WO2023184833A1

WO2023184833A1 - Detection result processing method and apparatus, device, medium, and computer program product

Info

Publication number: WO2023184833A1
Application number: PCT/CN2022/114478
Authority: WO
Inventors: 李建; 张辉
Original assignee: 上海商汤智能科技有限公司
Priority date: 2022-03-28
Filing date: 2022-08-24
Publication date: 2023-10-05
Also published as: CN114627098A

Abstract

Embodiments of the present invention provide a detection result processing method and apparatus, a device, a medium, and a computer program product. The detection result processing method comprises: obtaining a detection result of an image to be detected; determining, on the basis of a historical detection result, whether the detection result satisfies a preset condition; and in response to the detection result satisfying the preset condition, executing preset processing corresponding to the detection result.

Description

Test result processing methods and devices, equipment, media and computer program products

Cross-references to related applications

This disclosed embodiment is based on a Chinese patent application with application number 202210314170.1, application date is March 28, 2022, and the application name is "Detection result processing method and device, equipment, medium", and claims the priority of this Chinese patent application , the entire content of this Chinese patent application is hereby incorporated by reference into this disclosure.

Technical field

The present disclosure relates to, but is not limited to, the technical field of image processing, and in particular, to a detection result processing method and device, equipment, media and computer program products.

Background technique

In related technologies, due to the wide variety of images to be detected and detection environments, the detection results are inaccurate and the credibility of the product is reduced.

Contents of the invention

Embodiments of the present disclosure provide at least a detection result processing method and device, equipment, media and computer program products.

Embodiments of the present disclosure provide a detection result processing method, which includes: obtaining the detection result of the image to be detected; judging whether the detection result satisfies the preset condition based on the historical detection result; in response to the detection result satisfying the preset condition, executing the detection result corresponding to Preset processing.

Embodiments of the present disclosure provide a detection result processing device, including: a detection part configured to obtain detection results of images to be detected; a judgment part configured to judge whether the detection results meet preset conditions based on historical detection results; and an execution part , configured to execute preset processing corresponding to the detection result in response to the detection result meeting the preset condition.

An embodiment of the present disclosure provides an electronic device, including a memory and a processor. The processor is configured to execute program instructions stored in the memory to implement the above detection result processing method.

Embodiments of the present disclosure provide a computer-readable storage medium on which program instructions are stored. When the program instructions are executed by a processor, the above detection result processing method is implemented.

Embodiments of the present disclosure provide a computer program product. The computer program product includes a computer program or instructions. When the computer program or instructions are run on an electronic device, the above detection results are achieved when the electronic device is executed. Approach.

In the embodiment of the present disclosure, after obtaining the detection result of the image to be detected, and judging that the detection result satisfies the preset conditions based on the historical detection results, the preset processing corresponding to the detection result is executed. Compared with after performing the detection, , directly executing the preset processing corresponding to the detection result can reduce the probability of mistakenly executing the preset processing.

It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory, and are not restrictive of the disclosure.

Description of drawings

The accompanying drawings herein are incorporated into and constitute a part of this specification. They illustrate embodiments consistent with the disclosure and, together with the description, serve to explain the technical solutions of the disclosure.

Figure 1 is a schematic flowchart of a detection result processing method provided by an embodiment of the present disclosure;

Figure 2 is a partial sub-flow schematic diagram showing step S12 of a detection result processing method provided by an embodiment of the present disclosure;

Figure 3 is a partial sub-flow schematic diagram showing step S12 of a detection result processing method provided by an embodiment of the present disclosure;

Figure 4 is a schematic flowchart of a detection result processing method provided by an embodiment of the present disclosure;

Figure 5 is a schematic structural diagram of a detection result processing device provided by an embodiment of the present disclosure;

Figure 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure;

FIG. 7 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present disclosure.

Detailed ways

The solutions of the embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

In the following description, for purposes of explanation and not limitation, specific details such as specific system structures, interfaces, technologies, and the like are set forth in order to provide a thorough understanding of the present disclosure.

The term "and/or" in this article is an association relationship that describes related objects, indicating that there can be three relationships. For example, A and/or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. these three situations. In addition, the character "/" in this article generally indicates that the related objects are an "or" relationship. In addition, "many" in this article means two or more than two. In addition, the term "at least one" herein means any one of a plurality or any combination of at least two of a plurality, for example, including at least one of A, B, and C, which can mean including from A, Any one or more elements selected from the set composed of B and C.

In related technologies, due to the wide variety of images to be detected and detection environments, inaccurate detection results are a common problem encountered in the early stages of the project, and it is also a relatively serious problem. The generally adopted method includes re-training the detection model used for detection in the detection application system after it has been used for a period of time. In this way, before the detection model is retrained, because the detection results are inaccurate, the problem of incorrect execution of corresponding processing occurs frequently, thus reducing the credibility of the product.

Figure 1 is a schematic flowchart of a detection result processing method provided by an embodiment of the present disclosure. As shown in Figure 1, the detection result processing method may include the following steps S11 to S13, wherein:

Step S11: Obtain the detection result of the image to be detected.

Here, the images to be detected may include but are not limited to security monitoring images, medical images, hand-drawn images, animated images, etc. For example, the image to be detected is a security surveillance image. The security monitoring images may be captured by monitoring equipment, such as monitoring cameras installed on various streets.

In some application scenarios, the image to be detected can be taken in a bank scene, a street scene, a construction site scene, a school scene, etc.

The method of obtaining the detection result of the image to be detected may be performed using a detection model, or may be determined by performing data analysis on the image information of the image to be detected using preset rules. During implementation, whether using a detection model or using preset rules to detect images to be detected, erroneous detection results may occur.

For example, if the lens used for shooting is blocked by objects such as fog, water droplets, leaves, etc., the pictures taken will be different from those taken under normal circumstances, resulting in errors in the detection results of the images to be detected.

Step S12: Determine whether the detection results meet the preset conditions based on historical detection results.

Here, if the detection result meets the preset conditions, the execution device of the detection result processing method provided by the embodiment of the present disclosure determines that the detection result is accurate. Similarly, if the detection result does not meet the preset conditions, the execution device determines that the detection result is inaccurate. .

In some implementations, the determination method may be by comparing historical detection results stored in the blacklist database. By comparing the test results of this test with historical test results, it is judged whether the test results obtained by this test meet the preset conditions. The historical detection results may be historical detection results that do not meet preset conditions. The historical detection results may be obtained by this execution device or by other execution devices. For example, the historical detection results in the blacklist database are all obtained by this execution device.

In some application scenarios, the blacklist database can also store the shooting component information of the historical detection images corresponding to each historical detection result.

For example, the shooting component information may include but is not limited to the installation address, model, shooting parameters (exposure time, simulation gain, etc.) of the shooting component, etc. So that subsequent users can adjust the shooting component according to the shooting component information, thereby improving the quality of the images captured by the shooting component, or comparing the detection results of the images to be detected captured by the shooting component with the historical detection corresponding to the shooting component in the blacklist database. The results are compared and the corresponding judgment results are obtained.

Step S13: In response to the detection result satisfying the preset condition, execute the preset processing corresponding to the detection result.

Here, the preset processing corresponding to the detection results can be customized by the user, and is not limited in the embodiments of the present disclosure.

For example, the preset processing may be alarm processing, target recognition processing, uploading detection results and corresponding images to be detected to a preset recipient or a back-end business processing system, etc. For example, if an image to be detected taken on a city street is detected and it is found that there is a national protected animal that escaped from a zoo in the image, alarm processing will be performed so that the corresponding processor can rush to the street location corresponding to the image to be detected in time to deal with the problem. The animal is protected, and the detection results and the corresponding images to be detected are uploaded to the preset recipient, so that the preset recipient can perform corresponding operations based on the received detection results and the images to be detected.

In the embodiment of the present disclosure, after obtaining the detection result of the image to be detected, and judging that the detection result satisfies the preset condition based on the historical detection results, the preset processing corresponding to the detection result is executed. In this way, compared with directly executing the preset processing corresponding to the detection result after performing the detection, the probability of mistakenly executing the preset processing can be reduced.

In some embodiments, after performing step S12, step S14 may also be performed based on the determination result of step S12.

Step S14: In response to the detection result not meeting the preset condition, do not execute the preset processing corresponding to the detection result.

For example, if the detection result is that a fight event is detected, and the preset processing corresponding to the fight event is alarm processing, but because the detection result does not meet the preset conditions, the alarm processing will no longer be performed.

In some implementations, the above step S11 may be performed by a detection model. Historical detection results include historical false detection results with incorrect detection results. Among them, each historical misdetection result is stored in a preset blacklist database.

In some embodiments, the detection result processing method may also include the following steps S15 to S16, wherein:

Step S15: Receive the instruction to update the detection model;

Step S16: In response to the instruction, the detection model is updated. The updated detection model is trained by combining the historical misdetection results with the historical images to be detected corresponding to the historical misdetection results.

In some embodiments, the instruction is also used to train the detection model in combination with the historical to-be-detected images corresponding to the historical misdetection results, so that the trained detection model can subsequently be used to perform the step of detecting the to-be-detected images. The process of training the detection model using historical misdetection results and corresponding historical images to be detected can be executed by the execution device of the detection result processing method provided by the embodiment of the present disclosure, or by other devices. For example, the training process is performed by other devices.

In some application scenarios, the historical misdetection results in the blacklist library and the historical undetected images corresponding to the historical misdetection results are fed back to the training device at regular intervals, so that the training device can use the received historical misdetection results and historical misdetection results to The image to be detected performs targeted secondary training on the detection model, and then the trained detection model is fed back to the execution device, so that the execution device detects the image to be detected based on the detection model after the secondary training.

In the embodiment of the present disclosure, since retraining the detection model is time-consuming, and the training of the model depends on certain sample data, the detection result processing method provided by the embodiment of the present disclosure can be processed in a short time. The problem of event misdetection, and through accumulation over a period of time, provide more targeted training samples for secondary training on the algorithm side. Furthermore, during the secondary training process of the detection model, the detection result processing method provided by the embodiment of the present disclosure can be continuously executed.

In some embodiments, the model can also be forward trained based on non-historical false detection results and corresponding historical images to be detected. That is, the model is trained based on the historical misdetection results and the historical images to be detected corresponding to the historical misdetection results as negative samples, and the model is trained based on the non-historical misdetection results and the corresponding historical images to be detected as positive samples. Regarding the subsequent specific training method of the detection model, the embodiment of the present disclosure does not limit it.

In the embodiment of the present disclosure, the detection model is updated to a detection model trained by historical misdetection results and corresponding historical images to be detected, so that the accuracy of the next detection result is higher.

In some embodiments, the detection result includes the detection result type and the accuracy corresponding to the detection result type.

In some embodiments, the detection result is an object detection result, and the detection result type is an object type. Or the detection result is an event detection result, and the detection result type is an event type. The accuracy corresponding to the detection result type refers to the accuracy of the presence of objects of this object type or events of this event type in the image to be detected.

Among them, the objects here can include but are not limited to human bodies, animal bodies, vehicles and other non-fixed objects that are stored in a fixed position and cannot move by themselves and cannot be moved by external forces.

For example, object detection includes species detection, the detection result can be a species detection result, and the detection result type can be a category to which the detected species belongs. The accuracy of a detection result type refers to the probability that an object of that target category exists in the image to be detected.

Event detection can include but is not limited to corresponding event detection in different application scenarios. For example, any scenario can include detection of fight events, detection of red light running events in traffic scenarios, detection of vehicle speeding events, and preset impoliteness in bank scenarios. Customer behavior detection, smoking incident detection, etc. In event detection, the detection result may be a conclusion of whether the target event is contained in the image to be detected, and the type of the detection result may be the type of the target event. Among them, there can be many kinds of target events contained in the image to be detected, and the types of detection results also include many kinds.

For example, if the image to be detected contains speeding events and red light running events, the type of the detection result includes the speeding event type and the red light running event type. The accuracy of the detection result type refers to the probability that the target event type exists in the image to be detected.

In some application scenarios, before performing step S12, the following step S17 may also be performed, wherein:

Step S17: Determine whether the detection result type in the detection result is a preset detection result type.

Here, in response to the detection result type being the preset detection result type, step S12 is executed. The number of preset detection result types is at least one.

In some implementations, when the detection is object detection, the preset detection result type is the preset object type; when the detection is event detection, the preset detection result type is the preset event type.

For example, if the detection is event detection, the preset detection result type includes a red light running event type and a car accident event type. If the detection result type is a red light running event type or a car accident event type, step S12 is executed.

FIG. 2 is a partial sub-flow schematic diagram showing step S12 of a detection result processing method provided by an embodiment of the present disclosure. As shown in Figure 2, the above-mentioned step S12 includes the following steps S21 to step S22, wherein:

Step S21: Determine whether the accuracy corresponding to the detection result type is greater than or equal to the preset accuracy.

Here, the preset accuracy is based on the historical accuracy in historical detection results. The historical detection results are detection results based on historical images to be detected. Wherein, the preset accuracy may be dynamically determined. In some implementations, the preset accuracy may be obtained from the historical accuracy corresponding to historical misdetection results.

In some embodiments, before performing step S21, the method further includes step S20, wherein:

Step S20: Determine a preset accuracy based on the historical accuracy corresponding to historical detection results including the detection result type.

Here, the historical accuracy corresponding to different detection result types is the same or different.

In some embodiments, the history including the detection result type is the historical accuracy corresponding to the detection result, which is determined based on the historical accuracy corresponding to the historical false detection results with respect to the detection result type.

For example, the detection model supports the detection of m preset detection result types (for example, fighting event type, smoking event type, etc.). If the detection result type in this detection result is a fighting event type, the preset detection results in the blacklist database will be Result Type Historical Accuracy for Fight Event Type as Default Accuracy. By determining the corresponding historical accuracy for different detection result types, the corresponding preset accuracy can be determined for the detection results of different detection result types, thereby improving the accuracy of the judgment results.

In some application scenarios, in the early stages of project operation, if there are no historical detection results in the blacklist database, the preset accuracy is considered to be 0.

In some implementations, if there is no historical false detection result corresponding to the detection result type in the blacklist, the historical accuracy corresponding to the detection result type is considered to be 0.

In some embodiments, historical detection results containing a detection result type refer to historical detection results whose detection result type is a preset detection result type. Historical false detection results containing detection result types refer to historical detection results in which the detection result type is the preset detection result type and the detection result is incorrect.

For example, a total of 100 tests were performed and 100 test results were obtained. Among them, there were 10 test results with the test result type being the preset test result type (such as fighting event type, smoking event type, etc.), which means 90 of them The preset detection result type was not detected in the image to be detected. And 3 of the 10 test results are incorrect, and 7 test results are correct. Then there are 10 historical test results including the test result type, and there are 3 historical false test results in total.

In some embodiments, the above step S20 includes steps S201 to S202, wherein:

Step S201: Obtain the historical accuracy of each historical misdetection result including the detection result type.

Step S202: Average the historical accuracy corresponding to each historical misdetection result to obtain a preset accuracy.

In some embodiments, in addition to the average of the historical accuracy corresponding to each historical misdetection result of the detection result type, the preset accuracy can also be a result obtained by other statistical methods, such as each historical accuracy Median, maximum, minimum, etc. The disclosed embodiments are not limiting.

In the embodiment of the present disclosure, by counting the historical accuracy of each historical misdetection result regarding the detection result type, the historical accuracy corresponding to the detection result type is obtained, which can improve the accuracy of the judgment result.

For example, the detection model supports the detection of a total of m preset detection result types (for example, fighting event types, smoking event types, etc.), and the determination of the historical accuracy of the fighting event type is based on the correspondence of historical misdetection results of the fighting event type. The historical accuracy of It is (0.5+0.5+0.6+0.6)/4=0.55. Similarly, the historical accuracy corresponding to the smoking event type is also determined by the historical accuracy corresponding to the historical misdetection results of the smoking event type.

In some application scenarios, the blacklist database also includes information about the shooting components corresponding to each historical misdetection result. There may be multiple preset accuracies regarding detection result types, and different historical accuracies are determined by historical misdetection results corresponding to the corresponding shooting components.

For example, if the historical undetected images corresponding to the historical misdetection results under the detection result type are taken by the first number of shooting components, then the detection result type corresponds to the first number of preset accuracies. If the current detection result belongs to the detection result type, and the detection result is obtained from the image to be detected captured by one of the first number of photographing components, then the preset accuracy corresponding to the photographing component is used as the final preset accuracy.

In some embodiments, if the current detection result belongs to this detection result type, and the current detection result is not obtained from the image to be detected captured by any one of the first number of shooting components, then the first number of detection result types Statistics are performed on each preset accuracy to obtain the final preset accuracy. Specific statistical methods may include but are not limited to averaging, median, maximum, minimum, etc., which are not limited in the embodiment of the present disclosure.

Step S22: In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, determine that the detection result satisfies the preset condition.

For example, if the detection result type included in the detection result is a fight type, the accuracy corresponding to the detection result type is 0.6, and the preset accuracy is 0.55, then it is determined that the detection result meets the preset conditions.

In some embodiments, in response to the accuracy corresponding to the detection result type being less than the preset accuracy, it is determined that the detection result does not meet the preset condition. In some application scenarios, in response to the accuracy corresponding to the detection result type being less than the preset accuracy, the detection result is regarded as the historical misdetection result, and the accuracy in the detection result is regarded as the historical accuracy. During implementation, based on the detection result type in the detection result, a preset detection result type matching the detection result type in the blacklist library is determined, and the detection result is added to the preset detection result type set to re-determine the preset detection result type. Assume the historical accuracy corresponding to the detection result type. Alternatively, in response to the accuracy being less than the preset accuracy, the detection result of the image to be detected is discarded.

In the embodiment of the present disclosure, by determining the preset accuracy based on the historical accuracy of historical detection results, and judging whether the detection results meet the preset conditions based on the preset accuracy, the detection results are judged based on historical a priori information, Improved the accuracy of judgment results.

FIG. 3 is a partial sub-flow schematic diagram showing step S12 of a detection result processing method provided by an embodiment of the present disclosure. As shown in Figure 3, the above step S12 may include the following steps S31 to S33, wherein:

Step S31: Determine whether the accuracy corresponding to the detection result type is greater than or equal to the preset accuracy.

Here, the preset accuracy is based on the historical accuracy in historical detection results. For the specific step of determining whether the accuracy is greater than or equal to the preset accuracy, please refer to the determination process in the above-mentioned step S21.

Step S32: In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, obtain the user's judgment result about the detection result.

In some implementations, the image to be detected and the detection result are displayed on the display interface so that the user can check the detection result. That is, the user determines whether the detection result corresponding to the image to be detected is accurate. Among them, the user can make a positive result to indicate that the detection result is correct, and can also make a negative result to indicate that the detection result is incorrect. In some application scenarios, when the accuracy corresponding to the detection result type is greater than or equal to the accuracy, the detection results are reported so that users can make manual judgments.

Step S33: In response to the determination result being a positive result indicating that the detection result is correct, determine that the detection result satisfies the preset condition.

In some embodiments, in response to the determination result being a negative result indicating that the detection result is incorrect, it is determined that the detection result does not meet the preset condition. And, in response to the judgment result being a negative result indicating that the detection result is incorrect, the detection result is used as a historical false detection result. The detection degree in this detection result is regarded as historical accuracy. During implementation, based on the detection result type in the detection result, a preset detection result type matching the detection result type in the blacklist library is determined, and the detection result is added to the preset detection result type set to re-determine the preset detection result type. Set the preset accuracy corresponding to the detection result type so that it can be used when judging the detection results next time.

In the embodiment of the present disclosure, by obtaining the user's judgment result on whether the detection result is wrong, and after the user gives a positive result, it is determined that the detection result satisfies the preset conditions, so that the accuracy of the judgment result can be improved.

In some embodiments, in response to the detection result not meeting the preset condition, the detection result is used as a historical false detection result. Among them, the historical false detection results are used in the next detection process to determine whether the detection results obtained by the next detection meet the preset conditions. During implementation, the detection results that do not meet the preset conditions can be put into the blacklist library. And the detection results that do not meet the preset conditions and the corresponding images to be detected can be stored for subsequent use.

In the embodiments of the present disclosure, by not executing the preset processing corresponding to the detection result when the detection result does not meet the preset condition, the probability of incorrect execution of the preset processing can be reduced. In addition, when the detection result does not meet the preset conditions, the detection result is used as a historical misdetection result, which can improve the accuracy of the next detection result judgment.

In some embodiments, each historical false detection result and the feature sequence extracted from the corresponding historical image to be detected are recorded in the blacklist database. The above step S12 may include the following steps S121 to S123, wherein:

Step S121: Obtain the target feature sequence corresponding to the detection result.

Here, the target feature sequence is a feature sequence extracted from the image to be detected corresponding to the historical false detection results. The target feature sequence is used to obtain the detection result of the corresponding image to be detected. In some embodiments, feature extraction is performed on the image to be detected to obtain a feature sequence of the image to be detected, and then based on the feature sequence, a detection result of the image to be detected is obtained.

Step S122: Match the feature sequence of the image to be detected corresponding to the detection result with the target feature sequence to obtain a matching result.

Step S123: Determine whether the detection result satisfies the preset condition based on the matching result.

In some implementations, step S123 includes step S1231 and/or step S1232, wherein:

Step S1231: In response to the matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being greater than the preset matching degree, it is determined that the detection result does not meet the preset condition.

Step S1232: In response to the matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being less than or equal to the preset matching degree, obtain the user's judgment result about the detection result, and in response to the judgment result indicating that the detection result is An unmistakable positive result confirms that the test results meet the preset conditions.

In some application scenarios, in response to the judgment result being a negative result indicating that the detection result is incorrect, the detection result is regarded as a historical false detection result.

For example, the detection result type in this detection result is matched with the preset detection result type corresponding to each historical false detection result. If the matching is successful, the feature sequence of the image to be detected corresponding to the detection result is matched with the preset detection result type. The feature sequences of historical images corresponding to each historical false detection result under the detection result type are matched. If the matching degree is greater than or equal to the preset matching degree, the detection result is deemed not to meet the preset conditions. If the matching degree is less than the preset matching degree, the step of obtaining the user's judgment result regarding the detection result in step S32 is executed.

In some embodiments, before performing the above step S11, the following steps S18 to S19 may also be performed, wherein:

Step S18: Determine the false detection rate corresponding to each historical detection result within the preset historical time period.

Here, the false detection rate is obtained based on the number of historical false detection results and the number of total historical detection results within a preset time period.

In some embodiments, the false detection rate may be the ratio of the number of historical false detection results to the total number of historical detection results within a preset time period. The length of the preset historical time period can be set according to user needs, and is not limited in this disclosed embodiment.

For example, the preset historical time period may be several hours or several days before the current detection process is performed. Among them, some are one and above. For example, if the preset time period is m days, there are a total of historical detection results during these m days, and there are b historical false detection results, then the false detection rate is b/a.

In some embodiments, the total historical detection results are not the number of detections performed within the preset historical time period, but a collection of historical detection results whose detection result type is the preset detection result type. Historical false detection results refer to historical detection results with incorrect detection results. Please refer to the above for specific acquisition methods.

For example, the preset historical time period is 1 day before the current image to be detected is obtained, and the number of detections performed during this time period is 100 times to obtain 100 detection results. Among them, the detection result type is the preset detection result type (such as a fight event type, smoking event type, etc.), and 3 of the 10 do not meet the preset conditions, and 7 meet the preset conditions, then there are a total of 10 historical detection results, and historical false detections There were 3 copies in total.

Step S19: In response to the false detection rate being greater than the preset false detection rate, execute the above step S11.

In some embodiments, when it is determined that the false detection rate is less than or equal to the preset false detection rate, the detection result of the image to be detected is obtained, and preset processing corresponding to the detection result is performed.

Here, when the false detection rate is less than or equal to the preset false detection rate, the above step S12 may not be executed, and the preset processing corresponding to the detection result may be directly executed. Among them, the preset false detection rate can be customized by the user. For example, if the preset false detection rate is 5% and the false detection rate within the preset historical time period is 2%, it is determined that the detection result meets the preset conditions.

In some embodiments, obtaining the false detection rate corresponding to each historical detection result within the preset historical time period and comparing the false detection rate with the preset false detection rate is performed prior to the above step S12. If the false detection rate is greater than the preset false detection rate, the above step S12 is executed. Or, when the false detection rate is less than or equal to the preset false detection rate, step S12 is not performed and the detection result is directly confirmed to meet the preset conditions. That is, when the false detection rate is less than or equal to the preset false detection rate, the preset processing corresponding to the detection result is executed. When the false detection rate of the detection results is low, the probability of incorrect detection results is also lower, and the process of making judgments by using the false detection rate is relatively convenient.

In order to better understand the technical solution provided by the embodiment of the present disclosure, please also refer to FIG. 4 , which is a schematic flow chart of a detection result processing method provided by the embodiment of the present disclosure. As shown in Figure 4, the detection result processing method provided by the embodiment of the present disclosure includes the following steps S41 to S49, wherein:

Step S41: Obtain the image to be detected.

Here, the method of obtaining the image to be detected may be that after the photographing component captures the image to be detected, it receives the image to be detected sent by the photographing component through a pre-established communication link between the photographing component and the image to be detected. In some other embodiments, capturing the image to be detected is also performed by an execution device that executes the detection result processing method provided by the embodiments of the present disclosure.

Step S42: Detect the image to be detected and obtain the detection result of the image to be detected.

Here, the specific detection method is as shown above. In some embodiments, after performing step S42, continue to perform step S41, and perform step S42 on the newly acquired image to be detected.

Step S43: Determine whether the accuracy corresponding to the detection result type is greater than the preset accuracy.

Here, the specific judgment method is as shown above. If the judgment result is that the accuracy is less than the preset accuracy, step S44 is executed; otherwise, step S45 is executed.

Step S44: In response to the accuracy corresponding to the detection result type being less than the preset accuracy, determine that the detection result does not meet the preset condition.

Here, step S42 is re-executed, thereby cyclically executing the target detection method provided by the embodiment of the present disclosure. Furthermore, the detection results are used as historical misdetection results, and the accuracy corresponding to the detection results is used as historical accuracy, which is used to update the corresponding preset accuracy. In some embodiments, after step S44 is performed, step S41 may be re-executed and the cycle continues.

Step S45: In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, obtain the user's judgment result about the detection result.

Here, the specific acquisition method is as shown above.

Step S46: Determine whether the result is a positive result indicating that the detection result is correct.

Here, if the judgment result is a positive result indicating that the detection result is correct, step S48 is executed; otherwise, step S47 is executed.

Step S47: In response to the judgment result being a negative result indicating that the detection result is incorrect, determine that the detection result does not meet the preset condition.

Here, step S42 is executed again, and the detection result processing method provided by the embodiment of the present disclosure is cyclically executed.

Step S48: In response to the judgment result being a positive result indicating that the detection result is correct, determine that the detection result satisfies the preset condition.

Step S49: Execute preset processing corresponding to the detection result.

Here, the specific way of executing the preset processing corresponding to the detection result is as shown above.

In the embodiment of the present disclosure, after the image to be detected is detected, and when the detection result satisfies the preset conditions, the preset processing corresponding to the detection result is executed. Compared with directly executing the processing corresponding to the detection result after the detection, For preset processing, the probability of mistakenly executing the preset processing can be reduced.

In some application scenarios, the detection result processing method provided by the embodiments of the present disclosure can be applied to detection application systems. This detection application system can be applied to scenarios such as smart cities, public safety, and security. The detection application system includes a bottom layer and a business layer. The bottom layer performs detection of the image to be detected and determines whether the accuracy in the detection results is greater than or equal to the preset accuracy. When the accuracy is judged to be greater than or equal to the preset accuracy, the detection results are reported to the business layer. Otherwise, the detection results are directly stored in the blacklist database or discarded. The user will manually judge whether the detection result is accurate at the business layer. If it is accurate, the user will enter the business processing link to perform the preset processing corresponding to the detection result. Otherwise, the detection result will be stored in the blacklist database or discarded. In some embodiments, as the detection model continues to be trained based on historical misdetection results and corresponding images to be detected, the accuracy becomes higher and higher. When the false detection rate meets product expectations (for example, less than or equal to the preset False detection rate), you can choose to skip the comparison process of preset accuracy and the manual judgment process, and directly enter the business processing link to execute the preset processing corresponding to the detection results.

In the embodiment of the present disclosure, the execution subject of the detection result processing method may be a detection result processing device. For example, the detection result processing method may be executed by a terminal device or a server or other processing device, where the terminal device may be a user equipment (User Equipment). , UE), mobile devices, user terminals, terminals, cellular phones, cordless phones, personal digital assistants (Personal Digital Assistant, PDA), handheld devices, computing devices, vehicle-mounted devices, wearable devices, etc. In some implementations, the detection result processing method can be implemented by the processor calling computer readable instructions stored in the memory.

Those skilled in the art can understand that in the above-mentioned methods of specific embodiments, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be based on its function and possible The internal logic is determined.

Based on the same inventive concept, embodiments of the present disclosure also provide a detection result processing device corresponding to the detection result processing method. Figure 5 is a schematic structural diagram of a detection result processing device provided by an embodiment of the present disclosure. The detection result processing device 50 includes a detection part 51 , a judgment part 52 and an execution part 53 . The detection part 51 is configured to obtain the detection result corresponding to the image to be detected; the judgment part 52 is configured to judge whether the detection result satisfies the preset condition based on the historical detection results; the execution part 53 is configured to respond to the detection result satisfying the preset condition Conditions, execute the preset processing corresponding to the detection results.

In some embodiments, the detection results include the detection result type and the accuracy corresponding to the detection result type; the judgment part 52 is also configured to: determine whether the accuracy corresponding to the detection result type is greater than or equal to the preset accuracy, and the preset accuracy is The accuracy is obtained based on the historical accuracy in historical detection results; in response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, it is determined that the detection result satisfies the preset conditions.

In some embodiments, the detection results include the detection result type and the accuracy corresponding to the detection result type; the judgment part 52 is also configured to: determine whether the accuracy corresponding to the detection result type is greater than or equal to the preset accuracy, and the preset accuracy is The degree is obtained based on the historical accuracy in historical detection results; in response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, the user's judgment about the detection result is obtained; in response to the judgment result being an affirmation indicating that the detection result is correct As a result, it is determined that the detection results meet the preset conditions.

In some embodiments, the judgment part 52 is further configured to determine the preset accuracy based on the historical accuracy corresponding to historical detection results including detection result types, where the historical accuracy corresponding to different detection result types is the same or different.

In some embodiments, the historical detection results include historical misdetection results with incorrect detection results. The judgment part 52 is also configured to: obtain the historical accuracy of each historical misdetection result including the detection result type; The historical accuracy of the inspection results is averaged to obtain the preset accuracy.

In some embodiments, the judgment part 52 is further configured to: in response to the judgment result being a negative result indicating that the detection result is incorrect, use the detection result as a historical misdetection result.

In some embodiments, the judgment part 52 is also configured to: determine the false detection rate corresponding to each historical detection result within the preset historical time period, and the false detection rate is based on the number and total number of historical false detection results within the preset time period. The number of historical detection results is obtained; the detection part 51 is also configured to: in response to the false detection rate being greater than the preset false detection rate, execute acquisition of the detection results of the image to be detected.

In some embodiments, the detection part 51 is also configured to: obtain the detection result of the image to be detected when it is determined that the false detection rate is less than or equal to the preset false detection rate; the execution part 53 is also configured to: execute Preset processing corresponding to the detection results.

In some embodiments, the historical detection results include historical misdetection results with incorrect detection results. The judgment part 52 is also configured to: obtain a target feature sequence corresponding to the detection result, and the target feature sequence is corresponding to the historical misdetection results. The feature sequence extracted from the image to be detected, the target feature sequence is used to obtain the detection result corresponding to the image to be detected; the feature sequence of the image to be detected corresponding to the detection result is matched with the target feature sequence to obtain the matching result; determined based on the matching result Check whether the test results meet the preset conditions.

In some embodiments, the judgment part 52 is also configured to: in response to the matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being greater than the preset matching degree, determine that the detection result does not meet the preset condition. ; and/or, in response to the matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being less than or equal to the preset matching degree, obtain the user's judgment result about the detection result, and respond to the judgment result as indicating A positive result that the test result is correct, confirming that the test result meets the preset conditions.

In some embodiments, the detection result processing device 50 includes an update part (not shown). The step of detecting the image to be detected is performed by the detection model. The update part is configured to: receive an instruction to update the detection model, and update the detection model in response to the instruction. The updated detection model is combined with historical misdetection results. The historical images to be detected corresponding to the historical false detection results are obtained by training the detection model.

In some embodiments, the detection result is an object detection result, and the detection result type is an object type; or the detection result is an event detection result, and the detection result type is an event type.

For descriptions of the processing flow of each part in the device and the interaction flow between the parts, please refer to the relevant descriptions in the above method embodiments.

In the embodiments of the present disclosure and other embodiments, "part" may be part of a circuit, part of a processor, part of a program or software, etc., of course, it may also be a unit, it may be a module or it may be non-modular.

Based on the same inventive concept, an electronic device is also provided in an embodiment of the present disclosure. FIG. 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure. As shown in FIG. 6 , the electronic device 60 includes a memory 61 and a processor 62 . The processor 62 is used to execute program instructions stored in the memory 61 to implement the steps in any of the above detection result processing method embodiments. In some implementation scenarios, the electronic device 60 may include but is not limited to: security equipment, medical equipment, microcomputers, desktop computers, and servers. In addition, the electronic device 60 may also include mobile devices such as laptop computers and tablet computers, which will not be discussed here. limited.

Among them, electronic equipment includes driving devices and sensors. The sensor is used to obtain the motion parameters of the shooting device, and the driving device is connected to the processor and lens of the shooting device, and is used to receive instructions from the processor and drive the lens to move to adjust the posture of the lens.

In some embodiments, the processor 62 is used to control itself and the memory 61 to implement the steps in any of the above detection result processing method embodiments. The processor 62 may also be called a CPU (Central Processing Unit). The processor 62 may be an integrated circuit chip with signal processing capabilities. The processor 62 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. In addition, the processor 62 may be implemented by an integrated circuit chip.

Based on the same inventive concept, embodiments of the present disclosure also provide a computer-readable storage medium. FIG. 7 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present disclosure. The computer-readable storage medium 70 stores program instructions 71. When the program instructions 71 are executed by the processor, the steps in any of the above detection result processing method embodiments are implemented.

An embodiment of the present disclosure also provides a computer program product. The computer program product includes a computer program or instructions. When the computer program or instructions are run on an electronic device, the electronic device causes the electronic device to execute any of the above. Steps in the embodiment of the detection result processing method. Among them, the above-mentioned computer program product can be specifically implemented by hardware, software or a combination thereof. In some embodiments, the computer program product is embodied as a computer storage medium. In some embodiments, the computer program product is embodied as a software product, such as a software development kit (SDK) and so on.

In some embodiments, the functions or included parts of the device provided by the embodiments of the present disclosure can be used to perform the methods described in the above method embodiments. For specific implementation, reference can be made to the description of the above method embodiments.

The above description of various embodiments is intended to emphasize the differences between the various embodiments, and the similarities or similarities may be referred to each other.

In the several embodiments provided by the embodiments of this disclosure, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is schematic. For example, the division of parts or units is a logical function division. In actual implementation, there may be other division methods. For example, units or components may be combined or integrated into another unit. A system, or some features can be ignored, or not implemented. On the other hand, the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical or other forms.

In addition, each functional unit in various embodiments of the present disclosure may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above integrated units can be implemented in the form of hardware or software functional units. Integrated units may be stored in a computer-readable storage medium if they are implemented in the form of software functional units and sold or used as independent products. Based on this understanding, the technical solution of the present disclosure is essentially or contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including a number of instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the various implementation methods of the present disclosure. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code. .

Industrial applicability

Embodiments of the present disclosure provide a detection result processing method and device, equipment, media and computer program products. The detection result processing method includes: obtaining the detection result of the image to be detected; judging whether the detection result satisfies the preset based on the historical detection results. Condition; in response to the detection result meeting the preset condition, execute the preset processing corresponding to the detection result. The above solution can filter the detection results and reduce the probability of misoperation in preprocessing.

Claims

A method for processing test results, including:

Obtain the detection results of the image to be detected;

Determine whether the detection results meet the preset conditions based on historical detection results;

In response to the detection result satisfying the preset condition, preset processing corresponding to the detection result is executed.
The method according to claim 1, wherein the detection result includes a detection result type and an accuracy corresponding to the detection result type; and determining whether the detection result satisfies a preset condition based on historical detection results includes:

Determine whether the accuracy corresponding to the detection result type is greater than or equal to a preset accuracy, which is obtained based on the historical accuracy in the historical detection results;

In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, it is determined that the detection result satisfies the preset condition.
The method according to claim 1, wherein the detection result includes a detection result type and an accuracy corresponding to the detection result type; and determining whether the detection result satisfies a preset condition based on historical detection results includes:

Determine whether the accuracy corresponding to the detection result type is greater than or equal to a preset accuracy, which is obtained based on the historical accuracy in the historical detection results;

In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, obtaining the user's judgment result regarding the detection result;

In response to the judgment result being a positive result indicating that the detection result is correct, it is determined that the detection result satisfies the preset condition.
The method according to claim 2 or 3, wherein before determining whether the accuracy corresponding to the detection result type is greater than or equal to a preset accuracy, the method further includes:

The preset accuracy is determined based on the historical accuracy corresponding to historical detection results including the detection result type, where the historical accuracy corresponding to different detection result types is the same or different.
The method according to claim 4, wherein the historical detection results include historical misdetection results with incorrect detection results; the preset is determined based on the historical accuracy corresponding to the historical detection results including the detection result type. Accuracy, including:

Obtain the historical accuracy of each historical false detection result including the detection result type;

The historical accuracy of each of the historical misdetection results is averaged to obtain the preset accuracy.
The method according to claim 4 or 5, wherein the method further includes:

In response to the judgment result being a negative result indicating that the detection result is incorrect, the detection result is used as a historical false detection result.
The method according to any one of claims 2 to 6, wherein before obtaining the detection result of the image to be detected, the method further includes:

Determine the false detection rate corresponding to each historical detection result within the preset historical time period, the false detection rate being obtained based on the number of historical false detection results within the preset time period and the number of total historical detection results;

In response to the false detection rate being greater than the preset false detection rate, the acquisition of the detection result of the image to be detected is performed.
The method of claim 7, further comprising:

When it is determined that the false detection rate is less than or equal to the preset false detection rate, obtain the detection result of the image to be detected;

Execute preset processing corresponding to the detection results.
The method according to any one of claims 1 to 8, wherein the historical detection results include historical misdetection results with incorrect detection results; and judging whether the detection results meet preset conditions based on the historical detection results, include:

Obtain a target feature sequence corresponding to the detection result. The target feature sequence is a feature sequence extracted from the image to be detected corresponding to the historical misdetection result. The target feature sequence is used to obtain the detection result corresponding to the image to be detected. ;

Match the feature sequence of the image to be detected corresponding to the detection result with the target feature sequence to obtain a matching result;

It is determined based on the matching result whether the detection result satisfies a preset condition.
The method according to claim 9, wherein determining whether the detection result satisfies a preset condition based on the matching result includes at least one of the following:

In response to a matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being greater than a preset matching degree, it is determined that the detection result does not satisfy the preset condition;

In response to the matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being less than or equal to the preset matching degree, obtain the user's judgment result about the detection result, and respond to the The judgment result is a positive result indicating that the detection result is correct, and it is determined that the detection result satisfies the preset condition.
The method according to any one of claims 1 to 10, wherein the historical detection results include historical misdetection results with incorrect detection results, and the acquisition of the detection results of the image to be detected is performed by a detection model, so The above methods also include:

Receive instructions to update the detection model;

The detection model is updated in response to the instruction, and the updated detection model is trained by combining the historical false detection results with the historical to-be-detected images corresponding to the historical false detection results.
The method according to any one of claims 2 to 11, wherein the detection result is an object detection result and the detection result type is an object type; or the detection result is an event detection result and the detection result Type is event type.
A test result processing device, including:

The detection part is configured to obtain the detection results of the image to be detected;

The judgment part is configured to judge whether the detection results meet the preset conditions based on the historical detection results;

The execution part is configured to execute the preset processing corresponding to the detection result in response to the detection result satisfying the preset condition.
The device according to claim 13, wherein the detection result includes a detection result type and an accuracy corresponding to the detection result type;

The judgment part is further configured to: judge whether the accuracy corresponding to the detection result type is greater than or equal to a preset accuracy, where the preset accuracy is obtained based on the historical accuracy in the historical detection results; In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, it is determined that the detection result satisfies the preset condition.
The device according to claim 13, wherein the detection result includes a detection result type and an accuracy corresponding to the detection result type;

The judgment part is further configured to: judge whether the accuracy corresponding to the detection result type is greater than or equal to a preset accuracy, where the preset accuracy is obtained based on the historical accuracy in the historical detection results; In response to the accuracy corresponding to the detection result type being greater than or equal to the preset accuracy, obtaining the user's judgment result regarding the detection result; in response to the judgment result being a positive result indicating that the detection result is correct, determining The detection results meet the preset conditions.
The device according to claim 14 or 15, wherein,

The judgment part is further configured to: determine the preset accuracy based on the historical accuracy corresponding to historical detection results including the detection result type, wherein the historical accuracy corresponding to different detection result types is the same or different .
The device of claim 16, wherein:

The judgment part is also configured to: obtain the historical accuracy of each historical misdetection result including the detection result type; average the historical accuracy of each of the historical misdetection results to obtain the preset accuracy Spend.
The device according to claim 16 or 17, wherein,

The judgment part is further configured to: in response to the judgment result being a negative result indicating that the detection result is incorrect, use the detection result as a historical misdetection result.
The device according to any one of claims 14 to 16, wherein

The judgment part is also configured to: determine the false detection rate corresponding to each historical detection result within the preset historical time period, the false detection rate is based on the number of historical false detection results within the preset time period and the total history The number of test results is obtained;

The detection part is further configured to: in response to the false detection rate being greater than the preset false detection rate, execute the acquisition of the detection result of the image to be detected.
The method of claim 19, wherein:

The detection part is further configured to: obtain the detection result of the image to be detected when it is determined that the false detection rate is less than or equal to the preset false detection rate;

The execution part is also configured to execute preset processing corresponding to the detection result.
The device according to any one of claims 13 to 20, wherein the historical detection results include historical false detection results with incorrect detection results;

The judgment part is also configured to: obtain a target feature sequence corresponding to the detection result. The target feature sequence is a feature sequence extracted from the image to be detected corresponding to the historical misdetection result. The target feature sequence Used to obtain the detection result corresponding to the image to be detected; match the feature sequence of the image to be detected corresponding to the detection result with the target feature sequence to obtain a matching result; determine whether the detection result satisfies the predetermined condition based on the matching result. Set conditions.
The device according to claim 21, wherein the detection part is further configured to be at least one of the following:

In response to a matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being greater than a preset matching degree, it is determined that the detection result does not satisfy the preset condition;

In response to the matching degree between the feature sequence of the image to be detected corresponding to the detection result and the target feature sequence being less than or equal to the preset matching degree, obtain the user's judgment result about the detection result, and respond to the The judgment result is a positive result indicating that the detection result is correct, and it is determined that the detection result satisfies the preset condition.
The device according to any one of claims 13 to 22, wherein the historical detection results include historical misdetection results with incorrect detection results, and the acquisition of the detection results of the image to be detected is performed by a detection model, so The device also includes an update part;

The update part is also configured to: receive an instruction to update the detection model; update the detection model in response to the instruction, and the updated detection model is the combination of the historical false detection results and the It is obtained by training on the historical images to be detected corresponding to the historical false detection results.
The device according to any one of claims 14 to 23, wherein the detection result is an object detection result, and the detection result type is an object type; or the detection result is an event detection result, and the detection result Type is event type.
An electronic device includes a memory and a processor. The processor is configured to execute program instructions stored in the memory to implement the method according to any one of claims 1 to 12.
A computer-readable storage medium having program instructions stored thereon. When the program instructions are executed by a processor, the method of any one of claims 1 to 12 is implemented.
A computer program product, the computer program product comprising a computer program or instructions, when the computer program or instructions are run on an electronic device, causing the electronic device to execute any one of claims 1 to 12 Methods.