A method for improving image monitoring accuracy

Abstract

The application provides a method for improving image monitoring accuracy, comprising the following steps: a monitoring management subsystem acquires an object set Set_k_t reported by a certain video monitoring node Node_k in a video analysis subsystem at time t; objects with object characteristic values lower than a first threshold in Set_k_t are divided into an object subset LowSubset_k_t; from an object set A provided by a visit candidate subsystem, objects located within a radius L meters of Node_k at time point t+Delta_t are selected and defined as an object subset SubsetA; from SubsetA, objects most matched with objects in LowSubset_k_t are selected, and objects with matching values greater than or equal to a second threshold are defined as a subset B; and B is taken as an object successfully detected by Node_k at time t into a monitoring area of Node_k.

Description

[0001] This application is a divisional application, with parent application number 202011318828.3, application date 2020.11.23, and invention title: A method for improving the accuracy of image monitoring. Technical Field

[0002] This invention relates to the field of surveillance, and in particular to a method for improving the accuracy of image surveillance. Background Technology

[0003] With the rapid advancement of information and communication technologies and the rapid development of global professional division of labor, collaboration among business entities has become closer, mutual visits and exchanges have become more frequent, and the number of visits for business activities has increased exponentially. Correspondingly, due to the needs of personal safety management, information security management, target reconnaissance, and visit effectiveness management, it is necessary to conduct regional visit monitoring of a wide range of visitors in order to improve management effectiveness.

[0004] Typical scenarios include: in the power industry, personnel visits involve the safety management of visitors, requiring identification of whether visitors have entered dangerous areas for timely response and to avoid personal safety issues; in the high-tech industry, personnel visits involve information confidentiality and security issues, requiring identification of whether visitors have entered the confidential areas of core products and timely response to prevent the leakage of core technologies; in tourism, personnel visits involve the safety management of tourists after entering scenic spots, requiring identification of whether visitors have mistakenly entered restricted areas and timely handling to avoid personal safety issues for tourists; in commercial sales and customer referral effectiveness management, it is necessary to identify whether customers are visiting for the first time in order to confirm the market guidance of customers. All these scenarios require accurate information on whether personnel have entered the target monitoring area in order to take corresponding measures or make factual judgments and other related actions.

[0005] Regarding the monitoring of visited areas, existing technologies primarily employ the following approaches:

[0006] Video surveillance: By installing cameras in the target area, the cameras capture video and extract images for object recognition to identify visitors to the area (this solution is prone to misjudgment or missed detection due to facial angles, crowd obstruction, and facial decorations).

[0007] Entry registration (this method makes it difficult to ensure that all visitors are registered, leading to missed checks. In addition, this method is difficult to achieve refined management of visits within the area);

[0008] Registration via WIFI access authentication (this method is difficult to guarantee that all visitors will be registered, resulting in missed checks; in addition, this method is difficult to achieve refined management of visits within a region).

[0009] As analyzed above, due to the imperfections of existing monitoring technologies, there are problems such as the inability to achieve refined management of regional visits or low accuracy of visit monitoring, resulting in the failure of regional visit management. Therefore, proposing a method to improve the accuracy of image monitoring, thereby enhancing the level of refined management of regional visits and improving the accuracy of monitoring, is a problem that the industry needs to solve. Summary of the Invention

[0010] The technical problem to be solved by this invention is: This invention proposes a method to improve the accuracy of image monitoring, which realizes refined regional monitoring and management through video surveillance. When the object identification is blurred during the video surveillance process, at least one additional dimension is introduced to collaboratively verify the validity of the object's visit to the region, thereby improving the coverage and accuracy of monitoring, and finally realizing a refined, high-coverage, and high-accuracy regional visit monitoring solution, thereby improving the effectiveness of visit management.

[0011] The technical solution adopted by this invention to solve the problems existing in the prior art is:

[0012] This invention provides a method for improving the accuracy of image monitoring, comprising the following steps:

[0013] Step 1: The monitoring and management subsystem obtains the set of objects Set_k_t reported by a certain video monitoring node Node_k in the video analysis subsystem at time t;

[0014] Step 2: Divide objects in Set_k_t whose feature values ​​are lower than the first threshold into the object subset LowSubset_k_t;

[0015] Step 3: The monitoring and management subsystem selects objects located within a radius of L meters of Node_k at time point t+Delta_t from the object set A provided by the visiting candidate subsystem, and defines them as object subset SubsetA;

[0016] Step 4: The monitoring and management subsystem selects the object from SubsetA that best matches each object in LowSubset_k_t, and defines the objects with matching values ​​greater than or equal to the second threshold as subset B;

[0017] Step 5: The monitoring and management subsystem successfully detected B as an object entering the monitoring area of ​​node Node_k at time t;

[0018] Each element in the Set_k_t set includes at least three pieces of information: object ID, object content, and object feature value. The object feature value refers to a comprehensive detection value used to indicate the reliability of the existence of an object.

[0019] In step 3, the visitor selection subsystem locates the personnel's location distribution information in each area through wireless signals. The object set information provided by the visitor selection subsystem is the set of objects detected by the visitor selection subsystem with a radius G centered on the reference location after the monitoring and management subsystem provides the reference location information to the visitor selection subsystem. The reference location information is the center location of the monitoring and management subsystem's monitored area.

[0020] Preferably, in step 1, the monitoring and management subsystem obtains the object set Set_k_t reported by a certain video monitoring node Node_k in the video analysis subsystem at time t. The method for obtaining this set specifically includes:

[0021] After detecting the presence of an object, the video surveillance node Node_k proactively reports the set of detected objects;

[0022] or

[0023] The monitoring and management subsystem is configured to periodically report the set of objects detected by the video analytics subsystem;

[0024] or

[0025] After the monitoring and management subsystem sends a query request to the video analytics subsystem, the video analytics subsystem reports the set of detected objects.

[0026] Preferably, in step 1, the object ID refers to a number in the database used to uniquely identify the object, and the ID can also identify the corresponding monitoring area; the object content refers to the initial image information of the object.

[0027] Preferably, in step 5, the information written to the data includes at least the Node_k identifier, time information t, object ID, and object initial information.

[0028] Preferably, the wireless signal may include any one or a combination of several of the following: mobile communication signals, Bluetooth, WIFI, LoRa, and NB-IoT.

[0029] Preferably, in step 4, the construction process of subset B is as follows:

[0030] Step 4.1: Select any object LowSubset_k_t_i from LowSubset_k_t that has not yet been matched;

[0031] Step 4.2: Select any object SubsetA_j from SubsetA that has not yet been matched with LowSubset_k_t_i;

[0032] Step 4.3: Calculate the matching degree between LowSubset_k_t_i and SubsetA_j to obtain the matching value M_k_t_ij;

[0033] Step 4.4: Determine whether all objects in SubsetA match LowSubset_k_t_i. If yes, proceed to step 4.5; otherwise, proceed to step 4.2.

[0034] Step 4.5: Determine whether all objects in LowSubset_k_t have completed the matching operation. If yes, proceed to step 4.6; otherwise, proceed to step 4.1.

[0035] Step 4.6: Sort the matching values ​​of different j values ​​within the same i in M_k_t_ij from high to low to obtain PM_k_t_iy;

[0036] Step 4.7: Select the PM_k_t_p0 with the highest matching value under different i from PM_k_t_i0, and determine whether it is greater than or equal to the second threshold. If not, go to step 4.9. If yes, write the object information of SubsetA corresponding to PM_k_t_p0 into Subset B and delete the information in the PM_k_t_py column.

[0037] Step 4.8: Determine whether the object in SubsetA corresponding to PM_k_t_p0 appears in PM_k_t_i0. If it does, delete the object in PM_k_t_iy, where i is not equal to p.

[0038] Step 4.9: End the construction of subset B.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The method of this invention enables refined regional monitoring and management through video surveillance. When object identification is unclear during video surveillance, at least one additional dimension is introduced to collaboratively verify the validity of the object's visit to the region, thereby improving the coverage and accuracy of monitoring. Ultimately, a refined, high-coverage, and high-accuracy regional visit monitoring solution is achieved, improving the effectiveness of visit management. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0042] Figure 2 This is a schematic diagram of the system composition of the present invention. Detailed Implementation

[0043] To make the technical solution and beneficial effects of the present invention clearer, the embodiments of the present invention will be explained in further detail below.

[0044] like Figure 2 As shown, this invention provides a multi-dimensional collaborative monitoring system, including: a monitoring management subsystem, a visitor backup subsystem, and a video analysis subsystem. The functions of each subsystem are as follows:

[0045] Video analytics subsystem: This subsystem is responsible for video surveillance, image extraction, object detection, and reporting the object detection results to the monitoring management subsystem;

[0046] Visit alternative subsystems: Based on the coordinates submitted by the monitoring and management subsystem, obtain the objects within a radius G centered at those coordinates and send the information to the monitoring and management subsystem;

[0047] Monitoring and Management Subsystem: Based on the information provided by the video analysis subsystem and the visit alternative subsystem, object detection is performed using a multi-dimensional collaborative monitoring method, and the detection results are written into the database within the monitoring and management subsystem.

[0048] This invention also provides a multi-dimensional collaborative monitoring method, such as... Figure 1 As shown, the specific steps include:

[0049] Step 1: The monitoring and management subsystem obtains the set of objects Set_k_t reported by a certain video monitoring node Node_k in the video analysis subsystem at time t. Each element in the set of objects includes at least three pieces of information: object ID, object content, and object feature value.

[0050] Step 2: The monitoring and management subsystem divides objects in Set_k_t whose feature values ​​are lower than the threshold 1 into the object subset LowSubset_k_t, and objects whose feature values ​​are higher than or equal to the threshold 1 into the object subset HighSubset_k_t;

[0051] Step 3: The monitoring and management subsystem selects objects located within a radius of L meters of Node_k at time point t+Delta_t from the object set A provided by the visiting candidate subsystem, and defines them as object subset SubsetA;

[0052] Step 4: The monitoring and management subsystem selects the object from SubsetA that best matches each object in LowSubset_k_t, and defines the objects with matching values ​​greater than or equal to the threshold 2 as subset B;

[0053] Step 5: The monitoring and management subsystem records HighSubset_k_t and B as objects that successfully entered the monitoring area of ​​node Node_k at time t, and writes them to the database.

[0054] The method for improving the accuracy of image monitoring involved in steps 1, 2, 3, 4, and 5 consists of a monitoring management subsystem, a visitor backup subsystem, and a video analysis subsystem.

[0055] In step 1, the monitoring and management subsystem obtains the object set Set_k_t reported by a certain video monitoring node Node_k in the video analysis subsystem at time t. The method for obtaining this set specifically includes:

[0056] After detecting the presence of an object, the video surveillance node Node_k proactively reports the set of detected objects;

[0057] or

[0058] The monitoring and management subsystem is configured to periodically report the set of objects detected by the video analytics subsystem;

[0059] or

[0060] After the monitoring and management subsystem sends a query request to the video analysis subsystem, the video analysis subsystem reports the set of detected objects.

[0061] In step 1, the object ID refers to the number used to uniquely identify the object in the database. Preferably, the ID can also identify the corresponding monitoring area. The object content refers to the initial image information of the object. The object feature value refers to the comprehensive detection value used to indicate the reliability of the object's existence. Typical comprehensive detection values ​​include the three courts and five eyes comprehensive detection value, but there is no specific limitation.

[0062] In step 3, the visitor selection subsystem locates the personnel's location distribution information in each area through wireless signals. The wireless signals may include mobile communication signals, Bluetooth, WIFI, LoRa, NB-IoT, etc., and are not specifically limited.

[0063] In step 3, the object set information provided by the visiting candidate subsystem is the object set detected by the visiting candidate subsystem with the reference location as the center and a radius G after the monitoring and management subsystem provides the reference location information to the visiting candidate subsystem. The reference location information is the center location of the monitoring and management subsystem's monitored area.

[0064] In step 4, the construction process of subset B is as follows:

[0065] Step 4.1: Select any object LowSubset_k_t_i from LowSubset_k_t that has not yet been matched;

[0066] Step 4.2: Select any object SubsetA_j from SubsetA that has not yet been matched with LowSubset_k_t_i;

[0067] Step 4.3: Calculate the matching degree between LowSubset_k_t_i and SubsetA_j to obtain the matching value M_k_t_ij;

[0068] Step 4.4: Determine whether all objects in SubsetA match LowSubset_k_t_i. If yes, proceed to step 4.5; otherwise, proceed to step 4.2.

[0069] Step 4.5: Determine whether all objects in LowSubset_k_t have completed the matching operation. If yes, proceed to step 4.6; otherwise, proceed to step 4.1.

[0070] Step 4.6: Sort the matching values ​​of different j values ​​within the same i in M_k_t_ij from high to low to obtain PM_k_t_iy;

[0071] Step 4.7: Select the PM_k_t_p0 with the highest matching value under different i from PM_k_t_i0, and determine whether it is greater than or equal to the threshold 2. If not, go to step 4.9. If yes, write the object information of SubsetA corresponding to PM_k_t_p0 into Subset B and delete the information in the PM_k_t_py column.

[0072] Step 4.8: Determine whether the object in SubsetA corresponding to PM_k_t_p0 appears in PM_k_t_i0 (i is not equal to p). If it does, delete the object in PM_k_t_iy.

[0073] Step 4.9: Complete the construction of subset B;

[0074] In step 5, the information written to the data includes at least the Node_k identifier, time information t, object ID, and object initial information.

[0075] The following specific examples describe a method for improving the accuracy of image monitoring:

[0076] Example: Figure 2As shown, the system in this embodiment consists of a monitoring and management subsystem, a visitor alternative subsystem, and a video analysis subsystem. Threshold 1 is set to 0.8, radius L is set to 60 meters, Delta_t is set to 30 seconds, radius G is set to 300 meters, and threshold 2 is set to 0.7. At time t, the Node_k node in the video analysis subsystem reports the object detection results shown in Table 1 to the monitoring and management subsystem. It can be seen from Table 1 that only the object with object ID 2 is greater than threshold 1, that is, it has been determined that the object entered the Node_k monitoring area at time t. Therefore, the object with object ID 2 in Table 1 is assigned to HighSubset_k_t (corresponding to HighSubset_k_t_0). As for the two objects with object IDs 0 and 1, since the detection results are less than threshold 1, there is ambiguity, so it cannot be confirmed whether they have entered the Node_k monitoring area. Therefore, the two objects with object IDs 0 and 1 in Table 1 are assigned to LowSubset_k_t (corresponding to LowSubset_k_t_0 and LowSubset_k_t_1 respectively).Because the detection is ambiguous and the result cannot be confirmed temporarily, the present invention introduces an additional dimension for collaborative judgment. The monitoring and management subsystem sends the reference location information RP under its jurisdiction to the visiting candidate subsystem. The visiting candidate subsystem performs positioning based on mobile communication signals and reports the candidate object set A, located within a radius G (i.e., 300 meters) of the reference location RP, to the monitoring and management subsystem. Details of the candidate object set A are shown in Table 2. Next, the monitoring and management subsystem filters out time points t+Delta_t that are located within a radius L meters of Node_k. The system calculates the matching degree between LowSubset_k_t_0 and SubsetA_f (where f takes values ​​of 0, 1, ..., 4) and then between LowSubset_k_t_1 and SubsetA_f (where f takes values ​​of 0, 1, ..., 4). The result is then used to calculate the matching degree between LowSubset_k_t_1 and SubsetA_f (where f takes values ​​of 0, 1, ..., 4). The final result is M_k_t_ij (where i... The values ​​of i and j are 0 and 1, respectively. Then, the matching degree corresponding to different j values ​​under the same i value in M_k_t_ij is sorted from high to low, resulting in the matching degree sorting results shown in Table 3. Next, the maximum value, PM_k_t_00, is found between PM_k_t_00 and PM_k_t_10. This value is determined to be greater than the threshold 2 (corresponding to a value of 0.7). Therefore, the object corresponding to it in SubsetA_j, i.e., SubsetA_0, is written into SubsetB, and the data in the PM_k_t_0y column is deleted. Simultaneously, it is determined that the PM_k_t_10 column contains... Whether SubsetA_0 is included or not. Since this real-time example includes it, the 0th element of PM_k_t_1y needs to be deleted, resulting in the result in Table 4. Then, PM_k_t_10 is selected from Table 4, and it is determined that it is greater than the threshold 2 (corresponding to a value of 0.7). Therefore, the object corresponding to it in SubsetA_j, namely SubsetA_1, is written into Subset B. Finally, the monitoring and management subsystem records HighSubset_k_t and B as objects that Node_k successfully detected entering the monitoring area of ​​node Node_k at time t, and writes them into the database. See Table 5 for details after the database update.

[0077] Table 1. Information on the detected objects reported by the video analysis subsystem.

[0078]

[0079] Table 2 Calculation results of M_k_t_ij

[0080]

[0081] Table 3 Calculation results of PM_k_t_iy

[0082]

[0083] Table 4 shows the element composition of PM_k_t_iy after deleting the relevant elements from the first multidimensional collaborative matching.

[0084]

[0085]

[0086] Table 5 Database Information

[0087] Node_k identifier Time information Object ID object initial information k t 0 Image 0 k t 1 Image 1 k t 2 Image 2

[0088] The method of this invention enables refined regional monitoring and management through video surveillance. When object identification is unclear during video surveillance, at least one additional dimension is introduced to collaboratively verify the validity of the object's visit to the region, thereby improving the coverage and accuracy of monitoring. Ultimately, a refined, high-coverage, and high-accuracy regional visit monitoring solution is achieved, improving the effectiveness of visit management.

[0089] In summary, the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the invention. Based on the above description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of the present invention is not limited to the contents of the specification; all equivalent variations and modifications of the shape, structure, features, and spirit described within the scope of the claims should be included within the scope of the claims.

Claims

1. A method of improving image monitoring accuracy, characterized by, The steps include the following: Step 1: the monitoring management subsystem obtains an object set Set_k_t reported by a certain video monitoring node Node_k in the video analysis subsystem at time t; Step 2: objects with feature values lower than a first threshold in Set_k_t are divided into an object subset LowSubset_k_t; Step 3: the monitoring management subsystem selects objects located within a radius L meters of Node_k at a time point t+Delta_t from an object set A provided by the visit candidate subsystem, and defines the objects as an object subset SubsetA; Step 4: the monitoring management subsystem selects objects most matched with objects in LowSubset_k_t from SubsetA, and defines objects with matching values greater than or equal to a second threshold as a subset B; Step 5: the monitoring management subsystem regards B as an object successfully detected by Node_k at time t into the monitoring area of Node_k; Each element in the object set Set_k_t includes at least object ID, object content, and object feature value, and the object feature value refers to a comprehensive detection value for indicating the reliability of the existence of the object, and the calculation method of the object feature value is as follows: Obtain detection values of each sub-item of the object three courts five eyes, including a detection value M from hairline to brow bone, a detection value N from brow bone to nose bottom, a detection value P from nose bottom to chin, and a five-eye detection value K; Set sub-item detection values less than 0.6 to 0, and record the number F of sub-items with detection values less than 0.6; The calculation method of the object feature value S is S=(M+N+P+K) / (4-F)-F*0.1; In step 3, the visit candidate subsystem detects the position distribution information of personnel in each area through wireless signal positioning, and the object set information provided by the visit candidate subsystem is based on the reference position information provided by the monitoring management subsystem to the visit candidate subsystem, and the object set detected by the visit candidate subsystem within a radius G range centered on the reference position, and the reference position information is the center position of the monitoring area of the monitoring management subsystem; In step 4, the construction process of the subset B is as follows: Step 4.1: select any object LowSubset_k_t_i in LowSubset_k_t that has not been matched; Step 4.2: select any object SubsetA_j in SubsetA that has not been matched with LowSubset_k_t_i; Step 4.3: calculate the matching degree of LowSubset_k_t_i and SubsetA_j to obtain a matching value M_k_t_ij; Step 4.4: determine whether all objects in SubsetA have completed matching with LowSubset_k_t_i, if yes, go to step 4.5, and if no, go to step 4.2; Step 4.5: determine whether all objects in LowSubset_k_t have completed matching, if yes, go to step 4.6, and if no, go to step 4.1; Step 4.6: determine whether all objects in LowSubset_k_t have completed matching, if yes, go to step 4.6, and if no, go to step 4.1; Step 4.6, sort the matching values of different j in the same i in M_k_t_ij from high to low to get PM_k_t_iy; Step 4.7, select the highest matching value PM_k_t_p0 from different i in PM_k_t_i0, and determine whether it is greater than or equal to the second threshold. If not, jump to step 4.

9. If yes, write the object information of SubsetA corresponding to PM_k_t_p0 into SubsetB, and delete the column information of PM_k_t_py; Step 4.8, determine whether the object in SubsetA corresponding to PM_k_t_p0 appears in PM_k_t_i0. If it appears, delete the object in PM_k_t_iy, i not equal to p; Step 4.9, end the construction of SubsetB.

2. The method for improving the accuracy of image monitoring according to claim 1, characterized in that: In step 1, the monitoring management subsystem obtains the object set Set_k_t reported by a video monitoring node Node_k in the video analysis subsystem at time t, and the obtaining method specifically includes: The video monitoring node Node_k actively reports the detected object set after detecting the existence of the object; Or, The monitoring management subsystem configures the video analysis subsystem to periodically report the detected object set; Or, The monitoring management subsystem sends a query request to the video analysis subsystem, and the video analysis subsystem reports the detected object set.

3. The method for improving the accuracy of image monitoring according to claim 1, characterized in that: In step 1, the object ID refers to a number used to uniquely identify the object in the database, and is also used to identify the corresponding monitoring area; the object content refers to the initial image information of the object.

4. The method for improving the accuracy of image monitoring according to claim 1, characterized in that: In step 5, the written data includes at least Node_k identification, time information t, object ID, and object initial information.

5. The method for improving the accuracy of image monitoring according to claim 1, characterized in that: The wireless signal can include any one or combination of several of mobile communication signal, Bluetooth, WIFI, LORA, and NBIOT.

Images