A method for assessing the risk of smuggling and transit and an electronic device

Abstract

Embodiments of the present application provide a smuggling transit risk assessment method and an electronic device, the method comprising: determining a first risk corresponding to each target object; determining at least one region condition that is satisfied by a target region among a plurality of preset region conditions, and determining a second risk coefficient; scaling the second risk coefficient to obtain a second risk after increasing; and aggregating the second risk to obtain a third risk of smuggling occurring in the target region. The application can dynamically adjust the risk coefficient according to the region condition of the target region when the target object appears in the target region, so as to improve the accuracy of the determined second risk coefficient, and further make the scaled risk coefficient more consistent with the real smuggling scene, so that the risk of smuggling occurring in the target region can be more accurately evaluated.

Description

Technical Field

[0001] This invention relates to the field of maritime inspection technology, and in particular to a method and electronic equipment for assessing the risks of smuggling and transshipment. Background Technology

[0002] Existing technologies often employ a uniform risk assessment method to evaluate smuggling risks. However, real-world smuggling scenarios are complex and ever-changing, making it difficult to apply fixed parameters or uniform standards to such diverse situations. This deficiency directly results in a significant discrepancy between risk assessment results and actual smuggling risks, leading to low accuracy in smuggling and transshipment risk assessment. Summary of the Invention

[0003] The purpose of this invention is to provide a method and electronic device for assessing smuggling and transshipment risks, thereby improving the accuracy of such risk assessments. The specific technical solution is as follows:

[0004] In a first aspect, embodiments of the present invention provide a method for assessing the risk of smuggling and transshipment, the method comprising:

[0005] For each target object appearing in the target area, the risk of the target object participating in smuggling is determined based on the object information of the target object, which is regarded as the first risk corresponding to the target object;

[0006] For each target object, at least one regional condition is determined from multiple preset regional conditions that the target region must satisfy when the target object appears in the target region. The first risk coefficients corresponding to all the determined regional conditions are aggregated to obtain the second risk coefficient corresponding to the target object.

[0007] For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0008] The second risks corresponding to each of the target objects are aggregated to obtain the third risk of smuggling occurring in the target area.

[0009] In one possible implementation, determining at least one regional condition that the target region satisfies when the target object appears in a target region from among multiple preset regional conditions, and aggregating the first risk coefficients corresponding to all determined regional conditions to obtain a second risk coefficient corresponding to the target object, includes:

[0010] Obtain the feature models pre-built for each of the multiple regional conditions;

[0011] The feature model and the currently matched regional conditions are determined as the conditions that the target region satisfies, and the preset first risk coefficient of each determined regional condition is obtained.

[0012] The first risk coefficients obtained are aggregated according to preset rules to obtain the second risk coefficient corresponding to the target object.

[0013] In one possible implementation, the step of aggregating all the obtained first risk coefficients according to a preset rule to obtain the second risk coefficient corresponding to the target object includes:

[0014] The obtained first risk coefficient is substituted into a preset calculation formula to calculate the second risk coefficient of the target object; or...

[0015] In a preset mapping table, the second risk coefficient corresponding to at least one combination of first risk coefficients determined by the current area conditions is found and used as the second risk coefficient corresponding to the target object. The preset mapping table includes second risk coefficients corresponding to various permutations and combinations of the first risk coefficients; or...

[0016] The determined regional conditions and the obtained first risk coefficient are input into the large model, and the large model is guided to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

[0017] In one possible implementation, determining at least one regional condition that the target region satisfies when the target object appears in a target region from among multiple preset regional conditions, and aggregating the first risk coefficients corresponding to all determined regional conditions to obtain a second risk coefficient corresponding to the target object, includes:

[0018] The regional information of the target area when the target object appears is input into the large model to guide the large model to deduce the regional conditions satisfied by the target area based on the scenario of the target area, and to deduce the second risk coefficient corresponding to the target object based on the deduced regional conditions and the first risk coefficient of the regional conditions; or...

[0019] When the target object appears in the target area, the area information of the target area is input into the large model, and the large model is guided to comprehensively consider the scene of the target area, the type of the determined area conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

[0020] In one possible implementation, the aggregation of the second risks corresponding to each of the target objects to obtain the third risk of smuggling occurring in the target area includes:

[0021] For each category, the second risk corresponding to each target object in the category is aggregated, and the smaller of the aggregation result and the preset cutoff threshold is taken as the fourth risk corresponding to the category.

[0022] The fourth risks corresponding to each of the aforementioned categories are aggregated to obtain the third risk, which serves as the risk of smuggling occurring in the target area.

[0023] In one possible implementation, the preset cutoff threshold is determined by the following methods:

[0024] Obtain object information for each sample object, wherein the sample object is an object that appears multiple times at the smuggling site within a preset period;

[0025] For each category, the degree of variation among the object information of the sample objects in that category is statistically calculated, and this degree of variation is used as the degree of variation corresponding to that category.

[0026] For each category, the smuggling proportion corresponding to that category is calculated based on the frequency of occurrence of the sample objects of that category at the smuggling site;

[0027] The objective weight of each category is determined based on the degree of variation and the proportion of smuggling for each category, wherein the objective weight is negatively correlated with the degree of variation and positively correlated with the proportion of smuggling.

[0028] The objective weight of each category is weighted and summed with the preset subjective weight of each category to obtain the comprehensive weight.

[0029] The preset cutoff threshold is determined based on the comprehensive weight, wherein the preset cutoff threshold is positively correlated with the comprehensive weight.

[0030] In one possible implementation, determining the risk of the target object participating in smuggling based on the object information of the target object, as the first risk corresponding to the target object, includes:

[0031] The object information of the target object is inferred using at least one inference model to obtain at least one inference result of the target object and the confidence level of each inference result;

[0032] Based on all the reasoning results of the target object, the risk of the target object participating in smuggling is determined as the first risk corresponding to the target object;

[0033] The method further includes:

[0034] Based on the confidence level of each of the inference results, the confidence level of the third risk is determined;

[0035] The corresponding third risk and its confidence level are displayed.

[0036] In one possible implementation, the correspondence demonstrates the third risk and the confidence level of the third risk, including:

[0037] The corresponding information includes the third risk, the confidence level of the third risk, and the reasoning results for each target object.

[0038] In response to the correction of the reasoning result, the object information of the object to which the corrected reasoning result belongs is determined, and the reasoning model of the corrected reasoning result is obtained.

[0039] The determined inference model is trained based on the corrected inference results and the identified object information.

[0040] In one possible implementation, the method further includes:

[0041] Acquire object information of each object collected by sensors set in the target area during a target time period, and first human-generated intelligence input for the target time period and the target area, wherein the first human-generated intelligence includes object information of potential smuggling targets; identify objects whose object information meets preset anomaly rules and objects whose object information is associated with the object information of the potential smuggling targets as target objects; and / or,

[0042] Acquire each second human intelligence input for the target area, wherein the second human intelligence is used to describe potential smuggling activities and the expected time period of the potential smuggling activities; among the potential smuggling activities described by each second human intelligence, find potential smuggling activities whose expected time period includes the target time period, and take them as target potential smuggling activities;

[0043] Based on the potential smuggling activities of each target, the risk of smuggling occurring in the target area during the target time period is determined as the fifth risk; the aggregation of the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area during the target time period includes: aggregating the second risks and the fifth risk corresponding to each target object to obtain the third risk of smuggling occurring in the target area during the target time period; and / or,

[0044] The step of determining the risk of the target object participating in smuggling based on the object information of the target object, as the first risk corresponding to the target object, includes: determining the category to which the target object belongs based on the object information of the target object, and the preset anomaly rule triggered by the category; determining the risk of the target object participating in smuggling based on the preset anomaly rule triggered by the category, as the first risk corresponding to the target object.

[0045] Secondly, embodiments of the present invention provide a smuggling and transshipment risk assessment device, the device comprising:

[0046] The first determining module is used to determine the risk of the target object participating in smuggling based on the object information of each target object appearing in the target area, and to use this as the first risk corresponding to the target object.

[0047] The second determining module is used to determine, for each target object, at least one regional condition that the target region satisfies when the target object appears in the target region from multiple preset regional conditions, and to aggregate the first risk coefficients corresponding to all the determined regional conditions to obtain the second risk coefficient corresponding to the target object.

[0048] The first scaling module is used to scale the first risk corresponding to each target object using the second risk coefficient corresponding to the target object, so as to obtain the scaled second risk corresponding to the target object.

[0049] The first aggregation module is used to aggregate the second risks corresponding to each of the target objects to obtain the third risk of smuggling occurring in the target area.

[0050] In one possible implementation, the second determining module includes:

[0051] The first determining submodule is used to obtain the feature models pre-constructed for each of the multiple regional conditions;

[0052] The second determining submodule is used to determine the feature model and the currently matched regional conditions as the conditions satisfied by the target region, and to obtain the preset first risk coefficient of each determined regional condition;

[0053] The third determining submodule is used to aggregate all the obtained first risk coefficients according to preset rules to obtain the second risk coefficient corresponding to the target object.

[0054] In one possible implementation, the third determining submodule includes:

[0055] The first determining unit is used to input the acquired first risk coefficient into a preset calculation formula to calculate the second risk coefficient of the target object; or...

[0056] The second determining unit is configured to search in a preset mapping table for a second risk coefficient corresponding to a combination of at least one first risk coefficient determined by the current area conditions, and use this second risk coefficient as the second risk coefficient corresponding to the target object. The preset mapping table includes second risk coefficients corresponding to various permutations and combinations of the first risk coefficients; or...

[0057] The third determining unit is used to input the determined regional conditions and the obtained first risk coefficient into the large model, and guide the large model to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

[0058] In one possible implementation, the second determining module includes:

[0059] The fourth determination submodule is used to input the regional information of the target area when the target object appears in the target area into the large model, so as to guide the large model to think about the regional conditions satisfied by the target area based on the scenario of the target area, and to think about the second risk coefficient corresponding to the target object based on the regional conditions obtained and the first risk coefficient of the regional conditions; or,

[0060] The fifth determination submodule is used to input the regional information of the target area when the target object appears in the target area into the large model, and guide the large model to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

[0061] In one possible implementation, the first aggregation module includes:

[0062] The first aggregation submodule is used to aggregate the second risks corresponding to each target object of each category for each category, and take the smaller of the aggregation result and the preset truncation threshold as the fourth risk corresponding to the category.

[0063] The second aggregation submodule is used to aggregate the fourth risks corresponding to each of the categories to obtain the third risk, which serves as the risk of smuggling occurring in the target area during the target time period.

[0064] In one possible implementation, the preset cutoff threshold is determined by the following methods:

[0065] Obtain object information for each sample object, wherein the sample object is an object that appears multiple times at the smuggling site within a preset period;

[0066] For each category, the degree of variation among the object information of the sample objects in that category is statistically calculated, and this degree of variation is used as the degree of variation corresponding to that category.

[0067] For each category, the smuggling proportion corresponding to that category is calculated based on the frequency of occurrence of the sample objects of that category at the smuggling site;

[0068] The objective weight of each category is determined based on the degree of variation and the proportion of smuggling for each category, wherein the objective weight is negatively correlated with the degree of variation and positively correlated with the proportion of smuggling.

[0069] The objective weight of each category is weighted and summed with the preset subjective weight of each category to obtain the comprehensive weight.

[0070] The preset cutoff threshold is determined based on the comprehensive weight, wherein the preset cutoff threshold is positively correlated with the comprehensive weight.

[0071] In one possible implementation, the first determining module includes:

[0072] The fifth determining submodule is used to infer the object information of the target object using at least one inference model, and to obtain at least one inference result of the target object and the confidence level of each inference result;

[0073] The sixth determining submodule is used to determine the risk of the target object participating in smuggling based on all the reasoning results of the target object, as the first risk corresponding to the target object;

[0074] The device further includes:

[0075] The third determining module is used to determine the confidence level of the third risk based on the confidence level of each of the inference results;

[0076] The first display module is used to display the third risk and the confidence level of the third risk.

[0077] In one possible implementation, the first display module includes:

[0078] The first display submodule is used to display the third risk, the confidence level of the third risk, and the reasoning results of each target object.

[0079] The second display submodule is used to, in response to the correction of the reasoning result, determine the object information of the object to which the corrected reasoning result belongs, and reason to obtain the reasoning model of the corrected reasoning result.

[0080] The third display submodule is used to train the determined inference model based on the corrected inference results and the determined object information.

[0081] In one possible implementation, the device further includes:

[0082] The fourth determining module is used to acquire object information of each object collected by sensors installed in the target area during the target time period, and first artificial intelligence input for the target time period and the target area, wherein the first artificial intelligence includes object information of potential smuggling objects; determine objects whose object information meets preset anomaly rules and objects whose object information is associated with the object information of the potential smuggling objects as target objects; and / or,

[0083] The fifth determining module is used to acquire each second piece of human-generated intelligence input for the target area, wherein the second piece of human-generated intelligence describes potential smuggling activities and the expected occurrence time of the potential smuggling activities; among the potential smuggling activities described by each piece of second piece of human-generated intelligence, potential smuggling activities whose expected occurrence time includes the target time period are identified as target potential smuggling activities; based on each target potential smuggling activity, the risk of smuggling activities occurring in the target area during the target time period is determined as the fifth risk; the aggregation of the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area during the target time period includes: aggregating the second risks and the fifth risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area during the target time period; and / or,

[0084] The sixth determining module is used to determine the risk of the target object participating in smuggling based on the object information of the target object, as the first risk corresponding to the target object, including: determining the category to which the target object belongs based on the object information of the target object, and the preset abnormal rules triggered by the category; determining the risk of the target object participating in smuggling based on the preset abnormal rules triggered by the category, as the first risk corresponding to the target object.

[0085] Thirdly, embodiments of the present invention provide an electronic device, including:

[0086] Memory, used to store computer programs;

[0087] When the processor executes the program stored in the memory, it implements the smuggling and transshipment risk assessment method of the first aspect mentioned above.

[0088] Fourthly, embodiments of this application also provide a computer program product containing instructions that, when run on a computer, cause the computer to execute the smuggling and transshipment risk assessment method described in the first aspect.

[0089] Beneficial effects of the embodiments of the present invention:

[0090] The present invention provides a method and apparatus for assessing the risk of smuggling and transshipment, which can dynamically adjust the risk coefficient according to the regional conditions of the target area when the target object appears in the target area, so as to improve the accuracy of the determined second risk coefficient, thereby making the scaled risk coefficient more closely match the real smuggling scenario, and thus more accurately assessing the risk of smuggling in the target area.

[0091] Of course, implementing any product or method of the present invention does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description

[0092] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings.

[0093] Figure 1 A flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention;

[0094] Figure 2 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0095] Figure 3 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0096] Figure 4 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0097] Figure 5 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0098] Figure 6 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0099] Figure 7 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0100] Figure 8 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0101] Figure 9This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0102] Figure 10 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0103] Figure 11 This is another flowchart illustrating the smuggling and transshipment risk assessment method provided by the present invention.

[0104] Figure 12 This is a schematic diagram of the smuggling and transshipment risk assessment device provided by the present invention;

[0105] Figure 13 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0106] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art based on this application are within the scope of protection of the present invention.

[0107] To address the problem that existing smuggling and transshipment risk assessment methods have poor applicability, resulting in low accuracy in some scenarios, this invention provides a smuggling and transshipment risk assessment method. (See attached...) Figure 1 The methods include:

[0108] S101, for each target object appearing in the target area, determine the risk of the target object participating in smuggling based on the object information of the target object, and take it as the first risk corresponding to the target object;

[0109] S102, for each target object, determine at least one regional condition that the target region must satisfy when the target object appears in the target region from multiple preset regional conditions, and aggregate the first risk coefficients corresponding to all the determined regional conditions to obtain the second risk coefficient corresponding to the target object.

[0110] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0111] S104, aggregate the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area.

[0112] Applying the above embodiments, for each target object appearing in the target area, the risk of the target object participating in smuggling is determined based on the object information of the target object, which is taken as the first risk corresponding to the target object. For each target object, at least one regional condition that the target area satisfies when the target object appears in the target area is determined from multiple preset regional conditions, and the first risk coefficients corresponding to all determined regional conditions are aggregated to obtain the second risk coefficient corresponding to the target object. For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object. The second risks corresponding to each target object are aggregated to obtain the third risk of smuggling occurring in the target area. By applying this method, the risk coefficient can be dynamically adjusted according to the regional conditions of the target area when the target object appears in the target area, so as to improve the accuracy of the determined second risk coefficient, and thus make the scaled risk coefficient more closely match the real smuggling scenario, thereby enabling a more accurate assessment of the risk of smuggling occurring in the target area.

[0113] The following will explain steps S101-S104:

[0114] In step S101, the target area is the geographical area that needs to be monitored for smuggling risks, such as wharves, ports, harbors, transportation routes, and regulatory areas.

[0115] The target object refers to the people, vehicles, ships, aircraft, cargo, and other objects that can be monitored and identified within the target area. It can be all objects within the target area or only some of them.

[0116] The target object's information includes, but is not limited to, its identity information, historical records, and behavioral information. Specifically, identity information includes unique identifiers for vehicles, ships, aircraft, and cargo, such as a person's ID number, a vehicle's license plate number, a ship's Automatic Identification System (AIS) code, an aircraft's product identification code, and cargo's cargo identification code (ID). Historical records include whether there are prior smuggling offenses or abnormal entry / exit records. Behavioral information includes driving routes, duration of stay, driving speed, and detour behaviors.

[0117] The first risk is the probability or degree of risk of a single target being involved in smuggling, calculated for that target itself. It can be a risk score, risk level (low / medium / high), or probability value.

[0118] In step S102, the preset area conditions are various scenarios, states or factors that induce or cause the target object to engage in smuggling activities.

[0119] For example, based on information such as the geographical features, meteorological patterns, and historical risk events of the target area, several possible pre-set negative conditions are established, such as heavy rainfall, strong winds, visibility below a threshold, road icing, and road closures. For each target object, combining its behavioral characteristics, activity patterns, and historical trajectories, the pre-set regional conditions that the target area's environment must meet when the target object appears in the target area are determined, and these are used as the corresponding negative conditions for that target object.

[0120] Based on the pre-defined regional conditions satisfied by the identified target object, the first risk coefficients corresponding to each pre-defined regional condition are aggregated to obtain the second risk coefficient corresponding to the target object. Specifically, a corresponding risk coefficient is pre-set for each pre-defined regional condition; a higher risk coefficient indicates an environment conducive to smuggling, while a lower risk coefficient indicates an environment unfavorable to smuggling. For example, the risk coefficient is set to 1.3 for nighttime (e.g., 22:00-05:00); 1.4 for foggy weather (e.g., visibility <200 meters); 1.2 for high tide weather (e.g., tide height >3.5 meters); and 1.1 for holidays.

[0121] Furthermore, regional conditions can include both negative and positive regional conditions. Negative regional conditions refer to conditions that are conducive to smuggling, such as the aforementioned foggy weather and high tide weather. The first risk coefficient corresponding to negative regional conditions is greater than 1. Positive regional conditions refer to conditions that are unfavorable to smuggling, such as fog-free weather and calm sea conditions. The first risk coefficient corresponding to positive regional conditions is less than 1.

[0122] Furthermore, although the first risk coefficient corresponding to each regional condition in the aforementioned examples is a fixed value, in other possible examples, the first risk coefficient corresponding to the regional condition may not be a fixed value. For example, the first risk coefficient corresponding to the regional condition may be a range. For instance, the first risk coefficient corresponding to the fog condition is the range [1.1, 1.3]. When the fog is light fog, the first risk coefficient is 1.1, and when the fog is dense fog, the first risk coefficient is 1.3. Then, when it is between light fog and dense fog, the first risk coefficient is (1.1, 1.3). The specific value of the first risk coefficient within the range can be calculated according to a preset calculation rule. For example, assuming the visibility in the current scene is X, the visibility is Xmin in light fog and Xmax in dense fog, it can be calculated according to the formula a = 1.1 + 0.2(X - Xmin) / (Xmax - Xmin), where a is the specific value of the first risk coefficient. The specific value of the first risk coefficient within the range can also be obtained by model recommendation. This model can be a small model trained from multiple sample scene images, with each sample image labeled with the true first risk coefficient. This model can also be a large model.

[0123] In step S103, for each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the second risk corresponding to the target object.

[0124] In step S104, the second risk obtained in step S103 is aggregated, and the aggregation result is used as the risk of smuggling in the target area, which is also the third risk.

[0125] The third risk is the overall risk of smuggling occurring in the target area, under the combined influence of the current distribution of target objects and the current external environment.

[0126] In one possible embodiment, the second risk coefficient corresponding to the target object can be determined in the following manner, see [reference]. Figure 2 The methods include:

[0127] S101, for each target object appearing in the target area, determine the risk of the target object participating in smuggling based on the object information of the target object, and take it as the first risk corresponding to the target object;

[0128] S1021, For each target object, obtain the feature model pre-built for each of the multiple regional conditions;

[0129] S1022, determine the feature model and the currently matched regional conditions as the conditions that the target region must meet, and obtain the preset first risk coefficient of each determined regional condition;

[0130] S1023, Aggregate all the first risk coefficients obtained according to preset rules to obtain the second risk coefficient corresponding to the target object;

[0131] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0132] S104, aggregate the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area.

[0133] Steps S1021-S1023 are detailed steps of the aforementioned step S102. Steps S101, S103 and S104 have been explained in the preceding text and will not be repeated here.

[0134] In step S1021, a unique feature model is pre-constructed for each region condition through sample training, feature extraction, and other methods. This model is used to identify whether the environmental information meets the region condition. Each region condition corresponds to a unique feature model to ensure the accuracy of the identification.

[0135] Environmental information is used to describe the environment of the target area when the target object appears in the target area, including but not limited to meteorological environmental information, geographical environmental information (topography, roads, obstructions, monitoring coverage), and temporal environmental information.

[0136] In step S1022, "determining the feature model and the currently matched regional conditions as the conditions satisfied by the target region" means that during the matching process between the feature model and the current scene, the corresponding regional conditions are selected and the regional conditions are used as the conditions satisfied by the target region.

[0137] For example, regional information of the target area can be obtained, and the feature model of each acquired regional condition can be matched one by one with the regional information of the target area when the target object appears in the target area. The matching process is essentially to identify the regional information through the feature model and determine whether the current regional state meets the feature requirements of the regional condition. If the degree of matching between the features of the regional information and the feature model of a certain regional condition reaches a preset matching threshold, then the regional condition is determined to be a "condition satisfied by the region"; if the matching threshold is not reached, then the regional condition is determined to be not currently satisfied.

[0138] As mentioned earlier, a first risk coefficient is pre-set for the conditions of each region, and the corresponding first risk coefficient is obtained according to the conditions met.

[0139] In step S1023, the preset rules are aggregation algorithms or calculation logic pre-set according to the risk assessment needs of the actual application scenario. They are used to integrate the risk coefficients of multiple individual regional conditions. The preset rules include, but are not limited to: summation rules (simply add all the first risk coefficients, which are applicable to scenarios where the risks of multiple regional conditions are superimposed), weighted summation rules (set weights according to the importance of each regional condition, multiply the weights by the first risk coefficients and then sum them, which are applicable to scenarios where different regional conditions have different degrees of influence), and maximum value rules (select the maximum value among all the first risk coefficients as the aggregation result, which are applicable to scenarios where a single high-risk condition plays a dominant role). Specific rules can be flexibly set and pre-stored according to actual needs.

[0140] According to the above-mentioned preset rules, all the first risk coefficients obtained in step S1022 are calculated, and the final calculation result is the second risk coefficient corresponding to the target object.

[0141] Applying the above embodiments, for each target object, a feature model pre-constructed for multiple regional conditions is obtained; the feature model and the currently matched regional conditions are determined as the conditions satisfied by the target region, and a preset first risk coefficient for each determined regional condition is obtained; all obtained first risk coefficients are aggregated according to preset rules to obtain a second risk coefficient corresponding to the target object. By applying this method, the risk coefficient of the target object is accurately quantified through feature model matching, and the risk coefficients of multi-dimensional regional conditions are aggregated, making the assessment results more comprehensive and improving the accuracy and efficiency of risk assessment.

[0142] In one possible embodiment, in order to prevent the risk coefficient from being too large, a preset risk coefficient threshold can be set to truncate the second risk coefficient. The preset coefficient threshold can be set by professional technicians based on their work experience or industry regulations. For example, in this document, the preset coefficient threshold is 2.

[0143] like If the second risk coefficient is greater than 2, then a forced truncation is applied. =2, meaning the second risk coefficient is 2.

[0144] When aggregating all the first risk coefficients, in one possible implementation, aggregation can be performed using methods such as summation, weighted summation, or maximum value, as previously described.

[0145] In another possible embodiment, the first risk coefficient can be aggregated to obtain the second risk coefficient corresponding to the target object, see [link to relevant documentation]. Figure 3 The methods include:

[0146] S101, for each target object appearing in the target area, determine the risk of the target object participating in smuggling based on the object information of the target object, and take it as the first risk corresponding to the target object;

[0147] S1021, For each target object, obtain the feature model pre-built for each of the multiple regional conditions;

[0148] S1022, determine the feature model and the currently matched regional conditions as the conditions that the target region must meet, and obtain the preset first risk coefficient of each determined regional condition;

[0149] S10231, Substitute the obtained first risk coefficient into the preset calculation formula to calculate the second risk coefficient of the target object;

[0150] or,

[0151] S10232, In the preset mapping table, find the second risk coefficient corresponding to the combination of at least one first risk coefficient determined by the current area conditions, and use it as the second risk coefficient corresponding to the target object.

[0152] The preset mapping table includes a second risk coefficient corresponding to various permutations and combinations of the first risk coefficient;

[0153] or,

[0154] S10233, input the determined regional conditions and the obtained first risk coefficient into the large model, and guide the large model to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object;

[0155] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0156] S104, aggregate the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area.

[0157] Understandably, existing risk assessment methods often employ linear superposition or simple threshold judgments, which are prone to assessment distortion due to the overwhelming of risk factors or the lack of obvious coupling effects, failing to accurately depict the true risk level under complex nonlinear correlations. This embodiment, however, avoids simple linear superposition / threshold judgments, instead first quantifying the risk of a single object (first risk), then adjusting it in conjunction with the risk coefficient of the environment (second risk), and finally aggregating them to obtain the final risk (third risk). This reduces assessment distortion and thus solves the technical problem.

[0158] Steps S10231-S10233 are detailed steps of the aforementioned step S1023. Steps S101, S1021, S1022, S103 and S104 have been explained in the preceding text and will not be repeated here.

[0159] In step S10231, the preset calculation formula is a mathematical expression (such as a linear formula or a weighted summation formula) that pre-sets the correlation between the first risk coefficient and the second risk coefficient based on factors such as the target area scenario and risk type. That is, by using the preset, fixed mathematical calculation formula, the first risk coefficient of the target object is directly substituted into the calculation to obtain the second risk coefficient corresponding to the target object.

[0160] In step S10232, the preset mapping table is a list of all possible combinations of first risk coefficients (including single first risk coefficients and combinations of multiple first risk coefficients) compiled in advance based on a large amount of historical data, scenario testing, and expert experience, along with the second risk coefficient corresponding to each combination. Once at least one first risk coefficient determined by the current regional conditions of the target object is obtained, the second risk coefficient corresponding to the target object can be obtained by looking up the corresponding combination of first risk coefficients in the mapping table.

[0161] For example, see Table 1:

[0162] Table 1

[0163] The first risk combination currently obtained is "heavy fog + heavy rain + night". By querying the mapping table in Table 1, the second risk coefficient corresponding to the target object is determined to be 1.6. It should be understood that Table 1 is merely an example; users can exhaustively explore various combinations of solutions based on their actual circumstances.

[0164] In step S10233, instead of relying on a fixed formula or preset mapping table, the determined regional conditions of the target object and the obtained first risk coefficient are input into the trained large model. At the same time, the large model is guided to combine the scene of the target area, the type of regional conditions, the brightness of the current scene and other multi-dimensional factors to conduct comprehensive analysis and reasoning, and finally output the second risk coefficient corresponding to the target object.

[0165] The scene refers to the actual environment of the target area where the target object is located, including but not limited to the sea, roads, outdoor venues, indoor spaces, etc. The types of regional conditions include, but are not limited to, severe weather such as heavy fog, large waves, strong winds, heavy rain, heavy snow, hail, strong convection, and insufficient visibility, as well as good weather and environmental anomalies such as no fog and calm seas. The brightness of the current scene includes different lighting scenes such as daytime, nighttime, dusk, dawn, cloudy days, and strong light.

[0166] For example, the target area is location A at sea, the area condition is strong wind, and the current scene brightness is 11 PM, with dim lighting. The determined area condition is "strong wind," and the first risk coefficient obtained is 1.5 for "strong wind."

[0167] The large model combines the input information to make inferences: at location A at sea, the risk of smuggling is inherently high in windy weather; at the same time, the poor lighting at night makes it impossible to accurately identify smuggling activities, further amplifying the risk. Considering both factors, the second risk coefficient needs to be appropriately increased based on the first risk coefficient.

[0168] For example, the target area is location B at sea, the area condition type is fog-free, and the current scene brightness is daytime at 8 AM with sufficient light. The determined area conditions are "fog-free" and "daytime," and the obtained first risk coefficients are 0.7 for "fog-free" and 0.9 for "daytime."

[0169] The large model combines the input information to reason: at location B at sea, visibility is good in the absence of fog, the risk of smuggling is low, and there is sufficient light during the day, which further reduces the risk. Considering both factors, the second risk coefficient needs to be appropriately reduced based on the first risk coefficient, and the second risk coefficient is finally determined to be 0.5.

[0170] Applying the above embodiments, the obtained first risk coefficient is substituted into a preset calculation formula to calculate the second risk coefficient of the target object; alternatively, in a preset mapping table, the second risk coefficient corresponding to a combination of at least one first risk coefficient determined by the current regional conditions is searched and used as the second risk coefficient corresponding to the target object; wherein, the preset mapping table includes second risk coefficients corresponding to various permutations and combinations of first risk coefficients; alternatively, the determined regional conditions and the obtained first risk coefficient are input into a large model, and the large model is guided to comprehensively consider the scene of the target area, the type of the determined regional conditions, and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object. Using this method, the calculation method can be flexibly selected according to the complexity of the scene, balancing computational efficiency and result accuracy, thus improving the comprehensiveness of risk coefficient calculation.

[0171] In one possible embodiment, the second risk coefficient corresponding to the target object can also be obtained by the following method, see [link to relevant documentation]. Figure 4 The methods include:

[0172] S101, for each target object appearing in the target area, determine the risk of the target object participating in smuggling based on the object information of the target object, and take it as the first risk corresponding to the target object;

[0173] S1024, For each target object, input the regional information of the region where the target area is located when the target object appears in the target area into the large model, so as to guide the large model to think about the regional conditions that the region meets based on the scenario of the target area, and to think about the second risk coefficient corresponding to the target object based on the regional conditions obtained and the first risk coefficient of the regional conditions.

[0174] or,

[0175] S1025, For each target object, input the regional information of the target area when the target object appears in the target area into the large model, and guide the large model to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object;

[0176] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0177] S104, aggregate the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area.

[0178] Steps S1024 and S1025 are detailed steps of the aforementioned step S102. Steps S101, S103 and S104 have been explained in the preceding text and will not be repeated here.

[0179] In step S1024, the area information may include, but is not limited to: area type (such as remote alley, intersection, school area, etc.), area environmental characteristics (such as area weather, no surveillance, slippery road surface, sparse population, fire, etc.), and surrounding facilities (such as no lighting, no emergency equipment, etc.).

[0180] After the regional information is input into the large model, the model analyzes the regional conditions that might induce the target to engage in smuggling activities within that target area. Once the regional conditions are determined, the model combines these conditions with a pre-set first risk coefficient to further consider their impact on the current target, ultimately outputting a second risk coefficient for the target. This second risk coefficient is a modified value of the first risk coefficient, taking into account the target's specific circumstances, and thus more accurately reflects the actual risk level of the target in real-world scenarios.

[0181] In step S1025, the region information is input into the large model, guiding the model to calculate the second risk coefficient corresponding to the target object based on the scene of the target region, the type of the determined region conditions, and / or the brightness of the current scene. The scene of the target region, the type of region conditions, and the brightness of the current scene have been explained above and will not be repeated here.

[0182] Applying the above embodiments, for each target object, the regional information of the target area when the target object appears in the target area is input into the large model. This guides the large model to consider the scene of the target area to determine the regional conditions that the area must meet, and then to determine the second risk coefficient corresponding to the target object based on the determined regional conditions and the first risk coefficient of the regional conditions. Alternatively, for each target object, the regional information of the target area when the target object appears in the target area is input into the large model, and the large model is guided to comprehensively consider the scene of the target area, the type of determined regional conditions, and / or the brightness of the current scene to determine the second risk coefficient corresponding to the target object. By applying this method, by inputting the regional information of the target area into the large model and guiding it to comprehensively analyze the scene, regional conditions, and risk coefficient, the second risk coefficient corresponding to the target object can be obtained more accurately and efficiently.

[0183] When aggregating the second risks corresponding to each target object, the third risk of smuggling occurring in the target area can be obtained through the following method, see [link to relevant documentation]. Figure 5 The methods include:

[0184] S101, for each target object appearing in the target area, determine the risk of the target object participating in smuggling based on the object information of the target object, and take it as the first risk corresponding to the target object;

[0185] S102, for each target object, determine at least one regional condition that the target region must satisfy when the target object appears in the target region from multiple preset regional conditions, and aggregate the first risk coefficients corresponding to all the determined regional conditions to obtain the second risk coefficient corresponding to the target object.

[0186] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0187] S1041, For each category, aggregate the second risk corresponding to each target object of the category, and take the smaller of the aggregation result and the preset cutoff threshold as the fourth risk corresponding to the category;

[0188] S1042, aggregate the fourth risks corresponding to each category to obtain the third risk, which is the risk of smuggling occurring in the target area.

[0189] Existing smuggling risk assessment methods are often limited to single-object dimensions such as vehicles and ships, failing to establish risk correlation models among multiple elements such as ships, vehicles, personnel, and intelligence. This results in a one-sided assessment perspective, unable to reflect the overall and coordinated nature of smuggling activities. Furthermore, they lack the ability to model the real-time fusion of multi-source information and the dynamic evolution of risks, making the assessment process static and ill-suited to the rapidly changing realities of smuggling methods. Even within a single dimension, the features relied upon are relatively simple. For example, ship-related smuggling risk assessment heavily relies on location data such as Automatic Identification System (AIS) codes, but actual illegal vessels often disable AIS or tamper with location information, causing such methods to fail in critical scenarios, leading to difficulties in detection and identification, and a high false alarm rate.

[0190] In this embodiment, the environmental conditions of the target area when the target object appears are determined in real time, and a first risk coefficient is calculated. Subsequent second and third risks are based on the real-time acquired conditions of the object and the target area, rather than static thresholds, adapting to scenarios where smuggling methods change rapidly. Furthermore, this method abandons the limitation of relying solely on location data (such as AIS codes), taking "object information of the target object" as the core, while integrating the judgment of the appearing regional conditions. This avoids the problems of difficult identification and high false alarm rate caused by the failure of a single feature (such as AIS being turned off), broadens the feature coverage, and thus solves the three technical problems mentioned above.

[0191] Steps S1041 and S1042 are detailed steps of the aforementioned step S104. Steps S101-S103 have been explained in the preceding text and will not be repeated here.

[0192] In step S1041, the category is the classification method of the target object. For example, the target object can be classified by object type, such as small vehicles, large trucks, passenger vehicles, pedestrians, and ships; the target object can be classified by origin, taking vehicles as an example, such as local vehicles, out-of-town vehicles, and overseas vehicles; the target object can be classified by behavioral characteristics, again taking vehicles as an example, such as vehicles passing normally, vehicles stopping abnormally, and vehicles taking multiple detours.

[0193] The second risk of all target objects within the same category is aggregated, and the aggregation result is truncated according to a preset truncation threshold. This step employs a subjective-objective game-theoretic weighting method. Specifically, the aggregation result of the second risk for each category is confirmed using the subjective-objective game-theoretic weighting method. The aggregation result is compared with the preset truncation threshold to determine whether truncation is necessary. The resulting risk value is the fourth risk corresponding to the category.

[0194] For example, to prevent the risk scores corresponding to a category from growing indefinitely, an upper limit will be set for the category weights, i.e., a preset cutoff threshold. For example, the maximum category weights for ships, vehicles, and personnel are 0.35, 0.25, and 0.25, respectively, i.e., ships ≤ 35 points, vehicles ≤ 25 points, and personnel ≤ 25 points. That is, the preset cutoff thresholds are 35, 25, and 25, respectively. Scores exceeding the upper limit will be forcibly cut off to ensure that the risk of smuggling in the final target area is always kept within the range of [0, 100]. See below for details.

[0195] In step S1042, the fourth risks corresponding to each category are aggregated to obtain the third risk, which is the risk of smuggling occurring in the target area.

[0196] The third risk is calculated using the following formula (1):

[0197] ...Formula (1)

[0198] in, As the third risk, For weighted risk, R represents the fourth risk corresponding to category i. The weight of the fourth risk corresponding to category i. This is the attenuation coefficient, with a value of 0.8.

[0199] In one possible implementation, the preset cutoff threshold is determined through the following steps, see [link to relevant documentation] Figure 6 The methods include:

[0200] S601, obtain object information for each sample object.

[0201] The sample objects are those that appear at smuggling sites multiple times within a preset period.

[0202] S602, for each category, the degree of variation among the object information of the sample objects of the category is statistically obtained, and used as the degree of variation corresponding to the category.

[0203] S603: For each category, the smuggling proportion corresponding to the category is calculated based on the frequency of occurrence of sample objects of the category at the smuggling site.

[0204] S604, based on the degree of variation and the proportion of smuggling of each category, determines the objective weight of each category.

[0205] Among them, the objective weight is negatively correlated with the degree of variation, and the objective weight is positively correlated with the proportion of smuggling.

[0206] S605, the objective weights of each category and the preset subjective weights of each category are weighted and summed to obtain the comprehensive weight.

[0207] S606, determine the preset cutoff threshold based on the comprehensive weight.

[0208] In step S601, object information of objects that appear multiple times at the smuggling site within a preset period is obtained, that is, object information of objects at historical smuggling sites is obtained.

[0209] In step S602, the degree of variability refers to the magnitude of information differences among different sample objects within the same category. For example, in the "freight vehicle category," if all sample vehicles have highly consistent models and driving trajectories, it indicates a low degree of variability, suggesting that the behavior patterns of this category are relatively fixed. Conversely, if the sample vehicles exhibit significant differences in models and trajectories, it indicates a high degree of variability, suggesting that the behavior patterns of this category are more flexible and uncertain. The degree of variability can be obtained using common statistical methods such as standard deviation, variance, and coefficient of variation.

[0210] In step S603, for each category, the frequency of the sample objects under that category appearing at the smuggling site is counted, and the smuggling proportion of that category is calculated based on the frequency of appearance. For example, if 1000 historical cases are analyzed and the sample objects under category A appear 500 times, then the smuggling proportion of category A is 50%.

[0211] In step S604, the objective weight is negatively correlated with the degree of variation and positively correlated with the proportion of smuggling.

[0212] In this paper, the negative correlation between objective weight and the degree of variation means that, assuming that other factors affecting objective weight remain constant except for the degree of variation, objective weight decreases monotonically as the degree of variation increases. The monotonous decrease in this paper can refer to strict monotonous decrease or non-strict monotonous decrease.

[0213] In this paper, the positive correlation between objective weight and smuggling ratio means that, assuming that other factors affecting objective weight remain constant except for smuggling ratio, objective weight increases monotonically with the increase of smuggling ratio. This monotonous increase can refer to strict monotonous increase or non-strict monotonous increase.

[0214] The higher the degree of variation, the lower the objective weight; conversely, the lower the degree of variation, the higher the objective weight. Similarly, the higher the proportion of smuggling, the higher the objective weight; and vice versa. In smuggling investigation scenarios, categories with low variation (such as a certain type of freight vehicle whose trajectory and model are highly consistent across multiple smuggling sites) indicate a more stable and predictable correlation with smuggling activities, making them more valuable for risk assessment and thus assigned a higher objective weight. Conversely, categories with high variation (such as a certain type of personnel whose identities and activity trajectories vary greatly across multiple smuggling sites) have an unstable and unpredictable correlation with smuggling activities, making them less valuable for risk assessment and thus assigned a lower objective weight. Categories with a high proportion of smuggling activities indicate a more stable and predictable correlation with smuggling activities, making them more valuable for risk assessment and thus assigned a higher objective weight. Conversely, categories with a low proportion of smuggling have unstable and unpredictable correlations with smuggling activities, and therefore have lower reference value for risk assessment, thus being assigned lower objective weights.

[0215] For example, the smuggling proportion of sample objects in each category and the degree of variation among the object information of the sample objects are calculated using the following formula (2):

[0216] ...Formula (2)

[0217] in, Let k be the information entropy of the j-th category, and k be the normalization coefficient, k = 1 / ln(1000). Let represent the smuggling proportion of the i-th event under the j-th category, and 1000 represent the total number of smuggling events.

[0218] Based on this, the objective weight is calculated using the following formula (3):

[0219] ...Formula (3)

[0220] in, For objective weighting, Let be the information entropy of the j-th category.

[0221] In step S605, the preset subjective weights are determined using the Analytic Hierarchy Process (AHP). A judgment matrix is ​​constructed by technical professionals to ensure that the Consistency Ratio (CR) is less than 0.1, thus obtaining the preset subjective weights for each category. These preset subjective weights can be configured according to current needs. For example, large trucks are more likely to be involved in smuggling and therefore have a higher weight; ordinary private cars have a lower weight; and objects with abnormal trajectories have a higher weight.

[0222] After obtaining the objective weights and the preset subjective weights, the comprehensive weights are calculated using the following formula (4):

[0223] ...Formula (4)

[0224] Where W is the overall weight, and WAHP is the preset subjective weight. For objective weighting.

[0225] In step S606, a preset cutoff threshold is determined based on the comprehensive weight, and the preset cutoff threshold is positively correlated with the comprehensive weight.

[0226] Using the above embodiments, object information of each sample object is obtained; wherein, the sample object is an object that has appeared at the most recent smuggling sites; for each category, the degree of variation among the object information of the sample objects in the category is statistically obtained as the degree of variation corresponding to the category; for each category, the smuggling proportion corresponding to the category is calculated based on the frequency of occurrence of the sample objects of the category at the smuggling sites; based on the degree of variation and the smuggling proportion corresponding to each category, the objective weight of each category is determined; wherein, the objective weight is negatively correlated with the degree of variation and positively correlated with the smuggling proportion; the objective weight of each category is weighted and summed with the preset subjective weight of each category to obtain the comprehensive weight; a preset cutoff threshold is determined based on the comprehensive weight, wherein, the preset cutoff threshold is positively correlated with the comprehensive weight. By applying this method, objective weights are determined by combining the objective variability of sample object information at the smuggling site with the smuggling proportion corresponding to the frequency of sample object occurrence. These objective weights complement the preset subjective weights, dynamically determining the comprehensive weights. Based on the comprehensive weights, preset cutoff thresholds are determined, avoiding the one-sidedness and irrationality caused by purely subjective weight setting. This improves the accuracy and rationality of smuggling transshipment risk assessment, making it more closely aligned with actual smuggling scenarios.

[0227] In another possible embodiment, when the first risk corresponding to the target object is determined, the object information of the object can be inferred by an inference model to obtain the inference result. Based on the inference result, the risk of the target object participating in smuggling is determined as the first risk corresponding to the target object. For details, see [link to relevant documentation]. Figure 7 The methods include:

[0228] S1011, for each target object that appears in the target area during the target time period, use at least one inference model to infer the object information of the target object, and obtain at least one inference result of the target object and the confidence level of each inference result;

[0229] The target period is any time period, which is a pre-set time range for smuggling risk monitoring and assessment, such as a certain day, a certain week, or a certain specific time period.

[0230] S1012, Based on all the reasoning results of the target object, determine the risk of the target object participating in smuggling, which is the first risk corresponding to the target object;

[0231] S102, for each target object, determine at least one regional condition that the target region must satisfy when the target object appears in the target region from multiple preset regional conditions, and aggregate the first risk coefficients corresponding to all the determined regional conditions to obtain the second risk coefficient corresponding to the target object.

[0232] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0233] S1043, aggregate the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area within the target time period.

[0234] Steps S1011 and S1012 are detailed steps of step S101, and step S1043 is a detailed step of step S104. Steps S102-S103 have been explained in the preceding text and will not be repeated here.

[0235] In steps S1011 and S1012, the reasoning model is any model that can infer whether an object to which the object information belongs is involved in smuggling based on the object information, such as a Bayesian network model, a neural network model, etc.

[0236] The inference model analyzes information about the target entity and outputs at least one inference conclusion regarding whether it is involved in smuggling, i.e., the inference result. It also provides the credibility level of each inference result, i.e., the confidence score. By synthesizing all inference results, the risk level / probability of the target entity's involvement in smuggling is determined, i.e., the first risk.

[0237] Based on the above steps, the confidence level of the aforementioned third risk can be calculated according to the confidence level of each reasoning result, and finally the third risk itself and its confidence level are presented accordingly.

[0238] By applying the above embodiments, at least one inference model is used to infer the object information of the target object, obtaining at least one inference result for the target object and the confidence level of each inference result. Based on all the inference results for the target object, the risk of the target object participating in smuggling is determined as the first risk corresponding to the target object. Based on the confidence level of each inference result, the confidence level of the fourth risk is determined, and the fourth risk and its confidence level are displayed accordingly. By applying this method, combining multiple inference results and confidence levels to determine smuggling risks, the credibility of risks is quantified and displayed intuitively, improving the accuracy and transparency of smuggling risk assessment, and facilitating post-audit and error tracing.

[0239] When displaying the third risk and its confidence level, users can modify the inference results and optimize the inference model based on the modified results. For details, see [link to documentation]. Figure 8 The methods include:

[0240] S801 corresponds to displaying the third risk, the confidence level of the third risk, and the reasoning results for each target object.

[0241] S802, in response to the correction of the reasoning result, determines the object information of the object to which the corrected reasoning result belongs, and reasoning to obtain the reasoning model of the corrected reasoning result.

[0242] S803 trains the determined inference model based on the corrected inference results and the determined object information.

[0243] In steps S801-S803, when displaying the third risk, it can be achieved by displaying risk values, risk levels, etc. Taking the display of risk levels as an example, a four-level response system can be established. Specifically, green (0-29 points) corresponds to routine monitoring and daily patrols; yellow (30-59 points) corresponds to enhanced patrols and hourly video patrols by drones; orange (60-79 points) corresponds to key deployment, with regional management departments on standby and drones collecting evidence at specific locations for the incident; and red (80-100 points) corresponds to emergency response, cross-departmental incident reporting, and immediate dispatch of police.

[0244] The system simultaneously displays the confidence level corresponding to the third risk, that is, it demonstrates the credibility of the inference model's risk assessment results, while also showing the inference results for each target object. Furthermore, it can display the risk factor identification results, various risk scores, weight allocations, and the impact of risk coefficients in the smuggling and transshipment risk assessment, providing a basis for subsequent judgments and corrections.

[0245] When users / managers correct the inference results displayed in step S801, such as correcting false alarms, missed alarms, risk level judgment biases, or unreasonable confidence levels, the system responds to the correction operation. The system automatically identifies and determines the object information of the target object (such as area, equipment, personnel, event target, etc.) to which the corrected inference result belongs, and generates the inference model used for that corrected inference result. Using the manually corrected inference result as a real sample, the corresponding inference model determined in step S802 is trained to update the inference model. This enables the inference model to more accurately obtain at least one inference result for the target object and the confidence level of each inference result in subsequent risk assessments.

[0246] To further improve the accuracy of smuggling and transshipment risk assessment, in one possible embodiment, the smuggling and transshipment risk assessment can be conducted by combining intelligence containing object information of potential smuggling targets. Specifically, this includes the following steps:

[0247] Step 1: Obtain object information of each object collected by the sensors set in the target area during the target time period, as well as the first human intelligence input for the target time period and target area.

[0248] The first type of human intelligence includes object information of potential smuggling targets.

[0249] Step 2: Identify objects whose object information meets the preset anomaly rules and objects whose object information is associated with the object information of potential smuggling targets as target objects.

[0250] In step 1 above, the sensor collects data in real time on people, vehicles, goods, ships, and other objects that pass by or stay within a specified target time period, and obtains basic data of these objects, such as: location, trajectory, speed, dwell time, appearance characteristics, identity information, cargo information, etc., as object information for each object.

[0251] Sensors include, but are not limited to, cameras, radar, radio frequency identification (RFID), vehicle identification, and trajectory monitoring equipment.

[0252] The first human intelligence is information proactively input by personnel (such as anti-smuggling departments and intelligence analysts) based on experience, clues, reports, and historical cases, and is relevant to the target time period and target area. This first human intelligence should at least contain information about potential smuggling targets suspected of being involved in smuggling activities, such as: the identity of suspicious persons, license plates of suspicious vehicles, characteristics of suspicious goods, and information about suspicious vessels.

[0253] In step 2 above, the preset exception rules can be set by professional and technical personnel based on their work experience or industry regulations.

[0254] Specifically, if an object's information (trajectory, behavior, attributes, time, location, etc.) directly violates preset abnormal rules, such as deviating from the normal flight path / lane, staying for an extended period of time without reason, mismatched / obscured / disguised identity information, or activity in prohibited areas, such objects can be directly identified as target objects without relying on first-hand human intelligence.

[0255] Objects associated with potential smuggling targets are not necessarily obviously abnormal, but their information is related to potential smuggling targets in the first human intelligence. For example, people traveling with suspicious persons, vehicles traveling with suspicious vehicles, goods mixed with suspicious goods, vehicles connecting with suspicious goods, etc. These objects are related suspicious objects identified based on the first human intelligence.

[0256] Using the above embodiments, object information of each object collected by sensors set in the target area during the target time period is obtained, along with first human-generated intelligence input for the target time period and target area. The first human-generated intelligence includes object information of potential smuggling targets. Objects whose object information meets preset anomaly rules, and objects whose object information is associated with the object information of potential smuggling targets, are identified as target objects. This method, by combining objective sensor data collection with human-generated intelligence, enriches the risk assessment data source, enabling dual screening of target objects from both abnormal behavior and intelligence correlation perspectives. This improves the accuracy of suspicious target identification, reduces the false positive rate, and enhances the efficiency and accuracy of smuggling and transshipment risk assessment.

[0257] To further improve the accuracy of smuggling and transshipment risk assessment, in another possible embodiment, the smuggling and transshipment risk assessment can also be conducted by combining intelligence describing potential smuggling activities and the expected timeframes for such activities. For details, see [link to relevant documentation]. Figure 9 The methods include:

[0258] S901, acquire each second human intelligence input for the target area.

[0259] The second type of human intelligence is used to describe potential smuggling activities and the expected timeframe for such activities.

[0260] S902, among the potential smuggling activities described in each second human intelligence report, identify potential smuggling activities whose expected occurrence time includes the target time period, and designate them as target potential smuggling activities.

[0261] S903, based on the potential smuggling activities of each target, determines the risk of smuggling activities occurring in the target area within the target time period, as the fifth risk.

[0262] S904 aggregates the second and fifth risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area within the target time period.

[0263] In step S901 above, similar to the first human intelligence, the second human intelligence is actively input by humans (such as anti-smuggling departments or intelligence analysts) based on experience, clues, reports, historical cases, etc., and is related to the target area. This second human intelligence is used to describe potential smuggling activities and the expected time period of such activities.

[0264] In step S902 above, among the potential smuggling activities described in the second human intelligence, potential smuggling activities whose expected occurrence time includes the target time period are selected as target potential smuggling activities. That is, potential smuggling activities whose "expected occurrence time" covers the "target time period" to be assessed are selected as target potential smuggling activities to be used in this smuggling and transshipment risk assessment. There are one or more target potential smuggling activities.

[0265] In step S903 above, based on the identified potential smuggling activities, the risk of smuggling activities occurring in the target area during the target time period is determined as the fifth risk.

[0266] Specifically, when there is only one potential smuggling activity, the risk of that potential smuggling activity occurring in the target area within the target time period is defined as the fifth risk. When there are multiple potential smuggling activities, the maximum or average risk of each potential smuggling activity occurring in the target area within the target time period is defined as the fifth risk, without further specific limitations.

[0267] In step S904 above, the second and fifth risks corresponding to each target object are aggregated to obtain the third risk of smuggling occurring in the target area during the target time period.

[0268] The weight of the fifth risk is the weight of intelligence. Similarly, in order to prevent the accumulated score from growing indefinitely, an upper limit will be set for this weight. For example, the maximum weight of intelligence is 0.15, that is, intelligence ≤ 15 points.

[0269] Applying the above embodiments, second human-generated intelligence is obtained for each target area. This second human-generated intelligence describes potential smuggling activities and their expected occurrence time periods. Among the potential smuggling activities described in each second human-generated intelligence, those whose expected occurrence time includes the target time period are identified as target potential smuggling activities. Based on each target potential smuggling activity, the risk of smuggling occurring in the target area within the target time period is determined as the fifth risk. The second and fifth risks corresponding to each target are aggregated to obtain the third risk of smuggling occurring in the target area within the target time period. By introducing human-generated intelligence into risk assessment and performing time-matching filtering on the intelligence, this method effectively eliminates interference from irrelevant information. The fifth risk is determined based on human judgment, and the third risk is obtained based on the second and fifth risks, enriching the dimensions of risk assessment and thus improving the accuracy of smuggling risk assessment.

[0270] After obtaining the third risk, in order to keep the risk of smuggling in the target area within the final target time period always within the range of [0,100], the risk of smuggling in the target area within the final target time period can be obtained by truncating the third risk.

[0271] In one possible implementation, the first risk corresponding to the target object can be determined in the following manner, see [reference needed]. Figure 10 The methods include:

[0272] S1013, for each target object that appears in the target area during the target time period, determine the category to which the target object belongs based on the object information of the target object, and the preset exception rules triggered by the category to which it belongs;

[0273] S1014, Determine the risk of the target object participating in smuggling based on the preset abnormal rules triggered by the category to which it belongs, and use this as the first risk corresponding to the target object;

[0274] S102, for each target object, determine at least one regional condition that the target region must satisfy when the target object appears in the target region from multiple preset regional conditions, and aggregate the first risk coefficients corresponding to all the determined regional conditions to obtain the second risk coefficient corresponding to the target object.

[0275] S103, For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0276] S1043, aggregate the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area within the target time period.

[0277] Steps S1013 and S1014 are detailed steps of the aforementioned step S101. Steps S102-S1043 have been explained in the preceding text and will not be repeated here.

[0278] In step S1013, the preset anomaly rules refer to a set of rules pre-defined for each object category to determine whether a target object of that category is suspected of smuggling. Specifically, these rules can be set based on historical data of the target object's past involvement in smuggling activities, characteristics of smuggling activities, etc.

[0279] In step S1014, each preset exception rule corresponds to a risk, that is, there is a correspondence between object category, preset exception rule and risk.

[0280] For example, if the target object is a vehicle, the preset exception rule triggered is "the load weight deviates from the declared cargo weight by more than a preset ratio", and the corresponding risk is 0.4. Then 0.4 is the first risk corresponding to the target object. If the preset exception rule triggered is "the driving route deviates from the preset legal route", the corresponding risk is 0.3. Then 0.3 is the first risk corresponding to the target object.

[0281] Applying the above embodiments, the category to which the target object belongs is determined based on the object information of the target object, as well as the preset anomaly rules triggered by the category. The risk of the target object participating in smuggling is determined based on the preset anomaly rules triggered by the category, serving as the first risk corresponding to the target object. This method enables refined risk differentiation for different categories of target objects, improving the targeting and accuracy of smuggling risk identification, avoiding misjudgments or omissions caused by rule generalization, and thus improving the accuracy of risk assessment.

[0282] The following section will explain the entire process of the smuggling and transshipment risk assessment method using a flowchart. (See attached document.) Figure 11 The methods include:

[0283] S1101, Multi-source dynamic data input; wherein, the multi-source dynamic data includes the aforementioned object information, object category, environmental information, and human intelligence.

[0284] S1102, Dynamic feature extraction and data fusion; that is, equivalent to the aforementioned step S101;

[0285] S1103, bounded weighted aggregation evaluation of impact factors; that is, equivalent to the aforementioned steps S102-S103.

[0286] S1104, Regional comprehensive risk quantification output; that is, equivalent to the aforementioned step S104.

[0287] S1105, based on the dynamic optimization model of handling feedback; that is, equivalent to the aforementioned steps S801-S803.

[0288] Corresponding to the aforementioned smuggling and transshipment risk assessment method, this embodiment of the invention also provides a smuggling and transshipment risk assessment device, see [link to relevant documentation]. Figure 12 The device includes:

[0289] The first determining module 1201 is used to determine the risk of the target object participating in smuggling based on the object information of each target object that appears in the target area during the target time period, and to use this as the first risk corresponding to the target object.

[0290] The second determining module 1202 is used to determine, for each target object, at least one regional condition that the target region satisfies when the target object appears in the target region from multiple preset regional conditions, and to aggregate the first risk coefficients corresponding to all the determined regional conditions to obtain the second risk coefficient corresponding to the target object.

[0291] The first scaling module 1203 is used to scale the first risk corresponding to the target object for each target object using the second risk coefficient corresponding to the target object, so as to obtain the scaled second risk corresponding to the target object.

[0292] The first aggregation module 1204 is used to aggregate the second risks corresponding to each of the target objects to obtain the third risk of smuggling occurring in the target area during the target time period.

[0293] By applying the above embodiments, the risk coefficient can be dynamically adjusted according to the regional conditions of the target area when the target object appears in the target area, so as to improve the accuracy of the determined second risk coefficient. This makes the scaled risk coefficient more closely match the real smuggling scenario, and thus can more accurately assess the risk of smuggling in the target area.

[0294] In one possible implementation,

[0295] The second determining module includes:

[0296] The first determining submodule is used to obtain the feature models pre-constructed for each of the multiple regional conditions;

[0297] The second determining submodule is used to determine the feature model and the currently matched regional conditions as the conditions satisfied by the target region, and to obtain the preset first risk coefficient of each determined regional condition;

[0298] The third determining submodule is used to aggregate all the obtained first risk coefficients according to preset rules to obtain the second risk coefficient corresponding to the target object.

[0299] In one possible implementation, the third determining submodule includes:

[0300] The first determining unit is used to input the acquired first risk coefficient into a preset calculation formula to calculate the second risk coefficient of the target object; or...

[0301] The second determining unit is configured to search in a preset mapping table for a second risk coefficient corresponding to a combination of at least one first risk coefficient determined by the current area conditions, and use this second risk coefficient as the second risk coefficient corresponding to the target object. The preset mapping table includes second risk coefficients corresponding to various permutations and combinations of the first risk coefficients; or...

[0302] The third determining unit is used to input the determined regional conditions and the obtained first risk coefficient into the large model, and guide the large model to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

[0303] In one possible implementation, the second determining module includes:

[0304] The fourth determination submodule is used to input the regional information of the target area when the target object appears in the target area into the large model, so as to guide the large model to think about the regional conditions satisfied by the target area based on the scenario of the target area, and to think about the second risk coefficient corresponding to the target object based on the regional conditions obtained and the first risk coefficient of the regional conditions; or,

[0305] The fifth determination submodule is used to input the regional information of the target area when the target object appears in the target area into the large model, and guide the large model to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

[0306] In one possible implementation, the first aggregation module includes:

[0307] The first aggregation submodule is used to aggregate the second risks corresponding to each target object of each category for each category, and take the smaller of the aggregation result and the preset truncation threshold as the fourth risk corresponding to the category.

[0308] The second aggregation submodule is used to aggregate the fourth risks corresponding to each of the categories to obtain the third risk, which serves as the risk of smuggling occurring in the target area during the target time period.

[0309] In one possible implementation, the preset cutoff threshold is determined by the following methods:

[0310] Obtain object information for each sample object, wherein the sample object is an object that appears multiple times at the smuggling site within a preset period;

[0311] For each category, the degree of variation among the object information of the sample objects in that category is statistically calculated, and this degree of variation is used as the degree of variation corresponding to that category.

[0312] For each category, the smuggling proportion corresponding to that category is calculated based on the frequency of occurrence of the sample objects of that category at the smuggling site;

[0313] The objective weight of each category is determined based on the degree of variation and the proportion of smuggling for each category, wherein the objective weight is negatively correlated with the degree of variation and positively correlated with the proportion of smuggling.

[0314] The objective weight of each category is weighted and summed with the preset subjective weight of each category to obtain the comprehensive weight.

[0315] The preset cutoff threshold is determined based on the comprehensive weight, wherein the preset cutoff threshold is positively correlated with the comprehensive weight.

[0316] In one possible implementation, the first determining module includes:

[0317] The fifth determining submodule is used to infer the object information of the target object using at least one inference model, and to obtain at least one inference result of the target object and the confidence level of each inference result;

[0318] The sixth determining submodule is used to determine the risk of the target object participating in smuggling based on all the reasoning results of the target object, as the first risk corresponding to the target object;

[0319] The device further includes:

[0320] The third determining module is used to determine the confidence level of the third risk based on the confidence level of each of the inference results;

[0321] The first display module is used to display the third risk and the confidence level of the third risk.

[0322] In one possible implementation, the first display module includes:

[0323] The first display submodule is used to display the third risk, the confidence level of the third risk, and the reasoning results of each target object.

[0324] The second display submodule is used to, in response to the correction of the reasoning result, determine the object information of the object to which the corrected reasoning result belongs, and reason to obtain the reasoning model of the corrected reasoning result.

[0325] The third display submodule is used to train the determined inference model based on the corrected inference results and the determined object information.

[0326] In one possible implementation, the device further includes:

[0327] The fourth determining module is used to acquire object information of each object collected by sensors installed in the target area during the target time period, and first artificial intelligence input for the target time period and the target area, wherein the first artificial intelligence includes object information of potential smuggling objects; determine objects whose object information meets preset anomaly rules and objects whose object information is associated with the object information of the potential smuggling objects as target objects; and / or,

[0328] The fifth determining module is used to acquire each second piece of human-generated intelligence input for the target area, wherein the second piece of human-generated intelligence describes potential smuggling activities and the expected occurrence time of the potential smuggling activities; among the potential smuggling activities described by each piece of second piece of human-generated intelligence, potential smuggling activities whose expected occurrence time includes the target time period are identified as target potential smuggling activities; based on each target potential smuggling activity, the risk of smuggling activities occurring in the target area during the target time period is determined as the fifth risk; the aggregation of the second risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area during the target time period includes: aggregating the second risks and the fifth risks corresponding to each target object to obtain the third risk of smuggling occurring in the target area during the target time period; and / or,

[0329] The sixth determining module is used to determine the risk of the target object participating in smuggling based on the object information of the target object, as the first risk corresponding to the target object, including: determining the category to which the target object belongs based on the object information of the target object, and the preset abnormal rules triggered by the category; determining the risk of the target object participating in smuggling based on the preset abnormal rules triggered by the category, as the first risk corresponding to the target object.

[0330] This invention also provides an electronic device, such as... Figure 13As shown, it includes a processor 1301, a communication interface 1302, a memory 1303, and a communication bus 1304, wherein the processor 1301, the communication interface 1302, and the memory 1303 communicate with each other through the communication bus 1304.

[0331] Memory 1303 is used to store computer programs;

[0332] When processor 1301 executes a program stored in memory 1303, it performs the following steps:

[0333] For each target object that appears in the target area during the target time period, the risk of the target object participating in smuggling is determined based on the object information of the target object, which is regarded as the first risk corresponding to the target object;

[0334] For each target object, at least one regional condition is determined from multiple preset regional conditions that the target region must satisfy when the target object appears in the target region. The first risk coefficients corresponding to all the determined regional conditions are aggregated to obtain the second risk coefficient corresponding to the target object.

[0335] For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object;

[0336] The second risks corresponding to each of the target objects are aggregated to obtain the third risk of smuggling occurring in the target area during the target time period.

[0337] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0338] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0339] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0340] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0341] In another embodiment of the present invention, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of any of the above-described smuggling and transshipment risk assessment methods.

[0342] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the smuggling and transshipment risk assessment methods described in the above embodiments.

[0343] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape) or an optical medium (e.g., DVD), etc.

[0344] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0345] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0346] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for assessing the risk of smuggling and transshipment, characterized in that, The method includes: For each target object appearing in the target area, the risk of the target object participating in smuggling is determined based on the object information of the target object, which is regarded as the first risk corresponding to the target object; For each target object, at least one regional condition is determined from multiple preset regional conditions that the target region must satisfy when the target object appears in the target region. The first risk coefficients corresponding to all the determined regional conditions are aggregated to obtain the second risk coefficient corresponding to the target object. For each target object, the first risk corresponding to the target object is scaled using the second risk coefficient corresponding to the target object to obtain the scaled second risk corresponding to the target object; The second risks corresponding to each of the target objects are aggregated to obtain the third risk of smuggling occurring in the target area.

2. The method according to claim 1, characterized in that, The step of determining at least one regional condition that the target area must satisfy when the target object appears in a target area from multiple preset regional conditions, and aggregating the first risk coefficients corresponding to all determined regional conditions to obtain a second risk coefficient corresponding to the target object, includes: Obtain the feature models pre-built for each of the multiple regional conditions; The feature model and the currently matched regional conditions are determined as the conditions that the target region satisfies, and the preset first risk coefficient of each determined regional condition is obtained. The first risk coefficients obtained are aggregated according to preset rules to obtain the second risk coefficient corresponding to the target object.

3. The method according to claim 2, characterized in that, The step of aggregating all the obtained first risk coefficients according to preset rules to obtain the second risk coefficient corresponding to the target object includes: The obtained first risk coefficient is substituted into a preset calculation formula to calculate the second risk coefficient of the target object; or... In a preset mapping table, the second risk coefficient corresponding to at least one combination of first risk coefficients determined by the current area conditions is found and used as the second risk coefficient corresponding to the target object. The preset mapping table includes second risk coefficients corresponding to various permutations and combinations of the first risk coefficients; or... The determined regional conditions and the obtained first risk coefficient are input into the large model, and the large model is guided to comprehensively consider the scene of the target area, the type of the determined regional conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

4. The method according to claim 1, characterized in that, The step of determining at least one regional condition that the target area must satisfy when the target object appears in a target area from multiple preset regional conditions, and aggregating the first risk coefficients corresponding to all determined regional conditions to obtain a second risk coefficient corresponding to the target object, includes: The regional information of the target area when the target object appears is input into the large model to guide the large model to deduce the regional conditions satisfied by the target area based on the scenario of the target area, and to deduce the second risk coefficient corresponding to the target object based on the deduced regional conditions and the first risk coefficient of the regional conditions; or... When the target object appears in the target area, the area information of the target area is input into the large model, and the large model is guided to comprehensively consider the scene of the target area, the type of the determined area conditions and / or the brightness of the current scene to obtain the second risk coefficient corresponding to the target object.

5. The method according to claim 1, characterized in that, The aggregation of the second risks corresponding to each of the target objects to obtain the third risk of smuggling occurring in the target area includes: For each category, the second risk corresponding to each target object in the category is aggregated, and the smaller of the aggregation result and the preset cutoff threshold is taken as the fourth risk corresponding to the category. The fourth risks corresponding to each of the aforementioned categories are aggregated to obtain the third risk, which serves as the risk of smuggling occurring in the target area.

6. The method according to claim 5, characterized in that, The preset truncation threshold is determined in the following ways: Obtain object information for each sample object, wherein the sample object is an object that appears multiple times at the smuggling site within a preset period; For each category, the degree of variation among the object information of the sample objects in that category is statistically calculated, and this degree of variation is used as the degree of variation corresponding to that category. For each category, the smuggling proportion corresponding to that category is calculated based on the frequency of occurrence of the sample objects of that category at the smuggling site; The objective weight of each category is determined based on the degree of variation and the proportion of smuggling for each category, wherein the objective weight is negatively correlated with the degree of variation and positively correlated with the proportion of smuggling. The objective weight of each category is weighted and summed with the preset subjective weight of each category to obtain the comprehensive weight. The preset cutoff threshold is determined based on the comprehensive weight, wherein the preset cutoff threshold is positively correlated with the comprehensive weight.

7. The method according to claim 1, characterized in that, The step of determining the risk of the target object participating in smuggling based on the object information of the target object, as the first risk corresponding to the target object, includes: The object information of the target object is inferred using at least one inference model to obtain at least one inference result of the target object and the confidence level of each inference result; Based on all the reasoning results of the target object, the risk of the target object participating in smuggling is determined as the first risk corresponding to the target object; The method further includes: Based on the confidence level of each of the inference results, the confidence level of the third risk is determined; The corresponding third risk and its confidence level are displayed.

8. The method according to claim 7, characterized in that, The corresponding display of the third risk and the confidence level of the third risk includes: The corresponding information includes the third risk, the confidence level of the third risk, and the reasoning results for each target object. In response to the correction of the reasoning result, the object information of the object to which the corrected reasoning result belongs is determined, and the reasoning model of the corrected reasoning result is obtained. The determined inference model is trained based on the corrected inference results and the identified object information.

9. The method according to claim 1, characterized in that, The method further includes: Acquire object information of each object collected by sensors set in the target area, and first human-generated intelligence input for the target time period and the target area, wherein the first human-generated intelligence includes object information of potential smuggling targets; identify objects whose object information meets preset anomaly rules and objects whose object information is associated with the object information of the potential smuggling targets as target objects; and / or, Acquire each second piece of human-generated intelligence input for the target area, wherein the second piece of human-generated intelligence describes potential smuggling activities and the expected time period of the potential smuggling activities; among the potential smuggling activities described by each piece of second human-generated intelligence, find potential smuggling activities whose expected time period includes the target time period, as target potential smuggling activities; based on each target potential smuggling activity, determine the risk of smuggling activities occurring in the target area within the target time period, as a fifth risk; the aggregation of the second risks corresponding to each target object to obtain a third risk of smuggling occurring in the target area within the target time period includes: aggregating the second risks and the fifth risks corresponding to each target object to obtain a third risk of smuggling occurring in the target area within the target time period; and / or, The step of determining the risk of the target object participating in smuggling based on the object information of the target object, as the first risk corresponding to the target object, includes: determining the category to which the target object belongs based on the object information of the target object, and the preset anomaly rule triggered by the category; determining the risk of the target object participating in smuggling based on the preset anomaly rule triggered by the category, as the first risk corresponding to the target object.

10. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the method described in any one of claims 1-9.

Images