Image processing method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an image processing method, and the method comprises the steps: carrying out the behavior recognition processing based on a real-time chasing image when a person embracing a target child in a preset age group is recognized as a non-associated person in the real-time chasing image, and obtaining the first behavior data of the target child and the second behavior data of the non-associated person; calculating a lost risk value of the target child based on the first behavior data and the second behavior data; based on the lost risk value, determining early warning information corresponding to the target child; and sending the early warning information to the related personnel corresponding to the target child. When the target child is a non-associated person, behavior recognition processing is performed according to the real-time tracking image to obtain the first behavior data and the second behavior data, the missing risk value of the target child can be accurately calculated according to the first behavior data and the second behavior data, and then early warning information can be accurately determined according to the missing risk value; and the early warning information is sent to the related personnel, so that the passive lost early warning is realized.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an image processing method, device, electronic equipment and storage medium. Background Art

[0002] In today's society, children often go missing. Some children go missing due to ignorance and playfulness without intervention from others, while others go missing due to intervention from others. Passive loss is far more harmful than active loss. Traditionally, the investigation and warning of passive loss is often done by manually checking whether there are abnormalities in the images obtained by the Skynet system. When there are abnormalities, the results of the investigation are determined by manually querying the population archives and taking corresponding measures. However, the traditional method of judging abnormalities is subjective and its accuracy is not high. Therefore, how to provide an image processing method that can improve the accuracy of warning of passive loss has become an urgent problem to be solved. Summary of the invention

[0003] The embodiment of the present invention provides an image processing method, which aims to solve the problem of low accuracy of existing passive lost warning. By identifying the real-time tracking image, it is possible to determine whether the person holding the target child is an unrelated person. If the person is an unrelated person, behavior recognition processing is performed based on the real-time tracking image to obtain first behavior data and second behavior data, so that the target child's lost risk value can be accurately calculated based on the first behavior data and the second behavior data, and then the warning information can be accurately determined based on the lost risk value, and the warning information is sent to relevant personnel to achieve passive lost warning.

[0004] In a first aspect, an embodiment of the present invention provides an image processing method, the method comprising the following steps:

[0005] When a person holding a target child of a preset age group is identified as an unrelated person in a real-time tracking image, a behavior recognition process is performed based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person, and the unrelated person has no relative relationship with the target child;

[0006] Calculating a risk value of the target child being lost based on the first behavior data and the second behavior data;

[0007] Based on the risk value of getting lost, determining the warning information corresponding to the target child;

[0008] The warning information is sent to relevant personnel corresponding to the target child.

[0009] Optionally, when it is identified in the real-time tracking image that the person holding the target child of a preset age group is an unrelated person, before performing behavior recognition processing based on the real-time tracking image, the method further includes:

[0010] When a child meeting preset conditions is identified, a real-time tracking image corresponding to the child and a relationship map corresponding to the child are obtained, wherein the preset conditions include a preset age group and being held, and the relationship map includes associated persons who have a kinship relationship with the child;

[0011] In the relationship graph, the associated persons are extracted one by one for matching with the embraced persons, and matching results of the embraced persons are obtained;

[0012] Based on the matching result, determining whether the person holding the child is an unrelated person;

[0013] When the person holding the child is an unrelated person, the child is determined as the target child.

[0014] Optionally, the calculating the risk value of the target child being lost based on the first behavior data and the second behavior data includes:

[0015] Acquire the first frequently visited location information of the target child and the second frequently visited location information of the unrelated person;

[0016] When the child is in a preset occasion, determining a risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information;

[0017] When the child is not in a preset situation, the risk value of the target child being lost is determined based on the first behavior data, the second behavior data, the first frequented location information, and the second frequented location information.

[0018] Optionally, the second behavior data includes a travel direction, and when the child is in a preset occasion, determining the risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information, comprises:

[0019] Obtaining the identity of the non-associated person, where the identity includes a specific person and a non-specific person;

[0020] determining whether the traveling direction is opposite to the direction of the first frequently visited location information;

[0021] determining whether the traveling direction is the same as the direction of the second frequently visited location information;

[0022] When the child is in a preset situation, the non-associated person is a specific person, and the moving direction is opposite to the direction of the first frequently visited location information or is the same as the second frequently visited location information, the lost risk value is a preset first risk value;

[0023] When the child is in a preset occasion, the identity of the non-associated person is a non-specific person, and the traveling direction is opposite to the first frequented location information or the same as the second frequented location information, the lost risk value is a preset second risk value, and the first risk value is higher than the second risk value.

[0024] Optionally, the first behavior data includes a first behavior action and a degree of facial occlusion of the target child, the second behavior data includes a second behavior action and a moving speed of the unrelated person, the real-time tracking image corresponds to a collection location, and determining the target child's lost risk value based on the first behavior data, the second behavior data, the first frequently visited location information, and the second frequently visited location information includes:

[0025] Acquire the identity and second behavior habit information of the non-associated person, and the first behavior habit information corresponding to the target child;

[0026] Based on the identity of the person, determining an identity risk value corresponding to the target child;

[0027] Determining a distance risk value corresponding to the target child based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information;

[0028] Determining a speed risk value corresponding to the target child based on a ratio of the traveling speed to a preset basic speed;

[0029] Determining an occlusion risk value corresponding to the target child based on the degree of occlusion of the face;

[0030] Determining a habit risk value corresponding to the target child based on the first behavior habit information, the first behavior action, and the second behavior habit information, the second behavior action;

[0031] The identity risk value, the distance risk value, the speed risk value, the occlusion risk value and the habit risk value are added together to obtain the lost risk value.

[0032] Optionally, determining the distance risk value corresponding to the target child based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information includes:

[0033] Determining a first distance risk value corresponding to the target child based on a ratio between the first distance and a preset basic distance;

[0034] Determining a second distance risk value corresponding to the target child based on a ratio between the preset basic distance and the second distance;

[0035] When the identity of the non-associated person is a non-specific person, the first distance risk value is used as the distance risk value;

[0036] When the identity of the non-associated person is a specific person, the first distance risk value is added to the second distance risk value to obtain the distance risk value.

[0037] Optionally, determining the occlusion risk value corresponding to the target child based on the degree of occlusion of the face includes:

[0038] Obtaining current weather data corresponding to the target child;

[0039] Based on the current weather data and the degree of facial occlusion, a occlusion risk value corresponding to the target child is determined.

[0040] In a second aspect, an embodiment of the present invention further provides an image processing device, the image processing device comprising:

[0041] an identification module, configured to, when identifying in a real-time tracking image that a person holding a target child of a preset age group is an unrelated person, perform behavior identification processing based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person, wherein the unrelated person has no relative relationship with the target child;

[0042] A calculation module, configured to calculate a risk value of the target child being lost based on the first behavior data and the second behavior data;

[0043] A determination module, configured to determine the warning information corresponding to the target child based on the risk value of being lost;

[0044] The sending module is used to send the warning information to the relevant personnel corresponding to the target child.

[0045] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the image processing method provided in the embodiment of the present invention when executing the computer program.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the image processing method provided in the embodiment of the invention are implemented.

[0047] In an embodiment of the present invention, when it is identified in a real-time tracking image that the person holding a target child of a preset age group is a non-associated person, a behavior recognition process is performed based on the real-time tracking image to obtain the first behavior data of the target child and the second behavior data of the non-associated person, and the non-associated person has no relative relationship with the target child; based on the first behavior data and the second behavior data, the target child's risk of getting lost is calculated; based on the lost risk value, the warning information corresponding to the target child is determined; and the warning information is sent to the relevant person corresponding to the target child. By identifying the real-time tracking image, it is possible to determine whether the person holding the target child is a non-associated person. If the target child is a non-associated person, a behavior recognition process is performed based on the real-time tracking image to obtain the first behavior data and the second behavior data, so that the target child's risk of getting lost can be accurately calculated based on the first behavior data and the second behavior data, and then the warning information can be accurately determined based on the lost risk value, and the warning information is sent to the relevant person to achieve a warning of passive loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 is a flow chart of an image processing method provided by an embodiment of the present invention;

[0050] Figure 2 is a structural schematic diagram of an image processing device provided in an embodiment of the present invention;

[0051] Figure 3 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, Figure 1 is a flowchart of an image processing method provided by an embodiment of the present invention, comprising:

[0054] 101. When it is identified in the real-time tracking image that the person holding the target child of a preset age group is an unrelated person, a behavior recognition process is performed based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person.

[0055] In the embodiment of the present invention, the above-mentioned image processing method can be applied to a lost warning platform, which can be constructed by a server or a server cluster, and the above-mentioned server or server cluster can be any electronic device with functions such as data recognition, data processing, data transmission, and data storage. It should be noted that the above-mentioned lost warning platform can implement a lost warning of a child by executing any one of the image processing methods provided in the embodiment of the present invention.

[0056] The above-mentioned real-time tracking images can be collected by collecting cameras deployed in any public places such as roads, railway stations, schools, etc. The above-mentioned target children can be children in a preset age group and in someone's arms. The above-mentioned preset age group can be an age group that is easy to get lost and cannot walk independently (that is, they need to be held by others when traveling). The above-mentioned preset age group can be obtained based on the analysis of historical records of children getting lost. The above-mentioned non-associated persons have no relatives with the target children. The above-mentioned behavior recognition processing can be implemented by at least one behavior recognition model, and the above-mentioned behavior recognition model can be constructed by any deep learning algorithm.

[0057] For example, when the above-mentioned behavior data includes travel speed and behavior action, a behavior recognition model can be constructed by 3D CNNs algorithm or I3D algorithm, and the above-mentioned real-time tracking image is input into the above-mentioned behavior recognition model to obtain the above-mentioned travel speed and behavior action.

[0058] For another example, when the above-mentioned behavior data includes the degree of facial occlusion, a behavior recognition model can be constructed using the Openpose algorithm or the MaskR-CNN algorithm, and the above-mentioned real-time tracking image can be input into the above-mentioned behavior recognition model for behavior recognition processing to obtain the above-mentioned degree of facial occlusion.

[0059] For another example, when the behavior data includes a direction of travel, a behavior recognition model may be constructed using a Deep SORT algorithm or other optical flow tracking algorithm, and the real-time tracking image may be input into the behavior recognition model for behavior recognition processing to obtain the direction of travel.

[0060] It can be understood that the above-mentioned non-related person is the person holding the above-mentioned target child, and it can be understood that the above-mentioned non-related person is holding the above-mentioned target child, so the above-mentioned non-related person and the above-mentioned target child are in the above-mentioned real-time tracking image at the same time. Therefore, when performing behavior recognition processing, it is only necessary to perform behavior recognition processing on the real-time tracking image to obtain the first behavior data of the target child and the second behavior data of the non-related person.

[0061] 102. Calculate the risk value of the target child being lost based on the first behavior data and the second behavior data.

[0062] In an embodiment of the present invention, the first behavior data may include the behavior of the target child, the second behavior data may include the behavior of an unrelated person, the target child may correspond to a child profile, in which the first behavior habit information of the target child may be stored, the unrelated person may correspond to a personnel profile, in which the second behavior habit information of the unrelated person may be stored. The first behavior habit information may be understood as the behavior habit of the target child when being held. The behavior habit may be identified based on historical tracking images, and the second behavior habit information may be understood as the daily behavior habit of the unrelated person.

[0063] Specifically, the above-mentioned target child's behavior and actions can be matched with the above-mentioned first behavior habit information to obtain the target child's first habit matching degree, and the above-mentioned non-related person's behavior and actions can be matched with the above-mentioned second behavior habit information to obtain the non-related person's second habit matching degree. It should be noted that the higher the habit matching degree, the more normal the above-mentioned target child or non-related person's behavior and actions are, and conversely, the lower the habit matching degree, the more abnormal the above-mentioned target child or non-related person's behavior and actions are. Therefore, the above-mentioned lost risk value can be inversely proportional to the above-mentioned first habit matching degree and the second habit matching degree. When the above-mentioned first habit matching degree and the second habit matching degree are lower, the above-mentioned lost risk value is higher, and when the above-mentioned first habit matching degree and the second habit matching degree are higher, the above-mentioned lost risk value is lower.

[0064] More specifically, the current scene of the target child and the non-associated person can also be determined based on the real-time tracking image, and the scene behavior habit information in the same scene as the current scene can be extracted from the behavior habit information (including the first scene behavior habit information corresponding to the target child and the second scene behavior habit information corresponding to the non-associated person). Based on the first scene behavior habit information and the second scene behavior habit information, the first habit matching degree and the second habit matching degree are determined, thereby determining the risk value of getting lost. It is understandable that in different scenes, the habits of the target child and the non-associated person may change accordingly. For example, in a noisy environment, the target child may have a crying behavior habit due to environmental problems, and in a quiet environment, the target child may have a sleeping behavior habit due to environmental problems. Therefore, the scene behavior habit information is screened out according to the scene, so that a more accurate risk value of getting lost can be determined based on the scene behavior habit information.

[0065] 103. Based on the risk value of getting lost, determine the corresponding warning information for the target child.

[0066] In the embodiment of the present invention, the above-mentioned lost risk value may have an upper limit and a lower limit, for example, it may be 0-100, 0-120, etc. Therefore, different risk value intervals are divided according to the upper limit of the above-mentioned lost risk value, and the division of the above-mentioned risk value intervals can be set according to the historical child lost warning experience, or obtained through a limited number of experiments. Each risk value interval corresponds to a risk level.

[0067] According to the risk value interval of the lost risk value, the target risk level corresponding to the target child is determined. And according to the corresponding relationship between the risk level and the warning information, the warning information corresponding to the target risk level is determined.

[0068] For example, when the risk level is low, the warning information may be in text form, which may specifically be that the target child is at risk of getting lost, please pay attention in time, and provide the target child's current coordinates. When the risk level is medium, the warning information may be in voice form, which may specifically be that the target child is at a high risk of getting lost, please handle it in time, and provide the target child's current coordinates. When the risk level is high, the warning information may be in video form, which may specifically include the real-time tracking image of the target child and the corresponding voice prompt, and the target child's current coordinates are marked in the real-time tracking image.

[0069] 104. Send the warning information to the relevant personnel corresponding to the target child.

[0070] In an embodiment of the present invention, the above-mentioned relevant personnel may include associated personnel and managers of the area where the above-mentioned target child is located. The above-mentioned associated personnel may be persons who have a kinship relationship with the above-mentioned target child, and the above-mentioned kinship relationship may include close relatives and distant relatives.

[0071] Specifically, when the target risk level corresponding to the above-mentioned warning information is a low risk level, it means that the target child at the current moment is at risk of getting lost, but it is only a low risk. Therefore, the above-mentioned warning information can be sent only to persons who are close relatives of the above-mentioned target child. Therefore, at this time, the above-mentioned relevant persons only include related persons who are close relatives of the above-mentioned target child.

[0072] Alternatively, when the target risk level corresponding to the above-mentioned warning information is a medium risk level, it means that the target child at the current moment is at risk of getting lost and is relatively dangerous. Therefore, the above-mentioned warning information can be sent to the above-mentioned related persons (that is, all related persons who are close relatives or distant relatives of the above-mentioned target child). Therefore, at this time, the above-mentioned related persons include the above-mentioned related persons.

[0073] Alternatively, when the target risk level corresponding to the above-mentioned warning information is a high risk level, it means that the target child at the current moment is at risk of getting lost and is very dangerous. Therefore, the above-mentioned warning information can be sent to the above-mentioned related persons and the above-mentioned managers. Therefore, at this time, the above-mentioned related personnel can include the above-mentioned related persons and the above-mentioned managers.

[0074] In an embodiment of the present invention, when it is identified in a real-time tracking image that the person holding a target child of a preset age group is a non-associated person, a behavior recognition process is performed based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the non-associated person, and the non-associated person has no relative relationship with the target child; based on the first behavior data and the second behavior data, the target child's risk of getting lost is calculated; based on the lost risk value, the warning information corresponding to the target child is determined; and the warning information is sent to the relevant person corresponding to the target child. By identifying the real-time tracking image, it is possible to determine whether the person holding the target child is a non-associated person. If the target child is a non-associated person, a behavior recognition process is performed based on the real-time tracking image to obtain the first behavior data and the second behavior data, so that the target child's risk of getting lost can be accurately calculated based on the first behavior data and the second behavior data, and then the warning information can be accurately determined based on the lost risk value, and the warning information is sent to the relevant person to achieve a warning of passive loss.

[0075] It can be understood that in the specific implementation of the present application, it involves real-time tracking images, children's files, personnel files, personnel identities, relationship maps, frequently visited location information, behavioral habit information, weather data and other related data. When the embodiments in the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data and the construction, training and use of related models need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0076] It should be noted that the image processing method provided in the embodiment of the present invention can be applied to devices such as computers and servers that can perform image processing.

[0077] Optionally, when the person holding a target child of a preset age group is identified as a non-associated person in the real-time tracking image, before the step of performing behavior recognition processing based on the real-time tracking image, when a child meeting the preset conditions is identified, the real-time tracking image corresponding to the child and the relationship map corresponding to the child are obtained; in the relationship map, the associated persons and the holding person are extracted one by one for matching to obtain the matching results of the holding person; based on the matching results, it is determined whether the person holding the child is a non-associated person; when the person holding the child is a non-associated person, the child is determined as the target child.

[0078] In an embodiment of the present invention, the preset conditions include a preset age group and being held, the relationship map includes associated persons who have a kinship relationship with the child, and the lost warning platform can store personnel files of personnel and children files of children. The children files can be established according to different regions, that is, each region corresponds to a children's archive. The children's archive includes children files corresponding to all children in the preset age group in the region, and each child file can include a relationship map corresponding to the child. The region can be divided in any way such as cities, urban areas, towns, villages, etc. Similarly, the personnel files can also be established according to different regions, each region corresponds to a personnel archive, and the personnel archive includes personnel files of all personnel in the region, and can also include personnel files of specific personnel, and the specific personnel can be unaffected by the region, that is, the personnel archive of each region includes personnel files of specific personnel corresponding to all regions. The specific personnel can be a person with a record of violating the law.

[0079] Specifically, the facial features, behavioral actions and other data of children in the preset age group in the area can be collected through the image acquisition device deployed in public places in the above-mentioned area, as well as the facial features, hairstyles, heights, walking postures and the like of the companions who have walked with (or embraced) the children of the preset age group for many times. Combined with the information in the population archive and the companions corresponding to the above-mentioned children, the associated persons who have a kinship relationship with the above-mentioned children are determined, and according to the distance of the kinship relationship between the associated persons and the children, a relationship map is constructed. In the above-mentioned relationship map, the closer the associated persons are to the children, the closer their kinship relationship is (that is, the closer they are to close relatives, where the closest close relatives can be guardians, parents, etc.), and conversely, the farther the associated persons are from the children, the farther their kinship relationship is (that is, the closer they are to distant relatives). After obtaining the above-mentioned relationship map, the above-mentioned relationship map and the facial features, behavioral actions corresponding to the children, the facial features, hairstyles, heights, walking postures and other data corresponding to the companions are stored in the child file corresponding to each child. It should be noted that the above-mentioned population file can be obtained by statistics of relevant institutions.

[0080] Specifically, after obtaining the real-time tracking image corresponding to the child and the relationship map corresponding to the child, the associated persons and the embraced persons can be extracted one by one in the relationship map in a random order for matching, and the matching result of the embraced person can be obtained. The matching can be a facial feature matching, and when the face of the embraced person is blocked, a body feature matching can also be performed. When the matching result of the embraced person is that the corresponding associated person is not matched, the embraced person is a non-associated person, and the child is the target child.

[0081] Alternatively, in the above relationship map, related persons may be extracted one by one in the order of the above kinship from near to far, and matched with the above embraced persons to improve the matching efficiency.

[0082] Optionally, in the step of calculating the target child's risk of getting lost based on the first behavior data and the second behavior data, the target child's first frequent location information and the second frequent location information of an unrelated person can also be obtained; when the child is in a preset occasion, the target child's risk of getting lost is determined based on the second behavior data, the first frequent location information and the second frequent location information; when the child is not in the preset occasion, the target child's risk of getting lost is determined based on the first behavior data, the second behavior data, the first frequent location information and the second frequent location information.

[0083] In an embodiment of the present invention, the first frequently visited location information may be stored in the child file of the target child, and the second frequently visited location information may be stored in the personnel file of the non-associated person.

[0084] Specifically, trajectory analysis can be performed on children and their related persons based on a historical tracking image sequence captured by an image acquisition device in public places in the area, a map of places frequently visited by children and their related persons can be drawn, and the map of places frequently visited can be stored in the child's file.

[0085] It is also possible to perform trajectory analysis on all persons in the area based on the historical tracking image sequences acquired by the image acquisition device in public places in the area, draw a map of the places where each person frequently visits, and store the map of the places where each person frequently visits in the personnel file.

[0086] More specifically, the preset occasions may be noisy occasions, or may be occasions that the children have not been exposed to based on historical tracking images, such as vegetable markets, pedestrian streets, shopping malls, etc. When children are exposed to the preset occasions, they may behave abnormally due to the influence of the environment. Therefore, in the above occasions, the first behavior data of the children may not be considered in the calculation basis of the lost risk value, and only the second behavior data, the first frequented location information, and the second frequented location information need to be considered. In non-preset occasions, the first behavior data of the children may be considered.

[0087] Specifically, the first behavior data and the corresponding first behavior habit information can be matched to obtain the first habit matching degree, the second behavior data and the corresponding second behavior habit information can be matched to obtain the second habit matching degree, and the risk of loss to be adjusted can be determined based on the first habit matching degree and the second habit matching degree. At the same time, the second behavior data can include the direction of travel of non-associated persons, and the direction of travel can be matched with the direction of the first frequented location information and the direction of the second frequented location information. When the direction of travel is opposite to the direction of the first frequented location information or the same as the direction of the second frequented location information, the risk of loss to be adjusted is increased accordingly to obtain the risk of loss. On the contrary, when the direction of travel is the same as the first frequented location information or different from the second frequented location information, the risk of loss to be adjusted is reduced accordingly to obtain the risk of loss.

[0088] It should be noted that when the above-mentioned non-associated person enters the above-mentioned area for the first time and the above-mentioned non-associated person is not a specific person, there is no personnel file for the above-mentioned non-associated person in the personnel file library corresponding to the area. At this time, the personnel file of the above-mentioned non-associated person needs to be retrieved from the personnel file library of other areas, but the second frequented location information in the personnel file of other areas has no reference value in the current area. Therefore, when matching the travel direction with the frequented location information, only the travel direction and the direction of the first frequented location information can be considered for matching.

[0089] Optionally, in the step of determining the target child's lost risk value based on the second behavior data, the first frequent location information and the second frequent location information when the child is in a preset occasion, the personal identity of the non-associated person can also be obtained, and the personal identity includes a specific person and a non-specific person; determine whether the direction of travel is opposite to the direction of the first frequent location information; determine whether the direction of travel is the same as the direction of the second frequent location information; when the child is in the preset occasion, the personal identity of the non-associated person is a specific person, and the direction of travel is opposite to the direction of the first frequent location information or the same as the second frequent location information, the lost risk value is a preset first risk value; when the child is in the preset occasion, the personal identity of the non-associated person is a non-specific person, and the direction of travel is opposite to the first frequent location information or the same as the second frequent location information, the lost risk value is a preset second risk value.

[0090] In an embodiment of the present invention, the second behavior data includes a moving direction, which can be understood as the direction in which the person holding the target child moves when holding the target child. The specific person is a person with a record of violating the law. Correspondingly, the non-specific person may be a person without a record of violating the law.

[0091] Specifically, it can be determined at the same time whether the above-mentioned direction of travel is opposite to the direction of the first frequented location information, or the same as the direction of the second frequented location information. After the determination is completed, when the child is in a preset occasion, and the identity of the non-associated person is a specific person, and the direction of travel is opposite to the direction of the first frequented location information or the same as the second frequented location information, in such a case, the possibility of the child being lost is high, and the risk value is high, and it can be directly determined that the child is lost, and the lost risk value is determined as a preset first risk value, and the above-mentioned first risk value can be the upper limit of the above-mentioned lost risk value. For example, when the above-mentioned lost risk value is in the range of 0-100, the first risk value can be 100, and when the above-mentioned lost risk value is in the range of 0-120, the first risk value can be 120.

[0092] When the child is in a preset situation and the direction of travel is opposite to the first frequented location information or the same as the second frequented location information, but the identity of the non-associated person is not a specific person, in this case, the possibility of the child getting lost is also high, but the risk value is lower than that of the specific person. Therefore, the first risk value can be adjusted down accordingly to obtain the second risk value, and the second risk value can be used as the lost risk value. For example, when the first risk value is 100, the second risk value can be 80, and when the first risk value is 120, the second risk value can be 100.

[0093] Optionally, in the step of determining the target child's lost risk value based on the first behavior data, the second behavior data, the first frequent location information and the second frequent location information, the personal identity and second behavior habit information of the non-associated person, and the first behavior habit information corresponding to the target child can also be obtained; based on the personal identity, determine the identity risk value corresponding to the target child; based on the first distance between the collection location and the first frequent location information and the second distance between the collection location and the second frequent location information, determine the distance risk value corresponding to the target child; based on the ratio of the travel speed and the preset basic speed, determine the speed risk value corresponding to the target child; based on the degree of facial occlusion, determine the occlusion risk value corresponding to the target child; based on the first behavior habit information, the first behavior action and the second behavior habit information, the second behavior action, determine the habit risk value corresponding to the target child; add the identity risk value, distance risk value, speed risk value, occlusion risk value and habit risk value to obtain the lost risk value.

[0094] In an embodiment of the present invention, the first behavior data includes the first behavior action and the degree of facial occlusion of the target child, the second behavior data includes the second behavior action and the travel speed of the non-associated person, the real-time tracking image corresponds to a collection location, the first behavior data can be stored in the child file corresponding to the target child, and the second behavior data can be stored in the personnel file of the non-associated person. When calculating the risk value of getting lost, the first behavior data can be extracted from the child file, and the second behavior data can be extracted from the personnel file.

[0095] Specifically, each risk value can be divided into corresponding numerical sub-intervals according to the numerical interval in which the above-mentioned lost risk value is located. According to the upper limit of each numerical sub-interval, as well as the above-mentioned first behavior data, second behavior data, and collection location, the above-mentioned identity risk value, distance risk value, speed risk value, occlusion risk value and habit risk value can be calculated to obtain the above-mentioned lost risk value.

[0096] For example, assuming that when the identity of the non-associated person is a specific person, the numerical range of the lost risk value can be set to 0-120, and when the identity of the non-associated person is a non-specific person, the numerical range of the lost risk value can be set to 0-100.

[0097] At this time, the numerical sub-interval corresponding to the identity risk value is 0-40, the numerical sub-interval corresponding to the distance risk value is 0-40, and the numerical sub-intervals corresponding to the occlusion risk value, speed risk value, and habit risk value are all 0-20. The upper limit of the numerical sub-interval corresponding to the identity risk value is 40 (W1), the upper limit of the numerical sub-interval corresponding to the distance risk value is 40 (W2), the upper limit of the numerical sub-interval corresponding to the speed risk value is 20 (W3), the upper limit of the numerical sub-interval corresponding to the occlusion risk value is 20 (W4), and the upper limit of the numerical sub-interval corresponding to the habit risk value is 20 (W5).

[0098] Therefore, when the identity of the non-associated person is a specific person, the identity risk value may be the upper limit of the corresponding numerical sub-interval (i.e., 40). When the identity of the non-associated person is a non-specific person, the identity risk value may be the lower limit of the corresponding numerical sub-interval (i.e., 0).

[0099] The first distance risk value may be determined according to the ratio between the first distance and the base distance, the second distance risk value may be determined according to the ratio between the second distance and the base distance, and the distance risk value may be determined according to the first distance risk value and the second distance risk value.

[0100] The above speed risk value can be further explained by the following formula:

[0101] DRV=W3×(V / V0)^m

[0102] Among them, the above-mentioned DRV is represented by the speed risk value, the above-mentioned W3 is represented by the upper limit of the numerical sub-interval corresponding to the speed risk value, the above-mentioned V is represented by the above-mentioned travel speed, the above-mentioned V0 is represented by the basic speed, for example, it can be the normal walking speed of the age group of the above-mentioned non-associated personnel, and the above-mentioned m can be the speed adjustment coefficient, which is used to control the rate of increase of the speed risk value.

[0103] The occlusion risk value may be proportional to the degree of occlusion of the face. When the degree of occlusion of the face is higher, the occlusion risk value is higher. Conversely, when the degree of occlusion of the face is lower, the occlusion risk value is lower.

[0104] The habit risk value may be inversely proportional to the first habit matching degree and the second habit matching degree. The first habit matching degree is the matching degree between the first behavior habit information and the first behavior action, and the second habit matching degree is the matching degree between the second behavior habit information and the second behavior action.

[0105] It should be noted that since each risk value has an upper limit, during the calculation process, if the calculated value is greater than the upper limit, for example, the calculated value is 45 and the corresponding upper limit is 40, then it can be determined that the corresponding risk value is the upper limit 40, not 45.

[0106] Optionally, in the step of determining the distance risk value corresponding to the target child based on the first distance between the collection location and the first frequented location information and the second distance between the collection location and the second frequented location information, the first distance risk value corresponding to the target child can also be determined based on the ratio between the first distance and the preset basic distance; the second distance risk value corresponding to the target child can be determined based on the ratio between the preset basic distance and the second distance; when the personal identity of the non-associated person is a non-specific person, the first distance risk value is used as the distance risk value; when the personal identity of the non-associated person is a specific person, the first distance risk value is added to the second distance risk value to obtain the distance risk value.

[0107] In an embodiment of the present invention, the numerical interval corresponding to the above-mentioned first distance risk value and the numerical interval corresponding to the above-mentioned second distance risk value are both within the numerical sub-interval corresponding to the above-mentioned distance risk value. For example, when the numerical sub-interval corresponding to the above-mentioned distance risk value is 0-40, and its upper limit is 40 (W2), then the numerical interval corresponding to the above-mentioned first distance risk value can be 0-20, and its upper limit is 20 (W21), and the numerical interval corresponding to the above-mentioned second distance risk value can also be 0-20, and its upper limit is 20 (W22).

[0108] Therefore, the corresponding first distance risk value or second distance risk value can be determined according to the upper limit of the numerical range corresponding to each risk value, as well as the corresponding distance and the basic distance.

[0109] For example, the calculation process of the first distance risk value can be further described by the following formula:

[0110] FDRV=W21×(D / D0)^k

[0111] Among them, the above FDRV represents the above first distance risk value, the above W21 represents the upper limit of the numerical range corresponding to the above first distance risk value, the above D represents the above first distance, the above D0 represents the basic distance, which can be obtained through historical lost risk value calculation experience or through a limited number of tests, and can usually be a small and reasonable starting distance, and the above k represents the adjustment coefficient of the first distance risk value, which is used to control the rate at which the first distance risk value increases with the distance. That is, the longer the first distance, the higher the first distance risk value.

[0112] The calculation process of the above second distance risk value can be further explained by the following formula:

[0113] SDRV=W22×(D0 / D)^k

[0114] Among them, the above SDRV represents the above second distance risk value, the above W22 represents the upper limit of the numerical range corresponding to the above second distance risk value, the above D represents the above second distance, the above D0 represents the basic distance, which can be obtained through the historical lost risk value calculation experience or through a limited number of tests, and can usually be a small and reasonable starting distance, and the above k represents the adjustment coefficient of the second distance risk value, which is used to control the rate at which the second distance risk value increases with the distance. That is, the closer the second distance, the higher the second distance risk value.

[0115] It should be noted that when the above-mentioned non-related person is not a specific person, the impact on the child's loss is relatively small, so when it is not a specific person, the second distance risk value can be ignored. On the contrary, when the above-mentioned non-related person is a specific person, the impact on the child's loss is relatively large, so when it is a specific person, the second distance risk value can be considered. Taking the identity of the non-related person as one of the calculation bases of the distance risk value can improve the calculation accuracy of the distance risk value.

[0116] Optionally, in the step of determining the occlusion risk value corresponding to the target child based on the degree of facial occlusion, current weather data corresponding to the target child may also be obtained; and the occlusion risk value corresponding to the target child may be determined based on the current weather data and the degree of facial occlusion.

[0117] In the embodiment of the present invention, under normal circumstances, the degree of facial occlusion corresponds to the weather. For example, on sunny days, the light is relatively mild, providing sufficient sight for children, and the degree of facial occlusion is usually low. On windy, snowy, or sandy days, in order to prevent children from being affected by the wind, snow, or sand, the degree of facial occlusion is usually high. Therefore, when the current weather data does not match the degree of facial occlusion, the occlusion risk value is high, and when the current weather data can match the degree of facial occlusion, the occlusion risk value is low.

[0118] Specifically, when different weather data are determined based on historical tracking images, the corresponding degree of facial occlusion can be determined, and a correspondence table can be established for different weather data and corresponding degrees of facial occlusion. After obtaining the degree of facial occlusion of the target child and the current weather data, in the correspondence table, the reasonable degree of facial occlusion corresponding to the current weather data is determined based on the correspondence between the weather data and the degree of facial occlusion. The greater the difference between the reasonable degree of facial occlusion and the degree of facial occlusion of the target child, the higher the above-mentioned occlusion risk value, and conversely, the smaller the difference between the reasonable degree of facial occlusion and the degree of facial occlusion of the target child, the lower the above-mentioned occlusion risk value. It should be noted that the unit of the above-mentioned degree of facial occlusion can be a percentage.

[0119] like Figure 2 As shown, an embodiment of the present invention further provides an image processing device, comprising:

[0120] The identification module 201 is used to, when it is identified in the real-time tracking image that the person holding the target child of a preset age group is an unrelated person, perform behavior recognition processing based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person, and the unrelated person has no relative relationship with the target child;

[0121] A calculation module 202, configured to calculate a risk value of the target child being lost based on the first behavior data and the second behavior data;

[0122] A determination module 203 is used to determine the warning information corresponding to the target child based on the risk value of being lost;

[0123] The sending module 204 is used to send the warning information to relevant personnel corresponding to the target child.

[0124] Optionally, the image processing device further includes:

[0125] an acquisition module, configured to acquire a real-time tracking image corresponding to the child and a relationship map corresponding to the child when a child meeting preset conditions is identified, wherein the preset conditions include a preset age group and being held, and the relationship map includes associated persons who have a kinship relationship with the child;

[0126] An extraction module, used for extracting the associated persons and the embraced persons one by one in the relationship graph for matching, and obtaining matching results of the embraced persons;

[0127] A second determination module is used to determine whether the person holding the child is an unrelated person based on the matching result;

[0128] The third determination module is used to determine the child as the target child when the person holding the child is an unrelated person.

[0129] Optionally, the calculation module 202 includes:

[0130] A first acquisition submodule is used to acquire the first frequently visited place information of the target child and the second frequently visited place information of the unrelated person;

[0131] A first determination submodule, configured to determine a risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information when the child is in a preset occasion;

[0132] The second determination submodule is used to determine the risk value of the target child being lost based on the first behavior data, the second behavior data, the first frequented location information and the second frequented location information when the child is not in a preset occasion.

[0133] Optionally, the second behavior data includes a traveling direction, and the first determining submodule includes:

[0134] A first acquisition unit is used to acquire the identity of the non-associated person, where the identity includes a specific person and a non-specific person;

[0135] a first determination unit, configured to determine whether the traveling direction is opposite to the direction of the first frequently visited location information;

[0136] a second determination unit, configured to determine whether the traveling direction is the same as the direction of the second frequently visited location information;

[0137] a first processing unit, configured to, when the child is in a preset occasion, the non-associated person is a specific person, and the moving direction is opposite to the direction of the first frequently visited location information or is the same as the second frequently visited location information, the lost risk value is a preset first risk value;

[0138] The second processing unit is used for, when the child is in a preset occasion, the identity of the non-associated person is a non-specific person, and the travel direction is opposite to the first frequented location information or is the same as the second frequented location information, the lost risk value is a preset second risk value, and the first risk value is higher than the second risk value.

[0139] Optionally, the first behavior data includes a first behavior action and a degree of facial occlusion of the target child, the second behavior data includes a second behavior action and a moving speed of the unrelated person, and the second determination submodule includes:

[0140] A second acquisition unit is used to acquire the identity and second behavior habit information of the non-associated person, and the first behavior habit information corresponding to the target child;

[0141] A first determining unit, configured to determine an identity risk value corresponding to the target child based on the identity of the person;

[0142] a second determining unit, configured to determine a distance risk value corresponding to the target child based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information;

[0143] a third determining unit, configured to determine a speed risk value corresponding to the target child based on a ratio of the traveling speed to a preset basic speed;

[0144] a fourth determining unit, configured to determine an occlusion risk value corresponding to the target child based on the degree of occlusion of the face;

[0145] a fifth determining unit, configured to determine a habit risk value corresponding to the target child based on the first behavior habit information, the first behavior action, and the second behavior habit information, the second behavior action;

[0146] The third processing unit is used to add the identity risk value, the distance risk value, the speed risk value, the occlusion risk value and the habit risk value to obtain the lost risk value.

[0147] Optionally, the second determining unit includes:

[0148] A first determining subunit, configured to determine a first distance risk value corresponding to the target child based on a ratio between the first distance and a preset basic distance;

[0149] A second determining subunit is used to determine a second distance risk value corresponding to the target child based on a ratio between the preset basic distance and the second distance;

[0150] a first processing subunit, configured to use the first distance risk value as the distance risk value when the identity of the non-associated person is a non-specific person;

[0151] The second processing subunit is configured to add the first distance risk value and the second distance risk value to obtain the distance risk value when the identity of the non-associated person is a specific person.

[0152] Optionally, the fourth determining unit includes:

[0153] A first acquisition subunit is used to acquire current weather data corresponding to the target child;

[0154] The third determination subunit is used to determine the occlusion risk value corresponding to the target child based on the current weather data and the degree of occlusion of the face.

[0155] like Figure 3 As shown, an embodiment of the present invention further provides an electronic device, characterized in that it includes a processor, and the processor can execute any one of the above-mentioned image processing methods.

[0156] Specifically, it includes a processor 301 and a memory 302, and a computer program for executing the image processing method which is stored in the memory 302 and can be run on the processor 301, wherein:

[0157] The processor 301 runs the computer program of the image processing method stored in the memory 302 to perform the following steps:

[0158] When a person holding a target child of a preset age group is identified as an unrelated person in a real-time tracking image, a behavior recognition process is performed based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person, and the unrelated person has no relative relationship with the target child;

[0159] Calculating a risk value of the target child being lost based on the first behavior data and the second behavior data;

[0160] Based on the risk value of getting lost, determining the warning information corresponding to the target child;

[0161] The warning information is sent to relevant personnel corresponding to the target child.

[0162] Optionally, when it is identified in the real-time tracking image that the person holding the target child of a preset age group is an unrelated person, before the behavior recognition process is performed based on the real-time tracking image, the method executed by the processor 301 further includes:

[0163] When a child meeting preset conditions is identified, a real-time tracking image corresponding to the child and a relationship map corresponding to the child are obtained, wherein the preset conditions include a preset age group and being held, and the relationship map includes associated persons who have a kinship relationship with the child;

[0164] In the relationship graph, the associated persons are extracted one by one for matching with the embraced persons, and matching results of the embraced persons are obtained;

[0165] Based on the matching result, determining whether the person holding the child is an unrelated person;

[0166] When the person holding the child is an unrelated person, the child is determined as the target child.

[0167] Optionally, the processor 301 calculates the target child's risk value of getting lost based on the first behavior data and the second behavior data, including:

[0168] Acquire the first frequently visited location information of the target child and the second frequently visited location information of the unrelated person;

[0169] When the child is in a preset occasion, determining a risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information;

[0170] When the child is not in a preset situation, the risk value of the target child being lost is determined based on the first behavior data, the second behavior data, the first frequented location information, and the second frequented location information.

[0171] Optionally, the second behavior data includes a traveling direction, and the processor 301 determines the risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information when the child is in a preset occasion, including:

[0172] Obtaining the identity of the non-associated person, where the identity includes a specific person and a non-specific person;

[0173] determining whether the traveling direction is opposite to the direction of the first frequently visited location information;

[0174] determining whether the traveling direction is the same as the direction of the second frequently visited location information;

[0175] When the child is in a preset situation, the non-associated person is a specific person, and the moving direction is opposite to the direction of the first frequently visited location information or is the same as the second frequently visited location information, the lost risk value is a preset first risk value;

[0176] When the child is in a preset occasion, the identity of the non-associated person is a non-specific person, and the traveling direction is opposite to the first frequented location information or the same as the second frequented location information, the lost risk value is a preset second risk value, and the first risk value is higher than the second risk value.

[0177] Optionally, the first behavior data includes a first behavior action and a degree of facial occlusion of the target child, the second behavior data includes a second behavior action and a moving speed of the unrelated person, the real-time tracking image corresponds to a collection location, and the processor 301 determines the target child's lost risk value based on the first behavior data, the second behavior data, the first frequently visited location information, and the second frequently visited location information, including:

[0178] Acquire the identity and second behavior habit information of the non-associated person, and the first behavior habit information corresponding to the target child;

[0179] Based on the identity of the person, determining an identity risk value corresponding to the target child;

[0180] Determining a distance risk value corresponding to the target child based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information;

[0181] Determining a speed risk value corresponding to the target child based on a ratio of the traveling speed to a preset basic speed;

[0182] Determining an occlusion risk value corresponding to the target child based on the degree of occlusion of the face;

[0183] Determining a habit risk value corresponding to the target child based on the first behavior habit information, the first behavior action, and the second behavior habit information, the second behavior action;

[0184] The identity risk value, the distance risk value, the speed risk value, the occlusion risk value and the habit risk value are added together to obtain the lost risk value.

[0185] Optionally, the determining, by the processor 301, of a distance risk value corresponding to the target child based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information includes:

[0186] Determining a first distance risk value corresponding to the target child based on a ratio between the first distance and a preset basic distance;

[0187] Determining a second distance risk value corresponding to the target child based on a ratio between the preset basic distance and the second distance;

[0188] When the identity of the non-associated person is a non-specific person, the first distance risk value is used as the distance risk value;

[0189] When the identity of the non-associated person is a specific person, the first distance risk value is added to the second distance risk value to obtain the distance risk value.

[0190] Optionally, the determining, based on the degree of facial occlusion, the occlusion risk value corresponding to the target child, performed by the processor 301 includes:

[0191] Obtaining current weather data corresponding to the target child;

[0192] Based on the current weather data and the degree of facial occlusion, a occlusion risk value corresponding to the target child is determined.

[0193] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the image processing method or the application-side image processing method provided by the embodiment of the present invention are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0194] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the above-mentioned computer program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the above-mentioned computer-readable storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0195] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. An image processing method, characterized in that: The method comprises the following steps: When a person holding a target child of a preset age group is identified as an unrelated person in a real-time tracking image, a behavior recognition process is performed based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person, and the unrelated person has no relative relationship with the target child; Calculating a risk value of the target child being lost based on the first behavior data and the second behavior data; Based on the risk value of getting lost, determining the warning information corresponding to the target child; The warning information is sent to relevant personnel corresponding to the target child.

2. The image processing method according to claim 1, characterized in that: When the person holding the target child of the preset age group is identified as an unrelated person in the real-time tracking image, before the behavior recognition process is performed based on the real-time tracking image, the method further includes: When a child meeting preset conditions is identified, a real-time tracking image corresponding to the child and a relationship map corresponding to the child are obtained, wherein the preset conditions include a preset age group and being held, and the relationship map includes associated persons who have a kinship relationship with the child; In the relationship graph, the associated persons are extracted one by one for matching with the embraced persons, and matching results of the embraced persons are obtained; Based on the matching result, determining whether the person holding the child is an unrelated person; When the person holding the child is an unrelated person, the child is determined as the target child.

3. The image processing method according to claim 1, characterized in that: The calculating the risk value of the target child being lost based on the first behavior data and the second behavior data includes: Acquire the first frequently visited location information of the target child and the second frequently visited location information of the unrelated person; When the child is in a preset occasion, determining a risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information; When the child is not in a preset situation, the risk value of the target child being lost is determined based on the first behavior data, the second behavior data, the first frequented location information, and the second frequented location information.

4. The image processing method according to claim 3, characterized in that: The second behavior data includes a traveling direction, and when the child is in a preset occasion, determining the risk value of the target child being lost based on the second behavior data, the first frequently visited location information, and the second frequently visited location information, includes: Obtaining the identity of the non-associated person, where the identity includes a specific person and a non-specific person; determining whether the traveling direction is opposite to the direction of the first frequently visited location information; determining whether the traveling direction is the same as the direction of the second frequently visited location information; When the child is in a preset situation, the non-associated person is a specific person, and the moving direction is opposite to the direction of the first frequently visited location information or is the same as the second frequently visited location information, the lost risk value is a preset first risk value; When the child is in a preset occasion, the identity of the non-associated person is a non-specific person, and the traveling direction is opposite to the first frequented location information or the same as the second frequented location information, the lost risk value is a preset second risk value, and the first risk value is higher than the second risk value.

5. The image processing method according to claim 3, characterized in that: The first behavior data includes a first behavior action and a degree of facial occlusion of the target child, the second behavior data includes a second behavior action and a moving speed of the unrelated person, the real-time tracking image corresponds to a collection location, and determining the lost risk value of the target child based on the first behavior data, the second behavior data, the first frequently visited location information, and the second frequently visited location information includes: Acquire the identity and second behavior habit information of the non-associated person, and the first behavior habit information corresponding to the target child; Based on the identity of the person, determining an identity risk value corresponding to the target child; Determining a distance risk value corresponding to the target child based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information; Determining a speed risk value corresponding to the target child based on a ratio of the traveling speed to a preset basic speed; Determining an occlusion risk value corresponding to the target child based on the degree of occlusion of the face; Determining a habit risk value corresponding to the target child based on the first behavior habit information, the first behavior action, and the second behavior habit information, the second behavior action; The identity risk value, the distance risk value, the speed risk value, the occlusion risk value and the habit risk value are added together to obtain the lost risk value.

6. The image processing method according to claim 5, characterized in that: The determining, based on a first distance between the collection location and the first frequently visited location information and a second distance between the collection location and the second frequently visited location information, a distance risk value corresponding to the target child includes: Determining a first distance risk value corresponding to the target child based on a ratio between the first distance and a preset basic distance; Determining a second distance risk value corresponding to the target child based on a ratio between the preset basic distance and the second distance; When the identity of the non-associated person is a non-specific person, the first distance risk value is used as the distance risk value; When the identity of the non-associated person is a specific person, the first distance risk value is added to the second distance risk value to obtain the distance risk value.

7. The image processing method according to claim 5, characterized in that: The determining, based on the degree of facial occlusion, a occlusion risk value corresponding to the target child includes: Obtaining current weather data corresponding to the target child; Based on the current weather data and the degree of facial occlusion, a occlusion risk value corresponding to the target child is determined.

8. An image processing device, characterized in that: The image processing device comprises: an identification module, configured to, when identifying in a real-time tracking image that a person holding a target child of a preset age group is an unrelated person, perform behavior identification processing based on the real-time tracking image to obtain first behavior data of the target child and second behavior data of the unrelated person, wherein the unrelated person has no relative relationship with the target child; A calculation module, configured to calculate a risk value of the target child being lost based on the first behavior data and the second behavior data; A determination module, configured to determine the warning information corresponding to the target child based on the risk value of being lost; The sending module is used to send the warning information to relevant personnel corresponding to the target child.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the image processing method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the image processing method according to any one of claims 1 to 7 are implemented.

Images