Monitoring key image recognition method and system based on drone

Abstract

The present application provides a method and system for identifying key monitoring images based on a drone. The configured image analysis thread is obtained by optimizing the coefficients of the original image analysis thread according to the regression analysis quality index values ​​corresponding to each of several original configuration examples. During the whole process, when configuring the original image analysis thread, the regression analysis of important monitoring image information is cascaded to the original image analysis thread, so that the original image analysis thread can understand the correct information related to the event from the important monitoring image information, and support the generation of monitoring area events, which can avoid the situation where the monitoring area event output produces anomalies to the monitoring area subject, and can obtain an image analysis thread that can output accurate monitoring area events. The generation of accurate monitoring area events can be achieved by inputting the monitoring area subject into the configured image analysis thread.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and more specifically, to a method and system for identifying key surveillance images based on drones. Background Art

[0002] At present, image recognition is a relatively mature technology, but it is relatively difficult to identify key image content in drone cruises because it requires relevant technical personnel to pre-set it. However, there is still a difference between the actual situation and the preset situation, which will interfere with the recognition of important content. Therefore, a technical solution is urgently needed to improve the above technical problems. Summary of the invention

[0003] In order to improve the technical problems existing in the related technologies, the present application provides a method and system for identifying key surveillance images based on drones.

[0004] In a first aspect, a method for identifying surveillance key images based on a drone is provided, comprising:

[0005] Obtaining several original configuration examples for configuring an original image analysis thread; the original configuration examples include an example monitoring area theme, an example monitoring area event, and important monitoring image information describing the content in the example monitoring area theme, the important monitoring image information is different from the example monitoring area event;

[0006] For each of the original configuration examples, inputting the example monitoring area subject in the original configuration example into the original image analysis thread to perform monitoring area event regression analysis and important image information regression analysis, and obtaining regression analysis monitoring area events and regression analysis important image information features;

[0007] Calculate the regression analysis quality index according to the example monitoring area event, the regression analysis monitoring area event, the monitoring important image information and the regression analysis important image information features to obtain the regression analysis quality index value corresponding to the original configuration example;

[0008] According to the regression analysis quality index values ​​corresponding to each of the several original configuration examples, the coefficients of the original image analysis thread are optimized to obtain a configured image analysis thread; the configured image analysis thread is used to output monitoring area events based on the input monitoring area theme.

[0009] In an independently implemented embodiment, the regression analysis quality index calculation is performed according to the example monitoring area event, the regression analysis monitoring area event, the monitoring important image information and the regression analysis important image information feature to obtain the regression analysis quality index value corresponding to the original configuration example, including:

[0010] Calculate the event regression analysis quality index value according to the example monitoring area event and the regression analysis monitoring area event, and calculate the important image information regression analysis quality index value according to the monitoring important image information and the regression analysis important image information features;

[0011] The event regression analysis quality index value and the important image information regression analysis quality index value are aggregated to obtain the regression analysis quality index value corresponding to the original configuration example.

[0012] In an independently implemented embodiment, the calculating the event regression analysis quality index value according to the example monitoring area event and the regression analysis monitoring area event includes:

[0013] Compressing each pixel in the example monitoring area event respectively to obtain pixel compression corresponding to each pixel in the example monitoring area event, and determining the regression analysis possibility corresponding to each pixel in the regression analysis monitoring area event;

[0014] For each pixel in the example monitoring area event, the regression analysis quality index value corresponding to the pixel is calculated according to the pixel compression corresponding to the pixel and the regression analysis possibility corresponding to the pixel at the same position in the regression analysis monitoring area event;

[0015] According to the regression analysis quality index value corresponding to each pixel in the example monitoring area event, the event regression analysis quality index value is obtained.

[0016] In an independently implemented embodiment, the calculating the important image information regression analysis quality index value according to the monitored important image information and the regression analysis important image information features includes:

[0017] Compressing the important monitoring image information to obtain important monitoring image information features corresponding to the important monitoring image information;

[0018] The commonality weight quality index value between the monitoring important image information feature and the regression analysis important image information feature is calculated, and the commonality weight quality index value is used as the important image information regression analysis quality index value.

[0019] In an independently implemented embodiment, the important monitoring image information is a monitoring area thing relationship; the monitoring area thing relationship is obtained by:

[0020] Counting the occurrence frequencies of monitoring area things appearing in the example monitoring area theme, determining a number of monitoring area things, and the occurrence frequencies of each of the several monitoring area things;

[0021] Extracting at least two key objects from the plurality of monitoring area objects according to the respective occurrence frequencies of the plurality of monitoring area objects;

[0022] Determine a key thing relationship between any two of the at least two key things according to the example monitoring area theme;

[0023] Based on the key thing relationship between any two key things among the at least two key things, a monitoring area thing relationship describing the content in the example monitoring area theme is determined.

[0024] In an independently implemented embodiment, determining the key thing relationship between any two key things among the at least two key things according to the example monitoring area theme includes:

[0025] Taking every two key things among the at least two key things as a group, a plurality of key thing binary groups are obtained;

[0026] For each key thing binary group, a regression analysis is performed on the relationship between the two key things in the key thing binary group according to the example monitoring area theme to obtain the key thing relationship between the two key things in the key thing binary group.

[0027] In an independently implemented embodiment, determining the monitoring area thing relationship describing the content in the example monitoring area theme based on the key thing relationship between any two key things among the at least two key things includes:

[0028] When the number of the at least two key things is two, the key thing relationship between the two key things is used as the monitoring area thing relationship that describes the content in the example monitoring area theme.

[0029] In an independently implemented embodiment, determining the monitoring area thing relationship describing the content in the example monitoring area theme based on the key thing relationship between any two key things among the at least two key things includes:

[0030] When the number of the at least two key items is greater than two, determining the two key items with the highest occurrence frequencies according to the respective occurrence frequencies of the at least two key items;

[0031] The key thing relationship of the two key things with the highest occurrence frequency is used as the monitoring area thing relationship for describing the content in the example monitoring area theme.

[0032] In an independently implemented embodiment, the coefficients of the original image analysis thread are optimized according to the regression analysis quality index values ​​corresponding to the several original configuration examples to obtain the configured image analysis thread, including:

[0033] According to the regression analysis quality index values ​​corresponding to the several original configuration examples, the coefficients of the original image analysis thread are optimized to obtain a configured original image analysis thread;

[0034] Obtaining an event generation reference thread; the event generation reference thread includes a classification instruction header and a network structure of the original image analysis thread, and is obtained by configuring the plurality of original configuration examples;

[0035] Obtaining several enhanced configuration examples, for each enhanced configuration example, using the example monitoring area theme in the enhanced configuration example, the selected monitoring area events and the unselected monitoring area events corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme, configuring the configured original image analysis thread and the event generation reference thread, and obtaining the enhanced configuration quality indicator value corresponding to the enhanced configuration example;

[0036] According to the enhanced configuration quality indicator values ​​corresponding to each of the plurality of enhanced configuration examples, the coefficients of the configured original image analysis thread are optimized to obtain the configured image analysis thread.

[0037] In an independently implemented embodiment, for each enhanced configuration example, the configured original image analysis thread and the event generation reference thread are configured using the example monitoring area theme in the enhanced configuration example, the selected monitoring area events and the unselected monitoring area events corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme, and the enhanced configuration quality index value corresponding to the enhanced configuration example is obtained, including:

[0038] For each enhanced configuration example, inputting the example monitoring area theme in the enhanced configuration example into the configured raw image analysis thread, determining the configured raw image analysis thread outputting a first possibility of the selected monitoring area event, a second possibility of the unselected monitoring area event, and a third possibility of the important monitoring image information;

[0039] Inputting the example monitoring area theme in the enhanced configuration example into the event generation reference thread, determining that the event generation reference thread outputs a fourth possibility of the selected monitoring area event, a fifth possibility of the unselected monitoring area event, and a sixth possibility of the important monitoring image information;

[0040] Determine an output event comparison quality index value according to the first possibility, the second possibility, the fourth possibility, and the fifth possibility, and determine an output important image information comparison quality index value according to the third possibility and the sixth possibility;

[0041] The enhanced configuration quality index value corresponding to the enhanced configuration example is obtained according to the output event comparison quality index value and the output important image information comparison quality index value.

[0042] In an independently implemented embodiment, determining the output event comparison quality index value according to the first possibility, the second possibility, the fourth possibility, and the fifth possibility includes:

[0043] Determine, according to the first possibility and the fourth possibility, a first output possibility ratio of the configured original image analysis thread and the event generation reference thread outputting the selected monitoring area event;

[0044] According to the second possibility and the fifth possibility, determining a second output possibility ratio of the unselected monitoring area event output by the configured original image analysis thread and the event generation reference thread;

[0045] An output event comparison quality indicator value is calculated based on the first output likelihood ratio value and the second output likelihood ratio value.

[0046] In an independently implemented embodiment, determining the comparison quality index value of the output important image information according to the third possibility and the sixth possibility includes:

[0047] Determine the configured original image analysis thread and the event generation reference thread according to the third possibility and the sixth possibility, output a third output possibility ratio of important monitoring image information while outputting the selected monitoring area event, and output a fourth output possibility ratio of important monitoring image information while outputting the unselected monitoring area event;

[0048] Based on the third output likelihood ratio and the fourth output likelihood ratio, an output important image information comparison quality index value is calculated.

[0049] In an independent implementation embodiment, the method further includes:

[0050] Get the monitoring area topics to be processed;

[0051] The monitoring area subject to be processed is input into the configured image analysis thread to obtain the monitoring area event corresponding to the monitoring area subject to be processed.

[0052] In a second aspect, a surveillance focus image recognition system based on a drone is provided, comprising a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute it to implement the above method.

[0053] The embodiment of the present application provides a method and system for identifying key monitoring images based on drones. On the basis of obtaining several original configuration examples for configuring the original image analysis thread, by inputting the example monitoring area theme in the original configuration example into the original image analysis thread for each original configuration example to perform monitoring area event regression analysis and important image information regression analysis, the regression analysis monitoring area events and regression analysis important image information features can be obtained, and then the example monitoring area events, regression analysis monitoring area events, monitoring important image information and regression analysis important image information features can be used to perform regression analysis quality index calculation, and the regression analysis quality index value corresponding to the original configuration example can be obtained, so that The configured image analysis thread is obtained by optimizing the coefficients of the original image analysis thread according to the regression analysis quality index values ​​corresponding to several original configuration examples. During the whole process, when the original image analysis thread is configured, the regression analysis of important monitoring image information is cascaded to the original image analysis thread, so that the original image analysis thread can understand the correct information related to the event from the important monitoring image information, and support the generation of monitoring area events, which can avoid the situation where the monitoring area event output causes abnormalities to the monitoring area subject, and can obtain an image analysis thread that can output accurate monitoring area events. The generation of accurate monitoring area events can be achieved by inputting the monitoring area subject into the configured image analysis thread. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 A flowchart of a method for identifying key surveillance images based on a drone provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0057] See also Figure 1 , shows a method for identifying surveillance focus images based on a drone, which may include the technical solutions described in the following steps 202 to 208.

[0058] Step 202, obtain several original configuration examples for configuring the original image analysis thread; the original configuration examples include example monitoring area topics, example monitoring area events, and important monitoring image information describing the content in the example monitoring area topics, and the important monitoring image information is different from the example monitoring area events.

[0059] The original image analysis thread refers to the original image analysis thread that has been coefficient-initialized and has set learning coefficients, but has not updated the coefficients. It is understandable that the coefficients in the original image analysis thread can be updated by learning from several original configuration examples to obtain the configured image analysis thread.

[0060] The original configuration example refers to a configuration example for configuring the original image analysis thread. The original configuration example includes an example monitoring area theme, an example monitoring area event, and important monitoring image information describing the content in the example monitoring area theme, which is different from the example monitoring area event. The example monitoring area theme refers to the monitoring area theme used as an example.

[0061] Specifically, when an image analysis thread needs to be configured, the server will obtain several original configuration examples for configuring the original image analysis thread, wherein the original configuration examples include example monitoring area topics, example monitoring area events, and important monitoring image information describing the content in the example monitoring area topics, and the important monitoring image information is different from the example monitoring area events.

[0062] In a specific application, the example monitoring area event can be obtained by extracting the monitoring area using the example monitoring area theme, and the important monitoring image information can be obtained by extracting the important image information using the example monitoring area theme. That is, when generating the original configuration example, the server will first obtain the example monitoring area theme, and then use the example monitoring area theme to extract the example monitoring area event and the important monitoring image information.

[0063] Step 204, for each original configuration example, input the example monitoring area theme in the original configuration example into the original image analysis thread to perform monitoring area event regression analysis and important image information regression analysis to obtain regression analysis monitoring area events and regression analysis important image information features.

[0064] Among them, the regression analysis monitoring area event refers to the example event that is regressed and analyzed by the original image analysis thread based on the example monitoring area theme in the original configuration example. The regression analysis important image information feature refers to the feature that characterizes the important image information for regression analysis. For example, the regression analysis important image information feature may specifically refer to a vector that characterizes the important image information for regression analysis. It can be understood that the regression analysis important image information can be uniquely determined through the regression analysis important image information features output by the original image analysis thread. The regression analysis important image information refers to the important image information for regression analysis that can be uniquely determined based on the regression analysis important image information features output by the original image analysis thread.

[0065] Specifically, for each original configuration example, the server will input the example monitoring area theme in the original configuration example into the original image analysis thread to perform regression analysis of monitoring area events and regression analysis of important image information, and obtain regression analysis monitoring area events and regression analysis of important image information features. In a specific application, after the example monitoring area theme is input into the original image analysis thread, the original image analysis thread will compress the example monitoring area theme to obtain a compression vector corresponding to the example monitoring area theme, and then use the compression vector corresponding to the example monitoring area theme to regress and analyze monitoring area events and important image information, and obtain regression analysis monitoring area events and regression analysis of important image information features.

[0066] In a specific application, the original image analysis thread may compress the example monitoring area subject by compressing each pixel in the example monitoring area subject to obtain pixel compression corresponding to each pixel in the example monitoring area subject, that is, the compression vector corresponding to the example monitoring area subject includes the pixel compression corresponding to each pixel in the example monitoring area subject.

[0067] For example, the original image analysis thread can compress each pixel in the example monitoring area theme through a preset pixel dictionary. The preset pixel dictionary can be configured according to the actual application scenario. In the preset pixel dictionary, each pixel or pixel mother has a unique pixel compression. Then, each pixel in the example monitoring area theme is compressed through the preset pixel dictionary, which actually maps each pixel in the example monitoring area theme to the pixel compression in the preset pixel dictionary.

[0068] In a specific application, when using the compression vector corresponding to the example monitoring area theme to regress and analyze the monitoring area event, the server can generate the monitoring area event by pixel-by-pixel regression analysis, that is, the first pixel in the regression analysis monitoring area event is obtained by the server directly based on the compression vector corresponding to the example monitoring area theme. For each pixel after the first pixel in the regression analysis monitoring area event, the server can combine the compression vector corresponding to the example monitoring area theme and the pixel obtained by the previous regression analysis to perform regression analysis, and obtain the pixel in the regression analysis monitoring area event of this regression analysis. In this way, each time a regression analysis is performed, the regression analysis can be performed in combination with the results of the previous regression analysis, and the generation of the regression analysis monitoring area event can be achieved on the basis of the above.

[0069] In a specific application, when using the compressed vector corresponding to the example monitoring area theme to regress and analyze important image information, in order to make the regression analysis of the important image information assist the regression analysis of the monitoring area event, the intermediate features output by the regression analysis of the monitoring area event part in the original image analysis thread can be used to regress and analyze the important image information. For example, the features output by the first pixel regression analysis network in the event regression analysis network in the original image analysis thread can be used to regress and analyze the important image information. Then, the output of the first pixel regression analysis network in the original image analysis thread will be used as the input of the important image information regression analysis network. The important image information regression analysis network can obtain the regression analysis of the important image information features by processing the output of the first pixel regression analysis network.

[0070] Step 206 , calculate the regression analysis quality index according to the example monitoring area events, regression analysis monitoring area events, monitoring important image information and regression analysis important image information features to obtain the regression analysis quality index value corresponding to the original configuration example.

[0071] Among them, the regression analysis quality index value corresponding to the original configuration example refers to the sum of the errors between the regression analysis monitoring area events and the example monitoring area events, and the errors between the monitoring important image information features corresponding to the monitoring important image information and the regression analysis important image information features.

[0072] Specifically, after the original image analysis thread outputs the regression analysis monitoring area events and the regression analysis important image information features, the server will calculate the event regression analysis quality index value based on the example monitoring area events and the regression analysis monitoring area events, and calculate the important image information regression analysis quality index value based on the monitoring important image information and the regression analysis important image information features, and then obtain the regression analysis quality index value corresponding to the original configuration example based on the event regression analysis quality index value and the important image information regression analysis quality index value.

[0073] Step 208, according to the regression analysis quality index values ​​corresponding to the several original configuration examples, the coefficients of the original image analysis thread are optimized to obtain a configured image analysis thread; the configured image analysis thread is used to output monitoring area events based on the input monitoring area theme.

[0074] Specifically, after obtaining the regression analysis quality index values ​​corresponding to several original configuration examples, the server will calculate the average of the regression analysis quality index values ​​corresponding to several original configuration examples, and return the average value to the original image analysis thread, calculate the coefficient gradient of each coefficient of the original image analysis thread, and then update each coefficient according to the coefficient gradient of each coefficient to obtain the configured original image analysis thread, and then obtain the configured image analysis thread according to the configured original image analysis thread. The configured image analysis thread is used to output monitoring area events based on the input monitoring area theme.

[0075] In specific applications, if the configured original image analysis thread is not further configured, the server can directly use the configured original image analysis thread as the configured image analysis thread. However, in order to enable the configured original image analysis thread to output more accurate monitoring area events, the server will further enhance the configured original image analysis thread to obtain the configured image analysis thread.

[0076] The above-mentioned drone-based monitoring key image recognition method, based on obtaining several original configuration examples for configuring the original image analysis thread, inputs the example monitoring area theme in the original configuration example into the original image analysis thread for monitoring area event regression analysis and important image information regression analysis for each original configuration example, and can obtain regression analysis monitoring area events and regression analysis important image information features, and then can use the example monitoring area events, regression analysis monitoring area events, monitoring important image information and regression analysis important image information features to perform regression analysis quality index calculation, and obtain the regression analysis quality index value corresponding to the original configuration example, so that it can be obtained by The regression analysis quality index values ​​corresponding to the original configuration examples are respectively used to optimize the coefficients of the original image analysis thread to obtain the configured image analysis thread. During the whole process, when configuring the original image analysis thread, the regression analysis of important monitoring image information is cascaded to the original image analysis thread, so that the original image analysis thread can understand the correct information related to the event from the important monitoring image information, and support the generation of monitoring area events, which can avoid the situation where the monitoring area event output causes abnormalities to the monitoring area subject, and can obtain the image analysis thread that can output accurate monitoring area events. By inputting the monitoring area subject into the configured image analysis thread, it is possible to generate accurate monitoring area events.

[0077] In one embodiment, the regression analysis quality index is calculated based on the example monitoring area events, regression analysis monitoring area events, monitoring important image information, and regression analysis important image information features, and the regression analysis quality index value corresponding to the original configuration example is obtained, including:

[0078] Calculate the event regression analysis quality index value based on the example monitoring area events and regression analysis monitoring area events, and calculate the important image information regression analysis quality index value based on the monitoring important image information and regression analysis important image information features;

[0079] The event regression analysis quality index values ​​and the important image information regression analysis quality index values ​​are aggregated to obtain the regression analysis quality index values ​​corresponding to the original configuration example.

[0080] The event regression analysis quality index value refers to the error between the example monitoring area event and the regression analysis monitoring area event. The important image information regression analysis quality index value refers to the error between the monitoring important image information features corresponding to the monitoring important image information and the regression analysis important image information features.

[0081] Specifically, when calculating the regression analysis quality index value, the server will calculate the event regression analysis quality index value based on the example monitoring area events and the regression analysis monitoring area events, and compress the monitoring important image information to obtain the corresponding monitoring important image information features of the monitoring important image information, and calculate the important image information regression analysis quality index value based on the monitoring important image information features and the regression analysis important image information features, and finally aggregate the event regression analysis quality index values ​​and the important image information regression analysis quality index values ​​to obtain the regression analysis quality index value corresponding to the original configuration example.

[0082] In a specific application, the regression analysis monitoring area event output by the original image analysis thread can be specifically a regression analysis monitoring area event vector. When calculating the event regression analysis quality index value, the server will first compress the example monitoring area event to obtain the compression vector corresponding to the example monitoring area event, and then determine the event regression analysis quality index value based on the compression vector corresponding to the example monitoring area event and the regression analysis monitoring area event vector.

[0083] In specific applications, the regression analysis important image information features output by the original image analysis thread are actually in vector form, which can be understood as the representation of the regression analysis important image information. Therefore, when calculating the important image information regression analysis quality index value, the server will first compress the monitoring important image information to obtain the corresponding monitoring important image information features of the monitoring important image information, and then determine the important image information regression analysis quality index value by comparing the monitoring important image information features with the regression analysis important image information features.

[0084] In a specific application, when collecting the event regression analysis quality index value and the important image information regression analysis quality index value, the server can directly superimpose the event regression analysis quality index value and the important image information regression analysis quality index value to obtain the regression analysis quality index value corresponding to the original configuration example, or pre-set the weighting coefficients corresponding to the event regression analysis quality index value and the important image information regression analysis quality index value, and then perform weighted summation of the event regression analysis quality index value and the important image information regression analysis quality index value according to the weighting coefficient to obtain the regression analysis quality index value corresponding to the original configuration example. The weighting coefficient can be configured according to the actual application scenario.

[0085] It is understandable that because the monitoring area event regression analysis is the main instruction and the important image information regression analysis is the auxiliary instruction, the weighted coefficient corresponding to the event regression analysis quality index value is usually much larger than the weighted coefficient corresponding to the important image information regression analysis quality index value.

[0086] In this embodiment, by first calculating the event regression analysis quality index value based on the example monitoring area events and regression analysis monitoring area events, and calculating the important image information regression analysis quality index value based on the monitoring important image information and regression analysis important image information features, the aggregated event regression analysis quality index value and the important image information regression analysis quality index value can be used to determine the regression analysis quality index value corresponding to the original configuration example.

[0087] In one embodiment, according to the example monitoring area events and regression analysis monitoring area events, calculating the event regression analysis quality index value includes:

[0088] Compressing each pixel in the example monitoring area event respectively, obtaining pixel compression corresponding to each pixel in the example monitoring area event, and determining the regression analysis possibility corresponding to each pixel in the regression analysis monitoring area event;

[0089] For each pixel in the example monitoring area event, the regression analysis quality index value corresponding to the pixel is calculated according to the pixel compression corresponding to the pixel and the regression analysis possibility corresponding to the pixel at the same position in the regression analysis monitoring area event;

[0090] According to the regression analysis quality index value corresponding to each pixel in the example monitoring area event, the event regression analysis quality index value is obtained.

[0091] Specifically, when calculating the event regression analysis quality index value, the server can first compress each pixel in the example monitoring area event by using a preset pixel code, obtain the pixel compression corresponding to each pixel in the example monitoring area event, and determine the regression analysis possibility corresponding to each pixel in the regression analysis monitoring area event, and then for each pixel in the example monitoring area event, according to the pixel compression corresponding to the pixel and the regression analysis possibility corresponding to the pixel at the same position in the regression analysis monitoring area event, calculate the regression analysis quality index value corresponding to the pixel, and finally, according to the regression analysis quality index value corresponding to each pixel in the example monitoring area event, obtain the event regression analysis quality index value. In a specific application, for each pixel in the example monitoring area event, according to the pixel compression corresponding to the pixel and the regression analysis possibility corresponding to the pixel at the same position in the regression analysis monitoring area event, it can be determined whether the pixel at the same position in the regression analysis monitoring area event is the same as the pixel at the same position, that is, according to the regression analysis possibility, the classification quality index corresponding to the pixel at the same position can be determined, so that the event regression analysis quality index value can be obtained in combination with the classification quality index value corresponding to each pixel in the example monitoring area event. In a specific application, the server may use the average value of the classification quality index value corresponding to each pixel in the example monitoring area event as the event regression analysis quality index value.

[0092] In this embodiment, by first compressing each pixel in the example monitoring area event, obtaining the pixel compression corresponding to each pixel in the example monitoring area event, and determining the regression analysis possibility corresponding to each pixel in the regression analysis monitoring area event, the pixel compression can be used to calculate the regression analysis quality index value corresponding to each pixel in the example monitoring area event, and then the regression analysis quality index value corresponding to each pixel in the example monitoring area event can be used to determine the event regression analysis quality index value.

[0093] In one embodiment, according to monitoring important image information and regressing and analyzing important image information features, calculating the important image information regression analysis quality index value includes:

[0094] Compressing the important monitoring image information to obtain the corresponding important monitoring image information features of the important monitoring image information;

[0095] Calculate the commonality weight quality index value between the important monitoring image information features and the important regression analysis image information features, and use the commonality weight quality index value as the important image information regression analysis quality index value.

[0096] Among them, the commonality weight quality index value is used to measure the similarity between the important image information features of monitoring and the important image information features of regression analysis. It can be understood that the more similar the important image information features of monitoring and the important image information features of regression analysis are, the smaller the commonality weight quality index value is, and the greater the gap between the important image information features of monitoring and the important image information features of regression analysis is, the greater the commonality weight quality index value is.

[0097] Specifically, when calculating the important image information regression analysis quality index value, the server will first compress the monitoring important image information to obtain the corresponding monitoring important image information features of the monitoring important image information, and then calculate the commonality weight quality index value between the monitoring important image information features and the regression analysis important image information features, and use the commonality weight quality index value as the important image information regression analysis quality index value. It can be understood that since the monitoring important image information features represent the monitoring important image information, and the regression analysis important image information features represent the regression analysis important image information, the similarity between the monitoring important image information and the regression analysis important image information can be determined by comparing the similarity between the monitoring important image information features and the regression analysis important image information features, and by improving the similarity between the regression analysis important image information and the monitoring important image information through quality index calculation, the original image analysis thread output can contain more such monitoring important image information, which is conducive to the accurate monitoring area event regression analysis.

[0098] In this embodiment, by compressing the important monitoring image information, the corresponding important monitoring image information features of the important monitoring image information are obtained. The important monitoring image information features and the regression analysis important image information features can be used to calculate the common weight quality index value between the two features, and then the common weight quality index value can be used to determine the regression analysis quality index value of the important image information.

[0099] In one embodiment, the important image information to be monitored is the relationship between things in the monitoring area; the relationship between things in the monitoring area is obtained by:

[0100] Counting the occurrence frequencies of monitoring area things appearing in the example monitoring area topic, determining a number of monitoring area things, and the occurrence frequencies of the respective monitoring area things;

[0101] According to the occurrence frequencies of the objects in the monitoring areas, no less than two key objects are extracted from the objects in the monitoring areas;

[0102] Based on the sample monitoring area theme, determine the key transaction relationship between any two key transactions of no less than two key transactions;

[0103] Based on a key thing relationship between any two key things among the at least two key things, a monitoring area thing relationship describing the content in the example monitoring area topic is determined.

[0104] In this embodiment, by counting the occurrence frequencies of monitoring area things that appear in the example monitoring area theme, a number of monitoring area things and the occurrence frequencies of each of the monitoring area things are determined, and the occurrence frequencies of each of the monitoring area things can be used to extract no less than two key things from the several monitoring area things. Then, by determining the key thing relationship between any two of the no less than two key things, accurate determination of the monitoring area thing relationship that describes the content in the example monitoring area theme can be achieved.

[0105] In one embodiment, according to the example monitoring area theme, determining the key transaction relationship between any two key transactions of at least two key transactions includes:

[0106] Taking every two key things among the at least two key things as a group, a number of key thing binary groups are obtained;

[0107] For each key thing binary group, a regression analysis is performed on the relationship between the two key things in the key thing binary group according to the example monitoring area theme to obtain the key thing relationship between the two key things in the key thing binary group.

[0108] Specifically, when determining the relationship between key things, the server will take every two key things out of no less than two key things as a group to obtain several key thing tuples. For each key thing tuple, a regression analysis is performed on the relationship between the two key things in the key thing tuple according to the example monitoring area theme to obtain the key thing relationship between the two key things in the key thing tuple.

[0109] In one embodiment, based on the key thing relationship between any two key things among the at least two key things, determining the monitoring area thing relationship describing the content in the example monitoring area topic includes:

[0110] When the number of the key things is not less than two, the key thing relationship between the two key things is used as the monitoring area thing relationship that describes the content in the example monitoring area theme.

[0111] Specifically, when the number of key things is not less than two, the server can directly use the key thing relationship between the two key things as the monitoring area thing relationship describing the content in the example monitoring area topic. In this way, the monitoring area thing relationship can be quickly determined.

[0112] In one embodiment, based on the key thing relationship between any two key things among the at least two key things, determining the monitoring area thing relationship describing the content in the example monitoring area topic includes:

[0113] When the number of at least two key items is greater than two, the two key items with the highest occurrence frequency are determined according to the occurrence frequency of each of the at least two key items;

[0114] The key thing relationship of the two key things that appear most frequently is used as the monitoring area thing relationship that describes the content in the example monitoring area topic.

[0115] Specifically, when the number of at least two key things is greater than two, the server needs to determine the two key things with the highest occurrence frequency according to the occurrence frequency of at least two key things respectively, so as to use the key thing relationship of the two key things with the highest occurrence frequency as the monitoring area thing relationship describing the content in the example monitoring area theme.

[0116] It can be understood that when the number of no less than two key things is greater than two, there will be a key thing relationship between every two key things. By using the frequency of occurrence of no less than two key things to screen the key thing relationships, it is possible to accurately characterize the relationship between things in the monitored area by using the key thing relationships of the key things based on determining the key things that can best promote the development of the monitored area.

[0117] In one embodiment, according to the regression analysis quality index values ​​corresponding to the several original configuration examples, the coefficients of the original image analysis thread are optimized, and the configured image analysis thread includes:

[0118] According to the regression analysis quality index values ​​corresponding to the several original configuration examples, the coefficients of the original image analysis thread are optimized to obtain the configured original image analysis thread;

[0119] Obtaining an event generation reference thread; the event generation reference thread includes a network structure of a classification instruction header and an original image analysis thread, and is obtained by configuring several original configuration examples;

[0120] Obtaining several enhanced configuration examples, for each enhanced configuration example, using the example monitoring area theme in the enhanced configuration example, the selected monitoring area events and the unselected monitoring area events corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme, configuring the configured original image analysis thread and the event generation reference thread, and obtaining the enhanced configuration quality indicator value corresponding to the enhanced configuration example;

[0121] According to the enhanced configuration quality indicator values ​​corresponding to the respective enhanced configuration examples, the coefficients of the configured original image analysis thread are optimized to obtain the configured image analysis thread.

[0122] Among them, the classification instruction head refers to a processing layer for executing classification instructions. It is understandable that the classification instruction head can output a classification result by analyzing the input features. In this embodiment, the classification result refers to the important image information of the regression analysis output by the event generation reference thread. The selected monitoring area event refers to an event that can accurately summarize the theme of the example monitoring area, and the unselected monitoring area event refers to an event that cannot summarize the theme of the example monitoring area. It is understandable that the selected monitoring area event summarizes the theme of the example monitoring area better than the unselected monitoring area event.

[0123] Specifically, the server can optimize the coefficients of the original image analysis thread according to the regression analysis quality index values ​​corresponding to several original configuration examples to obtain the configured original image analysis thread. On this basis, the server can obtain an event generation reference thread and several enhanced configuration examples to further enhance the configured original image analysis thread to obtain the configured image analysis thread. Among them, the event generation reference thread includes a classification instruction header and a network structure of the original image analysis thread, and is obtained through the configuration of several original configuration examples. The enhanced configuration examples include example monitoring area themes, selected monitoring area events and unselected monitoring area events corresponding to the example monitoring area themes, and important monitoring image information describing the content in the example monitoring area themes.

[0124] Specifically, based on obtaining an event generation reference thread and several enhanced configuration examples, for each enhanced configuration example, the server will use the example monitoring area theme in the enhanced configuration example, the selected monitoring area events and unselected monitoring area events corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme to respectively configure the configured original image analysis thread and the event generation reference thread, and calculate the enhanced configuration quality index value corresponding to the enhanced configuration example based on the regression analysis results output by the configured original image analysis thread and the regression analysis results output by the event generation reference thread, and then optimize the coefficient of the configured original image analysis thread based on the enhanced configuration quality index values ​​corresponding to each of the several enhanced configuration examples to obtain the configured image analysis thread.

[0125] Among them, by comparing the regression analysis results output by the two threads, the event generation effect of the configured original image analysis thread can be evaluated, and the enhanced configuration quality index value used to evaluate the event generation effect can be obtained. Then, the enhanced configuration quality index value can be used to optimize the coefficient of the configured original image analysis thread to optimize the event generation effect of the configured original image analysis thread and obtain the configured image analysis thread.

[0126] In specific applications, the main difference between the event generation reference thread and the original image analysis thread is that the event generation reference thread has an additional classification instruction header compared to the original image analysis thread. In the event generation reference thread, this classification instruction header is used in the important image information regression analysis network, that is, the output of the important image information regression analysis network of the event generation reference thread is no longer the regression analysis of important image information features, but the regression analysis of important image information. Therefore, when configuring the event generation reference thread, for each original configuration example, the server will first output the example monitoring area theme in the original configuration example to the original event generation reference thread for monitoring area event regression analysis and important image information regression analysis to obtain regression analysis monitoring area events and regression analysis of important image information, and then calculate the regression analysis quality index based on the example monitoring area events, regression analysis monitoring area events, monitoring important image information and regression analysis important image information to obtain the regression analysis quality index value corresponding to the original configuration example, and optimize the coefficient of the original event generation reference thread based on the regression analysis quality index values ​​corresponding to several original configuration examples to obtain the event generation reference thread.

[0127] In one embodiment, for each enhanced configuration example, the configured original image analysis thread and event generation reference thread are configured using the example monitoring area theme in the enhanced configuration example, the selected monitoring area events and the unselected monitoring area events corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme, and the enhanced configuration quality indicator value corresponding to the enhanced configuration example is obtained, including:

[0128] For each enhanced configuration example, inputting the example monitoring area theme in the enhanced configuration example into the configured raw image analysis thread, determining the configured raw image analysis thread outputting a first possibility of an event in the selected monitoring area, a second possibility of an event in the unselected monitoring area, and a third possibility of outputting important monitoring image information;

[0129] Inputting the example monitoring area theme in the enhanced configuration example into the event generation reference thread, determining that the event generation reference thread outputs a fourth possibility of a selected monitoring area event, a fifth possibility of an unselected monitoring area event, and a sixth possibility of an important monitoring image information;

[0130] According to the first possibility, the second possibility, the fourth possibility and the fifth possibility, determining the output event comparison quality index value, and according to the third possibility and the sixth possibility, determining the output important image information comparison quality index value;

[0131] According to the output event comparison quality index value and the output important image information comparison quality index value, the enhanced configuration quality index value corresponding to the enhanced configuration example is obtained.

[0132] Specifically, for each enhanced configuration example, the server inputs the example monitoring area theme in the enhanced configuration example into the configured original image analysis thread, and can use the configured original image analysis thread to perform regression analysis of monitoring area events and regression analysis of important image information, and then use the regression analysis results to calculate the configured original image analysis thread to output the first possibility of the selected monitoring area event, the second possibility of the unselected monitoring area event, and the third possibility of the important monitoring image information. At the same time, the server inputs the example monitoring area theme in the enhanced configuration example into the event generation reference thread, and can use the event generation reference thread to perform regression analysis of monitoring area events and regression analysis of important image information, and then use the regression analysis results to calculate the event generation reference thread to output the fourth possibility of the selected monitoring area event, the fifth possibility of the unselected monitoring area event, and the sixth possibility of the important monitoring image information.

[0133] Specifically, on the basis of obtaining the above possibilities, the server can calculate the output event comparison quality index value based on the first possibility, the second possibility, the fourth possibility and the fifth possibility related to the regression analysis monitoring area events, and use the third possibility and the sixth possibility related to the regression analysis important image information to determine the output important image information comparison quality index value, and then obtain the enhanced configuration quality index value corresponding to the enhanced configuration example by aggregating the output event comparison quality index values ​​and the output important image information comparison quality index values.

[0134] In a specific application, the first possibility of outputting the selected monitoring area event can be obtained based on the selected monitoring area event and the regression analysis monitoring area event in the regression analysis result, and the second possibility of outputting the unselected monitoring area event can be obtained based on the unselected monitoring area event and the regression analysis monitoring area event in the regression analysis result.

[0135] In this embodiment, by inputting the example monitoring area theme in the enhanced configuration example into the configured original image analysis thread and event generation reference thread, it is possible to accurately calculate the enhanced configuration quality index value corresponding to the enhanced configuration example through several possibilities determined based on the regression analysis results, and then the enhanced configuration quality index value can be used to optimize the configured original image analysis thread to obtain the configured image analysis thread.

[0136] In one embodiment, according to the first possibility, the second possibility, the fourth possibility and the fifth possibility, determining the output event comparison quality indicator value includes:

[0137] According to the first possibility and the fourth possibility, determining a first output possibility ratio of the events of the selected monitoring area output by the configured original image analysis thread and the event generation reference thread;

[0138] According to the second possibility and the fifth possibility, determining a second output possibility ratio of the unselected monitoring area events output by the configured original image analysis thread and the event generation reference thread;

[0139] An output event comparison quality indicator value is calculated based on the first output likelihood ratio value and the second output likelihood ratio value.

[0140] In one embodiment, according to the third possibility and the sixth possibility, determining the comparison quality index value of the output important image information includes:

[0141] According to the third possibility and the sixth possibility, determining the configured original image analysis thread and event generation reference thread, outputting a third output possibility ratio of important monitoring image information while outputting events of selected monitoring areas, and outputting a fourth output possibility ratio of important monitoring image information while outputting events of unselected monitoring areas;

[0142] Based on the third output likelihood ratio value and the fourth output likelihood ratio value, an output important image information comparison quality index value is calculated.

[0143] In a specific application, the possibility of outputting the important monitoring image information while outputting the selected monitoring area event can be obtained according to the possibility of outputting the selected monitoring area event, the important monitoring image information, and the regression analysis important image information in the regression analysis result. The possibility of outputting the important monitoring image information while outputting the unselected monitoring area event can be obtained according to the possibility of outputting the unselected monitoring area event, the important monitoring image information, and the regression analysis important image information in the regression analysis result.

[0144] In this embodiment, by first comparing the possibility of the configured original image analysis thread and the event generation reference thread outputting the selected monitoring area event and the outputting the important monitoring image information, and then comparing the possibility of the configured original image analysis thread and the event generation reference thread outputting the unselected monitoring area event and the outputting the important monitoring image information, the third output possibility ratio and the fourth output possibility ratio obtained by the two comparisons can be combined to realize the calculation of the comparison quality index value of the output important image information.

[0145] In one embodiment, the drone-based monitoring focus image recognition method further includes:

[0146] Get the monitoring area topics to be processed;

[0147] The monitoring area subject to be processed is input into the configured image analysis thread to obtain the monitoring area event corresponding to the monitoring area subject to be processed.

[0148] Specifically, when the configured image analysis thread is obtained, when the monitoring area theme needs to be processed, the server obtains the monitoring area theme to be processed, inputs the monitoring area theme to be processed into the configured image analysis thread, and obtains the monitoring area event corresponding to the monitoring area theme to be processed.

[0149] In this embodiment, since an image analysis thread capable of outputting accurate monitoring area events is configured, accurate monitoring area events can be generated by inputting the monitoring area subject into the configured image analysis thread. Based on the above, a monitoring focus image recognition system based on a drone is shown, including a processor and a memory communicating with each other, and the processor is used to read and execute a computer program from the memory to implement the above method.

[0150] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0151] In summary, based on the above scheme, on the basis of obtaining several original configuration examples for configuring the original image analysis thread, by inputting the example monitoring area theme in the original configuration example into the original image analysis thread for each original configuration example to perform monitoring area event regression analysis and important image information regression analysis, it is possible to obtain regression analysis monitoring area events and regression analysis important image information features, and then the example monitoring area events, regression analysis monitoring area events, monitoring important image information and regression analysis important image information features can be used to perform regression analysis quality index calculation, and the regression analysis quality index value corresponding to the original configuration example can be obtained, so that ... The corresponding regression analysis quality index values ​​of the examples are used to optimize the coefficients of the original image analysis thread to obtain the configured image analysis thread. During the whole process, when configuring the original image analysis thread, the regression analysis of important monitoring image information is cascaded to the original image analysis thread, so that the original image analysis thread can understand the correct information related to the event from the important monitoring image information, and support the generation of monitoring area events, which can avoid the situation where the monitoring area event output causes abnormalities to the monitoring area subject, and can obtain an image analysis thread that can output accurate monitoring area events. By inputting the monitoring area subject into the configured image analysis thread, it is possible to generate accurate monitoring area events.

[0152] It should be understood that the system and its modules shown above can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer executable instructions and / or included in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of the present application can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software such as executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (e.g., firmware).

[0153] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

Claims

1. A method for identifying key surveillance images based on drones, characterized in that: The method comprises: Obtaining several original configuration examples for configuring an original image analysis thread; the original configuration examples include an example monitoring area theme, an example monitoring area event, and important monitoring image information describing the content in the example monitoring area theme, the important monitoring image information is different from the example monitoring area event; For each of the original configuration examples, inputting the example monitoring area subject in the original configuration example into the original image analysis thread to perform monitoring area event regression analysis and important image information regression analysis, and obtaining regression analysis monitoring area events and regression analysis important image information features; Calculate the regression analysis quality index according to the example monitoring area event, the regression analysis monitoring area event, the monitoring important image information and the regression analysis important image information features to obtain the regression analysis quality index value corresponding to the original configuration example; According to the regression analysis quality index values ​​corresponding to each of the several original configuration examples, the coefficients of the original image analysis thread are optimized to obtain a configured image analysis thread; the configured image analysis thread is used to output monitoring area events based on the input monitoring area theme.

2. The method according to claim 1, characterized in that The regression analysis quality index calculation is performed according to the example monitoring area event, the regression analysis monitoring area event, the monitoring important image information and the regression analysis important image information feature to obtain the regression analysis quality index value corresponding to the original configuration example, including: Calculate the event regression analysis quality index value according to the example monitoring area event and the regression analysis monitoring area event, and calculate the important image information regression analysis quality index value according to the monitoring important image information and the regression analysis important image information features; The event regression analysis quality index value and the important image information regression analysis quality index value are aggregated to obtain the regression analysis quality index value corresponding to the original configuration example.

3. The method according to claim 2, characterized in that The calculating of the event regression analysis quality index value according to the example monitoring area event and the regression analysis monitoring area event comprises: Compressing each pixel in the example monitoring area event respectively to obtain pixel compression corresponding to each pixel in the example monitoring area event, and determining the regression analysis possibility corresponding to each pixel in the regression analysis monitoring area event; For each pixel in the example monitoring area event, the regression analysis quality index value corresponding to the pixel is calculated according to the pixel compression corresponding to the pixel and the regression analysis possibility corresponding to the pixel at the same position in the regression analysis monitoring area event; According to the regression analysis quality index value corresponding to each pixel in the example monitoring area event, the event regression analysis quality index value is obtained.

4. The method according to claim 2, characterized in that: Calculating the important image information regression analysis quality index value according to the monitored important image information and the regression analysis important image information features includes: Compressing the important monitoring image information to obtain important monitoring image information features corresponding to the important monitoring image information; The commonality weight quality index value between the monitoring important image information feature and the regression analysis important image information feature is calculated, and the commonality weight quality index value is used as the important image information regression analysis quality index value.

5. The method according to claim 1, characterized in that The important monitoring image information is the relationship between things in the monitoring area; the relationship between things in the monitoring area is obtained by: Counting the occurrence frequencies of monitoring area things appearing in the example monitoring area theme, determining a number of monitoring area things, and the occurrence frequencies of each of the several monitoring area things; Extracting at least two key objects from the plurality of monitoring area objects according to the respective occurrence frequencies of the plurality of monitoring area objects; Determine a key thing relationship between any two of the at least two key things according to the example monitoring area theme; Based on the key thing relationship between any two key things among the at least two key things, a monitoring area thing relationship describing the content in the example monitoring area theme is determined.

6. The method according to claim 5, characterized in that Determining the key thing relationship between any two key things among the at least two key things according to the example monitoring area theme includes: Taking every two key things among the at least two key things as a group, a plurality of key thing binary groups are obtained; For each key thing binary group, a regression analysis is performed on the relationship between the two key things in the key thing binary group according to the example monitoring area theme to obtain the key thing relationship between the two key things in the key thing binary group.

7. The method according to claim 5, characterized in that The determining of the monitoring area thing relationship describing the content in the example monitoring area theme based on the key thing relationship between any two key things among the at least two key things includes: When the number of the at least two key things is two, the key thing relationship between the two key things is used as the monitoring area thing relationship describing the content in the example monitoring area theme; Wherein, determining the monitoring area thing relationship describing the content in the example monitoring area theme based on the key thing relationship between any two key things among the at least two key things includes: When the number of the at least two key items is greater than two, determining the two key items with the highest occurrence frequencies according to the respective occurrence frequencies of the at least two key items; The key thing relationship of the two key things with the highest occurrence frequency is used as the monitoring area thing relationship for describing the content in the example monitoring area theme.

8. The method according to claim 1, characterized in that The step of optimizing the coefficients of the original image analysis thread according to the regression analysis quality index values ​​corresponding to the plurality of original configuration examples to obtain the configured image analysis thread includes: According to the regression analysis quality index values ​​corresponding to the several original configuration examples, the coefficients of the original image analysis thread are optimized to obtain a configured original image analysis thread; Obtaining an event generation reference thread; the event generation reference thread includes a classification instruction header and a network structure of the original image analysis thread, and is obtained by configuring the plurality of original configuration examples; Obtaining several enhanced configuration examples, for each enhanced configuration example, using the example monitoring area theme in the enhanced configuration example, the selected monitoring area events and the unselected monitoring area events corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme, configuring the configured original image analysis thread and the event generation reference thread, and obtaining the enhanced configuration quality indicator value corresponding to the enhanced configuration example; According to the enhanced configuration quality index values ​​corresponding to the plurality of enhanced configuration examples, the coefficients of the configured original image analysis thread are optimized to obtain the configured image analysis thread; Among them, for each enhanced configuration example, the configured original image analysis thread and the event generation reference thread are configured using the example monitoring area theme in the enhanced configuration example, the selected monitoring area event and the unselected monitoring area event corresponding to the example monitoring area theme, and the important monitoring image information describing the content in the example monitoring area theme, and the enhanced configuration quality indicator value corresponding to the enhanced configuration example is obtained, including: For each enhanced configuration example, inputting the example monitoring area theme in the enhanced configuration example into the configured raw image analysis thread, determining the configured raw image analysis thread outputting a first possibility of the selected monitoring area event, a second possibility of the unselected monitoring area event, and a third possibility of the important monitoring image information; Inputting the example monitoring area theme in the enhanced configuration example into the event generation reference thread, determining that the event generation reference thread outputs a fourth possibility of the selected monitoring area event, a fifth possibility of the unselected monitoring area event, and a sixth possibility of the important monitoring image information; Determine an output event comparison quality index value according to the first possibility, the second possibility, the fourth possibility, and the fifth possibility, and determine an output important image information comparison quality index value according to the third possibility and the sixth possibility; Obtaining, according to the output event comparison quality index value and the output important image information comparison quality index value, an enhanced configuration quality index value corresponding to the enhanced configuration example; Wherein, determining the output event comparison quality index value according to the first possibility, the second possibility, the fourth possibility and the fifth possibility includes: Determine, according to the first possibility and the fourth possibility, a first output possibility ratio of the configured original image analysis thread and the event generation reference thread outputting the selected monitoring area event; According to the second possibility and the fifth possibility, determining a second output possibility ratio of the unselected monitoring area event output by the configured original image analysis thread and the event generation reference thread; Calculating an output event comparison quality index value based on the first output likelihood ratio and the second output likelihood ratio; Wherein, determining the comparison quality index value of the output important image information according to the third possibility and the sixth possibility includes: Determine the configured original image analysis thread and the event generation reference thread according to the third possibility and the sixth possibility, output a third output possibility ratio of important monitoring image information while outputting the selected monitoring area event, and output a fourth output possibility ratio of important monitoring image information while outputting the unselected monitoring area event; Based on the third output likelihood ratio and the fourth output likelihood ratio, an output important image information comparison quality index value is calculated.

9. The method according to claim 1, characterized in that: The method further comprises: Get the monitoring area topics to be processed; The monitoring area subject to be processed is input into the configured image analysis thread to obtain the monitoring area event corresponding to the monitoring area subject to be processed.

10. A surveillance key image recognition system based on drone, characterized in that: The invention comprises a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the method according to any one of claims 1 to 9.

Images