula, thereby providing a quantitative indicator for the collision risk. In the formula, the height of the person, the speed of the person turning back, the volume of the person, and the severity of the collision interact with each other and jointly affect the calculation result of the collision risk index. Specifically, the formula takes into account the inverse proportion of the impact of the volume of the person on the collision, the square of the speed of the person turning back, and the exponential growth of the severity of the collision, ensuring a high-precision assessment of potential dangers. In particular The calculation results help the household robot to grasp the danger level of the collision in real time and provide data support for subsequent early warning and risk avoidance strategies. This enables the household robot to accurately predict potential safety risks.

Claims

1. A visual analysis method for a household robot, characterized in that: include: Acquire a real-time following image, and determine real-time behavior information and home environment information according to the real-time following image; Acquire follow-up information, and determine abnormal follow-up information according to the real-time behavior information, the follow-up information and the home environment information; According to the abnormal following information and the real-time following image, the following information is adjusted to determine the real-time following information, and the family members are followed according to the real-time following information; Determining the real-time behavior information and home environment information according to the real-time following image includes: Analyzing the real-time tracking image to determine the movement characteristics of key parts of the person being followed; Determining the walking speed and walking direction of the person being followed according to the movement characteristics; Incorporating the walking speed and the walking direction into the real-time behavior information; Determine the home space structure and home item types based on the real-time tracking image; Determining a following road width in the home according to the home space structure and the types of home items; Incorporating the following road width into the home environment information; The real-time behavior information includes walking stride, the following information includes following speed, following direction and following distance, and determining abnormal following information according to the real-time behavior information, the following information and the home environment information includes: determining whether the person being followed falls backwards according to the walking speed, the walking stride and the walking direction; If the person falls backwards, the falling direction, falling speed and height of the person being followed are determined based on the real-time following image; Predicting a backward backward distance of the person being followed according to the backward direction, the backward speed and the height of the person; Determining a reverse trajectory of the person according to the following road width, the reverse direction, and the reverse speed; Determining a following trajectory according to the following speed and the following direction; According to the backward distance, determining whether the following trajectory and the backward trajectory of the person have an intersection; If an intersection is generated, the intersection position is determined according to the following moving trajectory and the backward trajectory of the person, and the intersection position is determined as the collision position; The collision position is incorporated into the abnormal follow-up information.

2. The method according to claim 1, characterized in that The step of adjusting the following information according to the abnormal following information and the real-time following image to determine the real-time following information includes: Determining the volume and height of the person being followed according to the real-time following image; Determine the volume of the space where the followed person falls backwards according to the collision position, the person's volume, the person's height and the person's falling trajectory; Determine a real-time avoidance angle according to the volume of the backward space, the following speed, the volume of the person, the following distance and the height of the person; The real-time avoidance angle is incorporated into the real-time following information.

3. The method according to claim 2, characterized in that The determining of the real-time avoidance angle according to the volume of the backward space, the following speed, the volume of the person, the following distance and the height of the person includes: determining, according to the real-time following image, an inclination angle formed between the body of the followed person and the wall when the followed person falls backwards; The real-time avoidance angle is determined according to the volume of the backward space, the following speed, the inclination angle, the volume of the person, the following distance and the height of the person, and is calculated according to the following formula: ； in, represents the real-time avoidance angle, represents the volume of the person, represents the following distance, represents the height of the person, represents the following speed, represents the tilt angle.

4. The method according to claim 3, characterized in that After determining the real-time avoidance angle according to the volume of the backward space, the following speed, the volume of the person, the following distance and the height of the person, the method further includes: Acquire a rear monitoring image, and determine whether a rear obstacle moves according to the rear monitoring image; If movement occurs, determining the movement speed, movement direction and volume of the obstacle based on the rear monitoring image; Determining a direction angle according to the moving direction and the real-time avoidance angle; According to the movement speed, the movement direction, the direction angle, the obstacle volume and the real-time avoidance angle, the secondary avoidance angle of the robot when an obstacle appears behind is determined and calculated according to the following formula: ； in, represents the secondary avoidance angle, represents the real-time avoidance angle, represents the direction angle, represents the obstacle volume, Indicates the speed of the movement.

5. The method according to claim 1, characterized in that The determining abnormal following information according to the real-time behavior information, the following information and the home environment information includes: Determining whether the dropped object bounces according to the real-time tracking image; If a bounce occurs, the volume of the object, the bounce height and the bounce direction of the object after it falls are determined based on the real-time tracking image; According to the bouncing height and the bouncing direction, determining whether the walking speed and the walking direction affect the bouncing of the object; If so, determining the degree of influence of the person's behavior on the bouncing of the object according to the walking speed, the walking direction and the volume of the object; Determining a bouncing trajectory of the object according to the walking impact degree, the bouncing height, and the bouncing direction; Determining a track contact position and a track contact height according to the bouncing track and the following traveling track; The track contact position and the track contact height are incorporated into the abnormal following information.

6. The method according to claim 5, characterized in that The step of adjusting the following information according to the abnormal following information and the real-time following image to determine the real-time following information includes: Determining the relative distance between the robot and the bouncing object according to the real-time following image; According to the following speed, the track contact height, the relative distance, the object volume and the bouncing direction, the moving distance of the robot when avoiding the bouncing object is determined and calculated according to the following formula: ； in, represents the moving distance, represents the relative distance, represents the following speed, Indicates the volume of the item, represents the track contact height, Indicates the bounce direction; The moving distance is incorporated into the real-time following information.

7. The method according to claim 4, characterized in that After determining the secondary avoidance angle of the robot when an obstacle appears behind the robot according to the movement speed, the movement direction, the direction angle, the obstacle volume and the real-time avoidance angle, the method further includes: Determining the collision position of the personnel according to the real-time tracking image; Determining the severity of the collision of the part of the person according to the reversing speed, the reversing direction and the collision part of the person; Determining a collision risk index according to the collision severity, the height of the person, the reversing speed, and the volume of the person; Acquiring a danger index level, and determining a danger alarm frequency according to the danger index level and the collision danger index; Incorporate the hazard alarm frequency into the hazard alarm strategy.

8. The method according to claim 7, characterized in that The collision risk index is determined according to the severity of the collision, the height of the person, the reversing speed and the volume of the person, and is calculated according to the following formula: ； in, represents the collision risk index, represents the height of the person, Indicates the severity of the collision, represents the volume of the person, Indicates the reverse speed.

Images