Anomaly identification method, storage medium and device based on Kylin operating system

Abstract

The present invention relates to the field of data processing technology, and in particular to an anomaly identification method, storage medium and device based on the Kylin operating system. It includes: obtaining the image marking information of the object to be tested through a target model, if the object category identifier in the current image marking information is different from the object category identifier of the last image marking information in the target storage sequence, then correcting the current image marking information to generate the corresponding corrected image marking information; to generate the information sequence to be tested corresponding to the object to be tested; according to the information corresponding to the current detection time sliding window in the information sequence to be tested, the abnormal information of the object to be tested is generated. In the present invention, the object category identifier in the current image marking information is corrected by the category confidence corresponding to different object category identifiers in the historical image marking information, so as to further ensure the accuracy of each image marking information in the finally generated information sequence to be tested, and then ensure the accuracy of the finally generated abnormal information.

Description

Background Art

[0002] With the development and progress of technology, AI technology has been introduced in most scenarios to assist operations. For example, in some illegal behavior detection scenarios, the target image recognition model can identify each target to be detected in the marked target area, and the target image tracking model can continuously track each target to be detected. Based on the above information, the behavior of each target to be detected can be detected and judged.

[0003] However, since the target image recognition model and the target image tracking model will inevitably generate erroneous image tag information during the process of generating image tag information, the accuracy of abnormal behavior determination will be reduced. Summary of the invention

[0004] In view of the above technical problems, the technical solution adopted by the present invention is:

[0005] According to one aspect of the present invention, a method for identifying anomalies based on a Kylin operating system is provided, the method comprising the following steps:

[0006] The image tag information of the object to be tested is obtained through the target model; the target model runs on the Kylin operating system; wherein the i-th image tag information A obtained i =(A i 1 , A i 2 , P i ), A i 1 A i The object ID in A i 2 A i The object category identification in P i Generate A for the target model i 2 The corresponding category confidence;

[0007] If the object ID in the current image marking information is the same as the object ID corresponding to any historical information storage sequence, the historical information storage sequence is used as the target storage sequence corresponding to the current image marking information;

[0008] If the object category identifier in the current image marking information is different from the object category identifier of the last image marking information in the target storage sequence, the current image marking information is corrected to generate corresponding corrected image marking information;

[0009] Adding the corrected image marking information to the target storage sequence to generate a sequence of information to be tested corresponding to the object to be tested;

[0010] Generate abnormal information of the object to be measured according to the information corresponding to the current detection time sliding window in the information sequence to be measured;

[0011] The correction process includes:

[0012] If P i > Y1, then use A i as the corresponding corrected image marking information; Y1 is the first confidence threshold;

[0013] If P i < Y1, then from the target storage sequence corresponding to A i obtain the K historical image marking information that is the closest in chronological order to A i ;

[0014] According to A i and the K historical image marking information, generate the first correction confidence B i 2 corresponding to A i 1 ; B i 1 satisfies the following conditions:

[0015] B i 1 = Max(P avg 1 , P avg 2 , …, P avg a , …, P avg f(i) )

[0016] where P avg a is the average class confidence corresponding to the a-th object class identifier in A i and the K historical image marking information; f(i) is the total number of types of object class identifiers in A i and the K historical image marking information, a = 1, 2, …, f(i);

[0017] Replace A i 2 with the object class identifier corresponding to B i 1 to generate the corrected image marking information corresponding to A i ;

[0018] According to a second aspect of the present invention, a non-transitory computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the above-mentioned exception identification method based on the Kylin operating system is implemented.

[0019] According to a third aspect of the present invention, there is provided an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned abnormality identification method based on the Kylin operating system when executing the computer program.

[0020] The present invention has at least the following beneficial effects:

[0021] The target model in the present invention includes a target image recognition model and a target image tracking model, and thus, each object to be tested in the target area can be continuously marked with information through the target model. Then, an image marking information sequence corresponding to each object to be tested is formed, that is, the information sequence to be tested corresponding to the object to be tested. The image marking information in the information sequence to be tested can reflect the behavior information of the target object to be tested in a certain period of time, and thus the corresponding detection time sliding window can be set to perform abnormal detection and judgment on the corresponding information in the information sequence to be tested in turn, and then generate abnormal information corresponding to the object to be tested.

[0022] In addition, when the recognition result output by the target model has a certain error rate, the same object to be tested may be misidentified as different object category identifiers in the image tag information. In the present invention, the object category identifier in the current image tag information is corrected by the category confidence corresponding to the different object category identifiers in the historical image tag information, so as to further ensure the accuracy of each image tag information in the final generated information sequence to be tested, and thus ensure the accuracy of the final generated abnormal information. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A flowchart of an abnormality identification method based on Kylin operating system provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of marking image marking information in an image frame of an actual scene provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0027] According to a first aspect of the present invention, Figure 1 As shown, a method for identifying anomalies based on the Kylin operating system is provided, and the method comprises the following steps:

[0028] S100: Obtain image tag information of the object to be tested through the target model. The target model runs on the Kylin operating system. Among them, the i-th image tag information A obtained i =(A i 1 , A i 2 , P i ), A i 1 A i The object ID in A i 2 A i The object category identification in P i Generate A for the target model i 2 The corresponding category confidence.

[0029] The target model can be an existing neural network model, which specifically needs to include a target image recognition model based on image recognition and a target image tracking model. When anomaly detection of related behaviors is required in a certain area, a camera needs to be installed first to obtain images of each object in the area, and then each frame of the image is input into the target model to obtain the image tag information corresponding to each object in the image. Figure 2 As shown in the figure, a unique object ID and object category identification corresponding to each object are obtained, such as the object ID identification is 201 and the object category identification is aircraft. At the same time, the target model will also configure a rectangular target recognition box corresponding to the selected object image for each identified object, such as the rectangular box outside the aircraft. At the same time, since the target model is a trained model for prediction, the category confidence corresponding to the object category identification will be output.

[0030] Typically, as time goes by, a series of image information of each object to be detected in the target area is acquired, and then a plurality of image marking information of each object to be detected is obtained, and then an information storage sequence of each object to be detected is generated.

[0031] S200: If the object ID in the current image marking information is the same as the object ID corresponding to any historical information storage sequence, the historical information storage sequence is used as the target storage sequence corresponding to the current image marking information.

[0032] S300: If the object category identifier in the current image marking information is different from the object category identifier of the last image marking information in the target storage sequence, the current image marking information is corrected to generate corresponding corrected image marking information.

[0033] Whenever a new image tag information (current image tag information) is acquired, the target storage sequence corresponding to the current image tag information is first found through the object ID identification, and then the object category identification in the current image tag information is verified to be correct. Since the target image recognition model has a certain probability of misjudgment, it is easier to misjudge the object category identification, resulting in multiple object category identifications for the same object. For example, taking the identification of abnormal behavior in the parking area of ​​the airport as an example, there are usually multiple operating vehicles in the parking area. Due to the large similarity of the appearance of the vehicles, and due to the influence of light or the obstruction of other objects, the image information of the vehicle obtained will be distorted when the vehicle passes through a dark area or an area obstructed by other objects. This will cause the target image recognition model to misjudge the type of vehicle.

[0034] S400: Adding the corrected image mark information to the target storage sequence to generate a sequence of information to be tested corresponding to the object to be tested.

[0035] S500: Generate abnormal information of the object to be detected according to the information corresponding to the current detection time sliding window in the information sequence to be detected.

[0036] After the correction process, the erroneous information can be corrected, and more accurate image marking information can be generated to generate a more accurate information sequence to be tested. In actual use, the corresponding objects in the target area can be marked according to the information in the information sequence to be tested. At the same time, the corresponding detection time sliding window is set to perform abnormal detection and judgment on the corresponding information in the information sequence to be tested in turn, and then the abnormal information corresponding to the object to be tested can be generated in real time, so as to generate alarm information more timely.

[0037] Corrective treatment includes:

[0038] S301: If P i >Y1, then A iAs the corresponding corrected image marking information. Y1 is the first confidence threshold. Y1 can be 0.9. When the corresponding class confidence is greater than 0.9, it means that the model has sufficient confidence to identify the class of the object, and it is possible that there is an error in the previous recognition result. Therefore, at this time, the new image marking information will not be corrected, and the result will be directly output.

[0039] S302: If P i < Y1, then from A i In the corresponding target storage sequence, obtain the K historical image marking information that is the closest in chronological order to A i Recently.

[0040] S303: According to A i And the K historical image marking information, generate A i 2 The corresponding first corrected confidence B i 1 . Usually, K is selected as an integer between 9 and 60. It is determined according to the actual situation. The determined K historical image marking information can represent the information of the object in a historical period. Usually, the length of this historical period is greater than the time when the object passes through the dark area or greater than the time when the object is occluded.

[0041] B i 1 Satisfies the following conditions:

[0042] B i 1 = Max(P avg 1 , P avg 2 , …, P avg a , …, P avg f(i) )

[0043] Among them, P avg a Is the mean value of the class confidence corresponding to the a-th object class identifier in A i And the K historical image marking information. f(i) is the total number of types of object class identifiers in A i And the K historical image marking information. a = 1, 2, …, f(i).

[0044] S304: Replace A i 2 With B i 1 The corresponding object class identifier, and generate A i The corresponding corrected image marking information.

[0045] Since, when a misjudgment occurs, the corresponding category confidence is usually low, basically between 0.2-0.6. The category confidence corresponding to the multiple correct judgment results close to it is high, basically between 0.6-0.9. Therefore, the category confidence mean can be used to determine whether a misjudgment has occurred, and Max(P avg 1 , P avg 2 ,…,P avg a ,…,P avg f (i) ) to correct the current image marking information.

[0046] In the present invention, the object category identification in the current image marking information is corrected by the category confidence corresponding to different object category identifications in the historical image marking information, so as to further ensure the accuracy of each image marking information in the finally generated information sequence to be tested, and thus ensure the accuracy of the finally generated abnormal information.

[0047] Furthermore, the image marking information is image information, and the image marking information also includes target position information. The target position information in the image marking information is generated according to the following steps:

[0048] S110: Obtaining position information of two diagonal points of a rectangular target recognition frame corresponding to the object to be detected in the image information.

[0049] S120: Generate the center position information of the rectangular object recognition frame according to the position information of the two diagonal points.

[0050] S130: Using the center position information of the rectangular target recognition frame as the target position information in the image marking information of the object to be detected.

[0051] After S100, the method further includes:

[0052] S310: If the current image marking information is different from the object IDs corresponding to all the historical information storage sequences, the current image marking information is corrected to generate corresponding corrected image marking information.

[0053] Similarly, since the target detection model has a certain misjudgment rate, it is possible that misjudgment will occur when generating object IDs, which may result in multiple object IDs appearing for one object. For example, when an object appears again after being blocked, the target image tracking model may consider the reappearing object to be a new target, so it will be marked with a new object ID. For example, the object ID corresponding to working vehicle A is 001. After being blocked and reappearing, the object ID corresponding to working vehicle A is 002.

[0054] The following correction processing is performed on the above-mentioned misjudged information, and the correction processing includes:

[0055] S311: acquiring an initial correction sequence from a plurality of historical information storage sequences according to target position information, wherein the distance between the last historical image mark information in the initial correction sequence and the current image mark information is less than a first distance threshold.

[0056] The first distance threshold needs to be configured accordingly according to the moving speeds of different types of objects.

[0057] S312: Obtain the most recent m historical image tag information from each initial correction sequence, and generate a target correction sequence corresponding to each initial correction sequence. The value of M is generally 10.

[0058] Since different objects usually move in different directions in the image, and their corresponding moving speeds are also different, the lengths of the moving trajectories generated by different objects in the last few frames of images are also different. Generally speaking, the moving trajectory of the same object is continuous, so the length of the trajectory between consecutive frames is usually the shortest.

[0059] At the same time, when the same object appears again, its corresponding object category identifier is the same as before, and the object category identifier is also likely to be different from the categories of other objects that appear around it.

[0060] S313: Generate a second correction confidence of each target correction sequence for the current image marking information according to the historical image marking information in the target correction sequence. Among them, the second correction confidence B corresponding to the bth target correction sequence is b 2 The following conditions must be met:

[0061]

[0062] Among them, P n bis the class confidence of the nth historical image marking information in the bth target correction sequence that has the same object class identifier as the current image marking information. Q is the total number of historical image marking information in the bth target correction sequence that has the same object class identifier as the current image marking information, and Q ≤ m. L r is the distance between the rth historical image marking information and the current image marking information in the bth target correction sequence. C r is L r The corresponding weight coefficient. The closer the generation time of the historical image marking information is to the current image marking information, the larger the corresponding weight coefficient. S is the total number of historical image marking information participating in the distance calculation in the bth target correction sequence, and S < m.

[0063] S314: If Max(B1 2 , B2 2 , …, B b 2 , …, B t 2 ) > Y2, then replace the object ID identifier in the current image marking information with the object ID identifier corresponding to Max(B1 2 , B2 2 , …, B b 2 , …, B t 2 ). Y2 = 0.8.

[0064] In the above steps, according to the similarity between the current image marking information and the object corresponding to the target correction sequence can be determined. According to the proximity of the distance between the current image marking information and the object corresponding to the target correction sequence can be determined. Based on the information in these two dimensions, the target object most similar to the object corresponding to the current image marking information can be determined, and then the object ID identifier of the current image marking information can be corrected more accurately.

[0065] S315: If Max(B1 2 , B2 2 , …, B b 2 , …, B t 2 ) < Y3, then create a corresponding information storage sequence for the current image marking information. Y3 = 0.3

[0066] When Max(B1 2 , B2 2 , …, B b 2 , …, B t 2) When Y3, it indicates that the similarity between the object corresponding to the current image marking information and all other existing objects is relatively low, and it should be a completely new object. Therefore, a new information storage sequence needs to be established to store the relevant image marking information of this new object.

[0067] As another possible embodiment of the present invention, after S400: adding the corrected image marking information to the target storage sequence, the method further includes:

[0068] S410: If the distance between the corrected image marking information and the adjacent historical image marking information in the target storage sequence does not belong to the reasonable distance interval, correct the rectangular target recognition frame in the corrected image marking information.

[0069] Since, usually, the rectangular target recognition frame of an object is circumscribed outside the image of the object, and thus can completely enclose the image of the object. As Figure 2 shown by the solid rectangular box corresponding to the aircraft. However, due to the misjudgment of the target image recognition model, the rectangular target recognition frame of the object may only frame a partial area of the image of the object. Such as only framing half of the area. As Figure 2 shown by the dashed rectangular box corresponding to the aircraft.

[0070] The reasonable distance interval [D1, D2] is determined according to the following steps:

[0071] The first side point D1 in the reasonable distance interval satisfies the following conditions:

[0072] D1 = E + 0.8 * D avg .

[0073] The second side point D2 in the reasonable distance interval satisfies the following conditions:

[0074] D2 = E + 1.2 * D avg .

[0075] Wherein, E is the target position information of the historical image marking information adjacent to the corrected image marking information in the target storage sequence. D avg is the average position offset between adjacent image marking information in the target storage sequence.

[0076] On the one hand, correcting the rectangular target recognition frame in the corrected image marking information includes:

[0077] S411: Re-input the image information corresponding to the corrected image marking information into the target model to re-generate the rectangular target recognition frame corresponding to each object to be measured in the image information.

[0078] On the other hand, correcting the rectangular target recognition frame in the corrected image marking information includes:

[0079] S421: Obtain target position information in a plurality of historical image marking information located before the revised image marking information in the target storage sequence.

[0080] S422: Generate an average position offset between two adjacent image mark information according to the multiple target position information.

[0081] S423: Generate offset direction information between the corrected image mark information and adjacent historical image mark information according to the plurality of target position information.

[0082] S424: Generate a rectangular target recognition frame corresponding to each object to be detected in the corrected image mark information according to the average position offset and the offset direction information.

[0083] Specifically, the rectangular target recognition frame of the corresponding object in the historical image marking information adjacent to the corrected image marking information is used to generate the corner point position information of the two diagonal corners of a new rectangular target recognition frame according to the average position offset and offset direction information. Then, based on the corner point position information of the two diagonal corners, a rectangular target recognition frame corresponding to the object to be measured is generated in the corrected image marking information.

[0084] According to the above, the steps can be used to infer the future motion trajectory of the object based on multiple historical image marking information, and then the specific position of the object in the corrected image marking information can be inferred. In addition, the interval between two adjacent image marking information is usually very short, so the shape and size of the rectangular target recognition frame in two adjacent image frames of the same object are basically the same. Therefore, the offset information of the rectangular target recognition frame can be obtained by the steps in this embodiment, and the rectangular target recognition frame of the corresponding object in the historical image marking information adjacent to the corrected image marking information can be adjusted through the offset information, so as to correct the rectangular target recognition frame of the corresponding object to be measured in the corrected image marking information.

[0085] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0086] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0087] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0088] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as a "circuit", "module" or "system".

[0089] The electronic device according to this embodiment of the present invention is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0090] The electronic device is presented in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: the at least one processor mentioned above, the at least one storage device mentioned above, and a bus connecting different system components (including storage devices and processors).

[0091] The storage stores program codes, which can be executed by the processor, so that the processor executes the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification.

[0092] The memory may include readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read only memory (ROM).

[0093] The storage may also include a program / utility having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include an implementation of a network environment.

[0094] The bus may represent one or more of several types of bus structures including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0095] The electronic device may also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may communicate with one or more devices that enable a user to interact with the electronic device, and / or may communicate with any device (e.g., routers, modems, etc.) that enables the electronic device to communicate with one or more other computing devices. Such communication may be performed through an input / output (I / O) interface. Furthermore, the electronic device may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through a network adapter. The network adapter communicates with other modules of the electronic device through a bus. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0096] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0097] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present invention may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present specification.

[0098] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0099] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0100] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0101] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0102] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0103] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0104] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An abnormality identification method based on Kylin operating system, characterized in that: The method comprises the following steps: S100: Obtaining image tag information of the object to be tested through a target model; the target model runs on a Kylin operating system; wherein the i-th image tag information A obtained i =(A i 1 , A i 2 , P i ), A i 1 A i The object ID in A i 2 A i The object category identification in P i Generate A for the target model i 2 The corresponding category confidence; S200: If the object ID in the current image marking information is the same as the object ID corresponding to any historical information storage sequence, the historical information storage sequence is used as the target storage sequence corresponding to the current image marking information; S300: if the object category identifier in the current image marking information is different from the object category identifier of the last image marking information in the target storage sequence, correct the current image marking information to generate corresponding corrected image marking information; S400: adding the corrected image marking information to the target storage sequence to generate a sequence of information to be tested corresponding to the object to be tested; S500: Generate abnormal information of the object to be detected according to the information corresponding to the current detection time sliding window in the information sequence to be detected; The correction process includes: If P i >Y1, then A i As the corresponding corrected image marking information; Y1 is the first confidence threshold; If P i < is less than Y1, then from A i corresponding to the target storage sequence, obtain the K historical image marking information that is closest in chronological order to A i in the most recent; According to A i And K historical image label information, generate A i 2 The corresponding first modified confidence B i 1 ; B i 1 The following conditions must be met: B i 1 =Max(P avg 1 、P avg 2 、…、P avg a 、…、P avg f(i) ) Among them, P avg a For A i And the category confidence mean corresponding to the a-th object category identification in the K historical image label information; f(i) is the category confidence mean corresponding to the a-th object category identification in A i and the total number of object category identifiers in the K historical image tag information, a=1, 2, …, f(i); A i 2 Replace with B i 1 The corresponding object category identification generates A i The corresponding corrected image tag information.

2. The method according to claim 1, characterized in that: The image marking information also includes target location information; After acquiring the image marking information of the object to be measured through the target model, the method further includes: If the current image tag information is different from the object IDs corresponding to all the historical information storage sequences, the current image tag information is corrected to generate corresponding corrected image tag information; The correction process includes: According to the target position information, an initial correction sequence is obtained from the plurality of historical information storage sequences; the distance between the last historical image mark information in the initial correction sequence and the current image mark information is less than a first distance threshold; Obtain the most recent m historical image labeling information from each initial correction sequence to generate a target correction sequence corresponding to each initial correction sequence; According to the historical image marking information in the target correction sequence, the second correction confidence of each target correction sequence for the current image marking information is generated; wherein the second correction confidence B corresponding to the bth target correction sequence is b 2 The following conditions must be met: Among them, P n b is the category confidence of the nth historical image labeling information with the same object category identifier as the current image labeling information in the bth target correction sequence; Q is the total number of historical image labeling information with the same object category identifier as the current image labeling information in the bth target correction sequence, Q≤m; L r The distance between the rth historical image label information and the current image label information in the bth target correction sequence; C r For L r The corresponding weight coefficient, the closer the generation time of the historical image tag information is to the current image tag information, the larger the corresponding weight coefficient; S is the total number of historical image tag information involved in the distance calculation in the b-th target correction sequence, S <m; If Max(B1 2 、B2 2 , …, B b 2 , …, B t 2 )>Y2, then the object ID in the current image tag information is replaced by Max (B1 2 、B2 2 , …, B b 2 , …, B t 2 ) corresponding to the object ID.

3. The method according to claim 1, characterized in that After generating a second correction confidence of each target correction sequence to the current image marking information according to the historical image marking information in the target correction sequence, the method further includes: If Max(B1 2 , B2 2 , …, B b 2 , …, B t 2 ) < Y3, a corresponding information storage sequence is newly created for the current image marking information.

4. The method according to claim 2, characterized in that: The image marking information is image information, and the target position information in the image marking information is generated according to the following steps: Obtain the position information of two diagonal points of the rectangular target recognition frame corresponding to the object to be detected in the image information; Generate the center position information of the rectangular target recognition frame according to the position information of the two diagonal points; The center position information of the rectangular target recognition frame is used as the target position information in the image marking information of the object to be detected.

5. The method according to claim 4, characterized in that After adding the modified image mark information to the target storage sequence, the method further includes: If the distance between the corrected image mark information and the adjacent historical image mark information in the target storage sequence does not belong to a reasonable spacing interval, the rectangular target recognition frame in the corrected image mark information is corrected; The reasonable spacing interval [D1, D2] is determined according to the following steps: The first edge point D1 in the reasonable spacing interval satisfies the following conditions: D1=E+0.8×D avg ; The second edge point D2 in the reasonable spacing interval satisfies the following conditions: D2=E+1.2×D avg ; Wherein, E is the target position information of the historical image mark information adjacent to the modified image mark information in the target storage sequence; D avg The target stores the average position offset between adjacent image marker information in the sequence.

6. The method according to claim 5, characterized in that Correcting the rectangular target recognition frame in the corrected image mark information includes: The image information corresponding to the corrected image marking information is re-inputted into the target model to regenerate a rectangular target recognition frame corresponding to each object to be detected in the image information.

7. The method according to claim 5, characterized in that Correcting the rectangular target recognition frame in the corrected image mark information includes: Acquire target position information in a plurality of historical image marking information located before the corrected image marking information in the target storage sequence; According to the multiple target position information, an average position offset between two adjacent image marking information is generated; Generating offset direction information between the corrected image mark information and adjacent historical image mark information according to the plurality of target position information; A rectangular target recognition frame corresponding to each object to be detected in the corrected image marking information is generated according to the average position offset and the offset direction information.

8. A non-transitory computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements an exception identification method based on the Kylin operating system as described in any one of claims 1 to 7.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements an exception identification method based on the Kylin operating system as described in any one of claims 1 to 7.

Images