Method and device for determining image recognition false positive rate, electronic equipment and medium

Abstract

The application relates to a method and device for determining an image recognition false positive rate, electronic equipment and a medium, and is applied to the technical field of image recognition. The method comprises the following steps: acquiring event information corresponding to a plurality of garbage classification violation events in a first time interval, including an event type and an event image; determining a target key detection factor according to the event type of a first garbage classification violation event, the first garbage classification violation event being any event in the plurality of garbage classification violation events; judging whether the event image of the first garbage classification violation event matches the target key detection factor; if not, determining that a false positive judgment result is a false positive; if yes, determining that the false positive judgment result is a non-false positive; and determining a false positive rate according to the false positive judgment result corresponding to each garbage classification violation event. The application can accurately determine the false positive rate of the garbage classification violation event, can verify the accuracy of the garbage classification violation event recognition algorithm, and improves the verification efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, in particular to a method and device for determining image recognition false positive rate, electronic equipment and medium. BACKGROUND

[0002] In the garbage classification industry, AI (Artificial Intelligence) image recognition technology has been widely used, which includes machine learning, deep learning, algorithm vision and image dataset, etc. It mainly identifies garbage images on the algorithm side to realize efficient, accurate and automatic garbage classification, recycling and processing, saving manpower and time cost, and improving environmental benefits.

[0003] The related technical solution is that the garbage classification model is used to classify garbage. When the trained garbage classification model is applied to the actual business scenario, the garbage classification result is obtained by using the garbage classification model according to the actual scene picture. Then, the garbage classification result is analyzed and classified by the violation event identification algorithm to generate a series of garbage classification related violation events, such as "garbage mixed throwing", "garbage missing", "recyclable material classification error", etc. At the same time, the event information corresponding to the violation event is obtained, such as event type, event time and collected image, etc. Then the generated garbage classification violation event and event information are uploaded to the front-end platform for display, and the manual verification and analysis of the violation event judgment result are carried out to accurately determine whether the event is really in violation of the rules, determine the false positive rate of the violation event, and feed back the false positive rate to the algorithm personnel for adjusting and optimizing the violation event identification algorithm. However, in the pilot community, many violation events can be generated in a short time, and this operation needs to consume a lot of manpower cost and has low efficiency. SUMMARY

[0004] In order to solve the problem that the accuracy of the garbage classification violation event identification algorithm needs to rely on manual verification in the prior art, resulting in low verification efficiency, the present application provides a method and device for determining image recognition false positive rate, electronic equipment and medium.

[0005] In the first aspect, the present application provides a method for determining image recognition false positive rate, which adopts the following technical solution:

[0006] A method for determining image recognition false positive rate, comprising:

[0007] Obtaining event information corresponding to each of a plurality of garbage classification violation events in a first time interval, wherein the event information includes event type and event image;

[0008] determine a target key detection factor according to an event type of a first garbage classification violation event; wherein the first garbage classification violation event is any event in the plurality of garbage classification violation events;

[0009] determine whether an event image of the first garbage classification violation event matches the target key detection factor; if not, determine a false positive determination result as a false positive; if yes, determine the false positive determination result as a non-false positive;

[0010] determine a false positive rate according to false positive determination results corresponding to respective garbage classification violation events.

[0011] By adopting the above technical solution, first, a key detection factor is determined according to an event type of a violation event. The key detection factor is a key feature that must exist or cannot exist in an event image corresponding to the event type. A feature that must exist in an event image corresponding to a certain event type is a positive key detection factor, and a feature that cannot exist is a negative key detection factor. When a positive key detection factor cannot be detected and / or a negative key detection factor is detected in an event image, it is determined that the corresponding violation event is a false positive. By using the key detection factor to screen false positives of violation events, an accurate garbage classification violation event false positive rate can be obtained to accurately verify a garbage classification violation event identification algorithm.

[0012] In a preferred example, the application can be further configured to: the false positive rate is determined according to false positive determination results corresponding to respective garbage classification violation events, including:

[0013] determine a second garbage classification violation event as a garbage classification violation event that is determined as a non-false positive in a first time interval;

[0014] divide the first time interval into a plurality of second time intervals;

[0015] when it is detected that the number of second garbage classification violation events of a same event type in a target second time interval exceeds a preset number threshold, adjust false positive determination results of the second garbage classification violation events of the same event type in the target second time interval according to the preset number threshold; wherein the target second time interval is any interval in the plurality of second time intervals;

[0016] determine a false positive rate according to latest false positive determination results corresponding to respective garbage classification violation events.

[0017] By adopting the technical scheme, part of the false positive events can be accurately screened out through the key detection factors, but since the number of the key detection factors is limited and cannot represent all the features in the event image, false positive events may exist in the events determined as non-false positive events through the key detection factors. Therefore, the scheme divides the second time interval. Generally, the number of continuous events of the same type in the second time interval is limited. When the number of continuous events of the same type in the second time interval exceeds the preset number threshold, the events exceeding the preset number threshold are determined as repeated events, and the repeated events are determined as false positive events, which can exclude too many repeated events in a short time due to network or algorithm problems, and improve the accuracy of the false positive rate.

[0018] In a preferred example, the application can be further configured to determine the false positive rate according to the latest false positive determination result of each garbage classification violation event, including:

[0019] The first time interval is divided into a plurality of third time intervals, wherein the third time interval is greater than the second time interval;

[0020] When it is detected that the number of third garbage classification violation events of the same event type in a target third time interval exceeds the preset number threshold, the false positive determination result of the second garbage classification violation event of the same event type in the target third time interval is adjusted based on the event image of the third garbage classification violation event corresponding to the target third time interval and the preset number threshold. The third garbage classification violation event is the violation event after the false positive determination result of the second garbage classification violation event is adjusted according to the preset number threshold, and the target third time interval is any interval in the plurality of third time intervals.

[0021] The false positive rate is determined according to the latest false positive determination result of each garbage classification violation event.

[0022] By adopting the technical scheme, the first time interval is divided into a third time interval larger than the second time interval, so as to repeatedly detect the violation event after the false positive determination result is adjusted by the number threshold in the previous example, which can improve the accuracy of the false positive rate determination.

[0023] In a preferred example, the application can be further configured to adjust the false positive determination result of the third garbage classification violation event of the same event type in the target third time interval based on the event image of the third garbage classification violation event corresponding to the target third time interval and the preset number threshold, including:

[0024] determine, based on the event images of the third garbage classification violation events corresponding to the target third time interval, a number of continuous similar events from the third garbage classification violation events of the same event type that are continuous in the target third time interval;

[0025] when it is detected that the number of continuous similar events exceeds the preset number threshold, adjust the false positive determination result of the continuous similar events according to the preset number threshold.

[0026] By adopting the above technical solution, the similarity of the event images of multiple events is determined, and similar events, i.e., repeated events, can be screened out from the multiple third time interval continuous same event type violation events. By adjusting the false positive determination result of the similar events, the accuracy of the false positive rate can be further improved.

[0027] In a preferred example, the application can be further configured to: the determination, based on the event images of the third garbage classification violation events corresponding to the target third time interval, of the number of continuous similar events from the third garbage classification violation events of the same event type that are continuous in the target third time interval includes:

[0028] perform feature extraction on the event images corresponding to the third garbage classification violation events of the same event type that are continuous in the target third time interval to obtain image features corresponding to each event image;

[0029] determine, based on the image features corresponding to each event image, a similarity of event images of two adjacent events from the multiple event images corresponding to the third garbage classification violation events of the same event type that are continuous in the target third time interval;

[0030] when the similarity of the event images of the two adjacent events exceeds a preset similarity threshold, determine that both of the two adjacent events are similar events;

[0031] determine, based on the similar event determination result, the number of continuous similar events from the third garbage classification violation events of the same event type that are continuous in the target third time interval.

[0032] By adopting the above technical solution, the similarity of each two event images is determined through the image features of the event images, and similar events can be accurately screened out from the multiple third time interval continuous same event type events, so that the number of similar events in the target third time interval continuous same event type third garbage classification violation events is more accurate.

[0033] In a preferred example, the application can be further configured to: the method further includes:

[0034] obtaining a plurality of abnormal situation detection factors; wherein the plurality of abnormal situation detection factors include at least one of the following: the number of garbage cans, the angle of the garbage can, the blur degree of the event image;

[0035] When the event image of the first garbage classification violation event satisfies any abnormal situation detection factor, determining that the first garbage classification violation event is an abnormal event; otherwise, determining that the first garbage classification violation event is a non-abnormal event;

[0036] storing the abnormality determination result of the first garbage classification violation event into a garbage classification violation event database.

[0037] By adopting the above technical solution, by setting abnormal situation detection factors, the abnormal situation in the event image corresponding to the garbage classification violation event is monitored in real time, the abnormality of the garbage classification garbage can and the monitoring equipment can be found in time, and the abnormal situation is uploaded to the database, so that the staff can check and on-site rectify the garbage classification abnormal situation.

[0038] In a preferred example, the application can be further configured to: after determining the false positive rate according to the false positive determination result of each garbage classification violation event, the method further includes at least one of the following:

[0039] storing the false positive determination result of each garbage classification violation event and the false positive rate into a garbage classification violation event database;

[0040] correcting the garbage classification violation event identification algorithm based on the false positive rate.

[0041] By adopting the above technical solution, the false positive determination result and the false positive rate of the garbage classification violation event are uploaded to the database, the staff can check the false positive event, the accuracy of the garbage classification violation event identification algorithm is verified according to the false positive rate, the efficiency of verifying the algorithm is improved, the garbage classification violation event identification algorithm is corrected, and the accuracy of the garbage classification identification algorithm is improved.

[0042] In a second aspect, the application provides a determination device for image recognition false positive rate, which adopts the following technical solution:

[0043] A determination device for image recognition false positive rate, comprising:

[0044] An obtaining module, configured to obtain event information corresponding to a plurality of garbage classification violation events in a first time interval, wherein the event information includes event type and event image;

[0045] The first determining module is configured to determine a target key detection factor according to an event type of a first garbage classification violation event, wherein the first garbage classification violation event is any event in the plurality of garbage classification violation events.

[0046] The determining module is configured to determine whether an event image of the first garbage classification violation event matches the target key detection factor, and if not, determine that the false positive determination result is a false positive, and if yes, determine that the false positive determination result is a non-false positive.

[0047] The second determining module is configured to determine a false positive rate according to the false positive determination result corresponding to each garbage classification violation event.

[0048] In a third aspect, the present application provides an electronic device, which adopts the following technical solution:

[0049] one or more processors;

[0050] a memory;

[0051] one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the image recognition false positive rate determination method according to any one of the first aspect.

[0052] In a fourth aspect, the present application provides a computer readable storage medium, which adopts the following technical solution:

[0053] A computer readable storage medium having a computer program stored thereon, when the computer program is executed in a computer, the computer executes the image recognition false positive rate determination method according to any one of the first aspect.

[0054] In summary, the present application has the following beneficial technical effects:

[0055] According to the present application, the target key detection factor is determined according to the event type of the plurality of garbage classification violation events in the first time interval, the key detection factor is a key feature that must exist or cannot exist in the event image corresponding to the event type, it is determined whether the event image of the first garbage classification violation event matches the target key detection factor, so as to determine whether the garbage classification violation event is a false positive, and then the false positive rate is determined according to the false positive determination result corresponding to each garbage classification violation event, which improves the accuracy of the false positive rate, verifies the accuracy of the garbage classification violation event identification algorithm through the false positive rate, reduces the labor cost, and improves the verification efficiency of the garbage classification violation event identification algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1is a flowchart of a method for determining an image recognition false positive rate according to an embodiment of the present application;

[0057] Figure 2 is a flowchart of a method for determining a false positive rate according to an embodiment of the present application;

[0058] Figure 3 is a flowchart of a method for updating a false positive determination result according to an embodiment of the present application;

[0059] Figure 4 is a flowchart of a method for adjusting a false positive determination result of a third garbage classification violation event of a same event type in a target third time interval according to an embodiment of the present application;

[0060] Figure 5 is a flowchart of a method for determining an event quantity of similar events according to an embodiment of the present application;

[0061] Figure 6 is a flowchart of a method for detecting an abnormal situation according to an embodiment of the present application;

[0062] Figure 7 is a flowchart of a method for determining an image recognition false positive rate according to an embodiment of the present application;

[0063] Figure 8 is a structural diagram of a device for determining an image recognition false positive rate according to an embodiment of the present application;

[0064] Figure 9 is a structural diagram of a device for determining an image recognition false positive rate according to an embodiment of the present application;

[0065] Figure 10 is a structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0066] The present application will be further described below in conjunction with the accompanying drawings.

[0067] The present embodiment is merely an explanation of the present application, and is not a limitation of the present application. Those skilled in the art can make modifications to the present embodiment without creative contribution after reading the present specification, and the present application is protected by the patent law as long as the modifications are within the scope of the claims of the present application.

[0068] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0069] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0070] In real-world business scenarios, waste sorting violation event identification algorithms recognize the results of waste sorting and generate multiple waste sorting violation events and normal events. However, due to network latency or algorithm errors, the waste sorting violation events generated by the algorithm may produce false alarms. That is, some events that are not in violation are judged as violations by the identification algorithm, or the violation type of the violation event is incorrectly determined by the identification algorithm.

[0071] Therefore, multiple violations generated by the garbage sorting violation identification algorithm can be judged as false alarms, and the false alarm rate of garbage sorting violation can be determined based on the false alarm judgment results, so as to verify the accuracy of the garbage sorting violation identification algorithm through the false alarm rate.

[0072] This application provides a method for determining the false alarm rate of image recognition, such as... Figure 1 As shown, the method provided in this application embodiment can be executed by an electronic device, which can be a server or a terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smartphone, tablet, laptop, desktop computer, etc., but is not limited to these. The terminal device and the server can be directly or indirectly connected via wired or wireless communication. This application embodiment does not impose any limitations on this connection. The method includes:

[0073] Step S101: Obtain the event information corresponding to each of the multiple garbage sorting violation events within the first time interval, wherein the event information includes the event type and the event image.

[0074] The garbage classification violation event identification algorithm uploads the generated garbage classification violation event information to the garbage classification violation event database. The event type of the garbage classification violation event includes garbage mixed investment, garbage packet loss, etc. Garbage mixed investment indicates the violation of putting other types of garbage into the garbage bin. Garbage packet loss indicates the violation of throwing garbage on the ground. The event image is collected by the image collection device set at the garbage throwing point. The image collection device can be a monitoring camera set near the garbage throwing point, which is used to collect garbage bin pictures and garbage ground pictures and video information. One event image can be one or more, which is related to the number of image collection devices set at the garbage throwing point. If two image collection devices monitor the same garbage throwing area, two event images with different shooting angles of the same event can be obtained at the same time. At the same time, the image collection device will record the event time corresponding to the collected event image. For each event, its event type, event image and event time correspond to each other.

[0075] The embodiment determines whether the garbage classification violation event is false. The event information of the historical garbage classification violation event in the first time interval can be obtained from the violation event database. The event set belonging to the same event type in the specified time period can also be retrieved according to the actual demand.

[0076] Specifically, the embodiment does not limit the specific length of the first time interval, which can be set according to the actual business scene. For example, the first time interval can be several hours to several days.

[0077] Step S102, determining the target key detection factor according to the event type of the first garbage classification violation event; wherein the first garbage classification violation event is any event in the plurality of garbage classification violation events.

[0078] The embodiment automatically verifies the determination result of the garbage classification violation event according to the event information of the garbage classification violation event. Before the false positive determination of the violation event, a key detection factor can be set in advance, the key detection factor corresponds to an event type of the garbage classification violation event, and any event type can correspond to one or more key detection factors. The key detection factor is divided into a positive key detection factor and a negative key detection factor. The positive key detection factor is a key feature that must exist in the event image corresponding to the event type, and the negative key detection factor is a key feature that cannot exist in the event image corresponding to the event type. The key detection factor corresponding to the event type can be set according to historical garbage classification violation event data, and can also be set artificially according to actual experience. The same event type can contain a positive key detection factor and a negative key detection factor at the same time, and the embodiment of the application is not limited. When the false positive determination of the violation event is performed, the key detection factor corresponding to the event type can be determined through the event type of the violation event.

[0079] In step S103, it is determined whether the event image of the first garbage classification violation event matches the target key detection factor. If not, it is determined that the false positive determination result is false positive. If yes, it is determined that the false positive determination result is not false positive.

[0080] The first garbage classification violation event is any event in a plurality of garbage classification violation events. When the key detection factor corresponding to the event type of the first garbage classification violation event is a positive key detection factor, feature extraction and recognition are performed on the event image. If it is detected that the event image of the first garbage classification violation event does not exist the feature image corresponding to the positive key detection factor, it is determined that the first garbage classification violation event is a false positive event, otherwise, it is determined that the first garbage classification violation event is not a false positive event. For example, for an event whose event type is littering garbage bags, suppose the corresponding positive key detection factor is a garbage bag. If there is no image of a garbage bag on the ground in the event image, it is determined that the violation event is a false positive event.

[0081] When the key detection factor corresponding to the event type of the first garbage classification violation event is a negative key detection factor, feature extraction and recognition are performed on the event image. If it is detected that the event image of the first garbage classification violation event exists the feature image corresponding to the negative key detection factor, it is determined that the first garbage classification violation event is a false positive event, otherwise, it is determined that the first garbage classification violation event is not a false positive event.

[0082] When the first garbage classification violation event contains both the positive key detection factor and the negative key detection factor, if it is detected that there is no feature image corresponding to the positive key detection factor in the event image of the first garbage classification violation event, or there is a feature image corresponding to the negative key detection factor, it is determined that the event is a false positive; if it is detected that there is a feature image corresponding to the positive key detection factor in the event image of the first garbage classification violation event, and there is no feature image corresponding to the negative key detection factor, it is determined that the event is not a false positive.

[0083] In step S104, the false positive rate is determined according to the false positive determination result corresponding to each garbage classification violation event.

[0084] Based on the event information of the garbage classification violation event, the false positive event can be accurately screened out. According to the total number of garbage classification violation events in the first time interval and the number of false positive events, the false positive rate of the garbage classification violation event in the first time interval can be determined. In combination with the event type, the false positive rate of the garbage classification violation event of a certain event type in the first event interval can also be determined. The present embodiment can obtain the accurate false positive rate of the garbage classification violation event by screening the false positive events of the violation event based on the key detection factor, so as to accurately verify the garbage classification violation event identification algorithm.

[0085] Further, in order to facilitate understanding, in one possible implementation manner of the present embodiment, the event type of the garbage classification violation event includes garbage mixed throwing and garbage missing, and the garbage mixed throwing can be specifically classified into which garbage mixed throwing, which can include: kitchen waste mixed throwing, recyclable garbage mixed throwing, hazardous garbage mixed throwing, and other garbage mixed throwing. For example, the kitchen waste mixed throwing is the behavior of throwing other types of garbage into the kitchen waste bin. The garbage missing refers to the behavior of throwing garbage on the ground outside the garbage bin in the monitoring area of the image acquisition device at the garbage classification site.

[0086] Specifically, the positive key detection factor of the garbage mixed throwing can be set as the garbage bin opening image corresponding to the garbage type, wherein the garbage type is one of kitchen waste, recyclable waste, hazardous waste, and other waste. It is specified that the first detection area of the garbage mixed throwing event must contain the garbage bin placement area, for example, for the kitchen waste mixed throwing event, the event image collected in the detection area must contain the kitchen waste bin opening, otherwise, it is determined as a false positive event. Specifically, in the process of detecting by using the algorithm, after the detection result is obtained, the image of the area where the garbage is located in the input image is cut out as the event image. In addition, for any garbage throwing behavior, multiple images are collected, and then the multiple images are identified, the image of the garbage bin opening at the time when the person throws the garbage into the garbage bin to the time when the human body completely leaves the image collection device detection area is selected as the event image of this garbage classification event, and then the garbage in the garbage bin is identified and classified to determine whether the garbage thrown this time is correctly classified. If the garbage classification is not correct, the garbage classification violation event recognition algorithm generates a garbage classification violation event with the event type of garbage mixed throwing. If the generated event image contains a human body image, it will interfere with the garbage classification identification, resulting in inaccurate violation event judgment. Therefore, the human body can also be set as the negative key detection factor of the garbage mixed throwing event, and when the human body image is detected in the event image, it is determined that the event is a false positive.

[0087] For the violation event with the event type of garbage mixed throwing, the garbage bin opening corresponding to the garbage type can be set as the key detection factor of the event type, and the garbage bin opening corresponding to the garbage type and the human body can also be set as the key detection factor of the garbage mixed throwing. The present embodiment is not limited.

[0088] In some cases, there is a situation of throwing garbage randomly near the garbage bin, i.e., garbage throwing event. For the event with the event type of garbage throwing, the image collection device monitoring area is the empty land beside the garbage bin placement area. The garbage and the human body can be set as the positive key detection factor, indicating that there is a human throwing garbage behavior. When the human body and the garbage exist in the obtained event image at the same time, it can be determined as the garbage throwing. The face can also be set as the positive key detection factor of the garbage throwing event, so as to track the face of the garbage classification violation event through face recognition, and determine the identity information of the person who commits the violation event.

[0089] In addition, the event type of the garbage classification violation event in the present application is not limited to garbage mixed throwing and garbage throwing. Correspondingly, the key detection factor corresponding to the event type can be one or more, which can be flexibly set according to the specific business scene, and the present embodiment is not limited.

[0090] In a possible implementation manner of the present embodiment, the false positive rate is determined according to the false positive judgment result of each garbage classification violation event in step S104, such as Figure 2As shown, comprising steps Sa1-Sa3, wherein:

[0091] Sa1, the first time interval is divided into several second time intervals.

[0092] In actual scenarios, the probability of continuously appearing multiple violation events of the same event type in a short period of time is small. Due to network lag or algorithm error, etc., the same event may be identified as multiple events, resulting in continuously appearing multiple violation events of the same event type in a short period of time. For example: in the same garbage classification detection area, the probability of continuously appearing two or more garbage mixed throwing events within one second is approximately zero, at which time it is determined that there are repeated events in the two or more garbage mixed throwing events.

[0093] In this embodiment, the first time interval is divided into several second time intervals according to the event time, and the number of second time intervals is greater than one, wherein each second time interval is equal in length, and the specific duration of the second time interval is not limited by the present application. Preferably, the second time interval is set to one second.

[0094] Sa2, when the number of consecutive second garbage classification violation events of the same event type in the target second time interval exceeds the preset number threshold, adjusting the false positive determination result of the consecutive second garbage classification violation events of the same event type in the target second time interval according to the preset number threshold; wherein the target second time interval is any interval in the plurality of second time intervals.

[0095] Specifically, the preset number threshold can be set to represent the maximum number of violation events of the same event type that occur consecutively in a second time interval. The specific value of the preset number threshold can be set artificially based on experience in actual application scenarios. Preferably, the preset number threshold is set to one.

[0096] When the number of consecutive second garbage classification violation events of the same event type in a second time interval exceeds the preset number threshold, the events exceeding the preset number threshold are determined to be repeated events. The repeated events can be determined to be false positive events, and the remaining number of events is determined to be non-false positive events.

[0097] Sa3, determining the false positive rate according to the latest false positive determination result of each garbage classification violation event.

[0098] This embodiment performs repeated determination on the violation events determined to be non-false positive by the key detection factor, improves the determination accuracy of false positive events, and can obtain a more accurate false positive rate.

[0099] For the case that the number of second garbage classification violation events of the same event type in the target second time interval is not more than the preset number threshold, one is not to adjust the false positive determination result, and the other is to further determine to improve the accuracy. For details, refer to the following embodiments.

[0100] In one possible implementation of the embodiments of the present application, step Sa3 determines the false positive rate according to the latest false positive determination result corresponding to each garbage classification violation event, as shown in the following table. Figure 3

[0101] Sb1, divide the first time interval into a plurality of third time intervals, wherein the third time interval is greater than the second time interval.

[0102] The above embodiments adjust the false positive result in combination with the number of events in the second time interval. In this embodiment, a larger third time interval is divided to perform secondary detection on the violation events after the number threshold adjustment of the false positive determination result in the previous example, which can improve the accuracy of the false positive rate determination.

[0103] Specifically, the third time interval can be flexibly set according to actual needs, and preferably, the third time interval can be set to three seconds.

[0104] Sb2, when it is detected that the number of third garbage classification violation events of the same event type in the target third time interval exceeds the preset number threshold, adjust the false positive determination result of the third garbage classification violation event of the same event type in the target third time interval based on the event image of the target third garbage classification violation event and the preset number threshold.

[0105] Wherein, the third garbage classification violation event is the violation event after adjusting the false positive determination result of the second garbage classification violation event according to the preset number threshold, and the target third time interval is any interval in the plurality of third time intervals.

[0106] Specifically, for the violation events after adjusting the false positive determination result by the number threshold in the previous example, when it is detected that the number of third garbage classification violation events of the same event type in the target third time interval exceeds the preset number threshold, it is preliminarily determined that there are repeated events in these violation events. The number of violation events of the same type in the third time interval can be further combined with the event image to screen out similar events in the violation events whose number exceeds the preset number threshold, and the similar events are the events determined repeatedly, and the false positive determination result is adjusted based on the similar events.

[0107] Sb3, determine the false positive rate according to the latest false positive determination result corresponding to each garbage classification violation event.

[0108] ​By adopting the technical solution, the first time interval is divided into a third time interval larger than the second time interval, so as to repeatedly detect the violation event after the false positive determination result of the previous example is adjusted by the quantity threshold value, and the accuracy of the false positive rate determination can be improved.

[0109] In a possible implementation of the embodiment of the present application, based on the event image of the third garbage classification violation event corresponding to the target third time interval and the preset quantity threshold value, the false positive determination result of the third garbage classification violation event of the same event type in the target third time interval is adjusted, such as Figure 4 As shown, the method comprises steps Sc1-Sc2, wherein:

[0110] Sc1, based on the event image of the third garbage classification violation event corresponding to the target third time interval, determining the number of consecutive similar events from the third garbage classification violation events of the same event type in the target third time interval.

[0111] The similar event is an event whose similarity to the event image of the adjacent event exceeds the preset similarity threshold value.

[0112] For the embodiment of the present application, it is determined whether the number of the third garbage classification violation events of the same event type in the target third time interval is greater than the preset quantity threshold value;

[0113] If yes, the similarity between the adjacent events is determined according to the event image corresponding to the third garbage classification violation event of the same event type, and if the similarity between the adjacent events reaches the preset similarity threshold value, the adjacent events are determined as similar events, and the number of similar events is obtained.

[0114] For example, the third garbage classification violation events of the same event type in the target third time interval are [a1, a2, a3, a4, a5], if the similarity of events a1 and a2 reaches the preset similarity threshold value, the similarity of events a2 and a3 does not reach the preset similarity threshold value, the similarity of events a3 and a4 reaches the preset similarity threshold value, and the similarity of events a4 and a5 reaches the preset similarity threshold value, at this time, events a1 and a2 are similar, and the number of consecutive similar events is 2; events a3, a4 and a5 are similar, and the number of consecutive similar events is 3.

[0115] In the previous embodiment, the number of violation events of the same event type in the second time interval is compared with the preset number threshold, and repeated events can be screened out, and the repeated events are determined as false positive events. In this embodiment, repeated detection is performed to further improve the accuracy of false positive event screening. However, since the third time interval is larger than the second time interval, the number of violation events of the same event type in the time interval is compared with the preset number threshold, and the accuracy is not high, so the event image similarity can be introduced to screen the repeated events.

[0116] Sc2, when the number of consecutive similar events is detected to exceed the preset number threshold, adjusting the false positive determination result of the consecutive similar events according to the preset number threshold.

[0117] For example, the first event of the obtained consecutive similar events can be selected as a non-false positive event, and the number of remaining violation events is determined as the number of false positive events.

[0118] According to the above embodiment, events a1 and a2 are similar, the number of consecutive similar events is 2, the preset number threshold is 1, the number of consecutive similar events exceeds the preset number threshold, and the false positive determination result of a2 is adjusted to false positive; events a3, a4 and a5 are similar, the number of consecutive similar events is 3, and the number of consecutive similar events exceeds the preset number threshold, and the false positive determination results of a4 and a5 are adjusted to false positive.

[0119] When the preset number threshold is 2, events a1 and a2 are similar, the number of consecutive similar events is 2, the number of consecutive similar events exceeds the preset number threshold, and a5 is not adjusted, events a3, a4 and a5 are similar, the number of consecutive similar events is 3, the number of consecutive similar events exceeds the preset number threshold, and the false positive determination result of a5 is adjusted to false positive.

[0120] In this embodiment, the similarity of the event images of multiple events is determined, and the events with high similarity, i.e. repeated events, can be screened out from the consecutive violation events of the same event type in multiple third time intervals. By adjusting the false positive determination result of the similar events, the accuracy of the false positive rate can be further improved.

[0121] In one possible implementation of the embodiment of the present application, based on the event images of the third garbage classification violation events corresponding to the target third time interval, the number of similar events is determined from the consecutive third garbage classification violation events of the same event type in the target third time interval, as shown in Figure 5 The method comprises steps Sd1-Sd3, wherein:

[0122] Sd1, feature extraction is performed on the event images corresponding to the third garbage classification violation events of the same event type that are continuous in the target third time interval, to obtain image features corresponding to each event image; and based on the image features corresponding to each event image, the similarity of the event images of each two adjacent events in the multiple event images corresponding to the third garbage classification violation events of the same event type that are continuous in the target third time interval is determined.

[0123] In this embodiment, feature extraction can be performed on the event images corresponding to the second garbage classification violation events of the same type that are continuous in the third time interval, and at least two important features, such as garbage features and garbage can opening features, are extracted for each event image. The features of two adjacent events in the multiple event images are compared to obtain the corresponding similarity.

[0124] Sd2, when the similarity of the event images of two adjacent events exceeds a preset similarity threshold, it is determined that both of the two adjacent events are similar events.

[0125] The event whose similarity exceeds the preset similarity threshold is a similar event, i.e., a repeated event. The preset similarity threshold can be set according to actual needs, and preferably, the preset similarity threshold is 95%.

[0126] Specifically, when the similarity of this event image and the event image of its adjacent event exceeds the preset similarity threshold, the two adjacent events corresponding to the two event images are the same event, and it is determined that the two adjacent events are similar events.

[0127] Sd3, based on the similar event determination result, the number of events of the similar events in the third garbage classification violation events of the same event type that are continuous in the target third time interval is determined.

[0128] This embodiment can accurately filter out similar events from the multiple events of the same event type that are continuous in the third time interval, so that the number of similar events in the third garbage classification violation events of the same event type that are continuous in the target third time interval is more accurate.

[0129] One possible implementation of the embodiment of the present application, after obtaining the event information corresponding to each of the multiple garbage classification violation events in the first time interval, the method further comprises:

[0130] The first time interval is divided into a plurality of deduplication intervals, and when it is detected that the number of garbage classification violation events of the same event type that are continuous in a target deduplication interval exceeds a preset deduplication number threshold, the garbage classification violation events of the same event type that are continuous in the target deduplication interval and exceed the deduplication number threshold are deleted, to obtain the deduplicated multiple garbage classification violation events; wherein the target deduplication interval is any interval in the plurality of deduplication intervals.

[0131] Specifically, deduplication is to delete duplicate events. The larger the deduplication interval is, the smaller the deduplication quantity threshold is, and the less accurate the duplicate events are. Therefore, a shorter time is generally selected as the deduplication interval. The deduplication interval and the deduplication quantity threshold can be flexibly set according to actual needs. Preferably, the deduplication interval is set to 0.5 seconds, and the deduplication quantity threshold is set to two.

[0132] By deleting duplicate events from the initial garbage classification violation events, important and accurate garbage classification violation event data can be obtained to obtain an accurate violation event false positive rate.

[0133] As shown in FIG. 1, the method further includes steps Se1-Se3. Figure 6

[0134] Se1, obtaining a plurality of abnormal situation detection factors; wherein the plurality of abnormal situation detection factors include at least one of the following: a number of garbage cans, a garbage can angle, and an event image blur degree.

[0135] In actual business scenarios, the garbage classification drop-off point can generate abnormal situations, such as insufficient number of garbage cans, incorrect angle of garbage cans, and excessively blurred collected event images. The event images containing abnormal situations are determined as abnormal events.

[0136] Specifically, the abnormal situation detection factors are set to determine whether the garbage classification drop-off point has abnormal situations. The abnormal situation detection factors can also include a garbage can proportion, indicating that the proportion of a certain type of garbage can is too small, and a garbage can position. For specific event types, the abnormal situation detection factors can be flexibly set.

[0137] Illustratively, the number of garbage cans can be set at a garbage classification point. The number of garbage cans for each type of garbage is at least one. For a specific type of garbage that needs to be detected for garbage classification, the proportion of the number of garbage cans of the type can be set to be no less than 20% of the total number of garbage cans. The preset angle threshold is the maximum angle at which the garbage can can be tilted. The preset angle threshold can be set according to experience. Preferably, the preset angle threshold is set to 30°.

[0138] Se2, when it is detected that the event image of the first garbage classification violation event satisfies any abnormal situation detection factor, the first garbage classification violation event is determined as an abnormal event; otherwise, the first garbage classification violation event is determined as a non-abnormal event.

[0139] Specifically, one event type can correspond to one or more abnormal situation detection factors. When it is detected that the event image satisfies any abnormal situation detection factor, the event is determined as abnormal.

[0140] ​Se3, store the abnormality determination result of the first garbage classification violation event into the garbage classification violation event database.

[0141] In addition, a general abnormality detection factor of each event type can also be set, such as the blurring degree of the event image. When the collected event image is too blurred, it indicates that the image acquisition device is blurred or blocked, and needs to be handled on site, and the violation event corresponding to the blurred event image is marked as an abnormal event.

[0142] Specifically, for the event initially determined as abnormal, the image can be re-collected during the determination process, and the newly collected image can be identified by the garbage classification violation event identification algorithm, which can improve the detection accuracy. The abnormality identified is uploaded to the database, which facilitates the staff to check the garbage classification abnormality and to on-site rectify the image acquisition device and garbage can.

[0143] In the present application, each detection factor can be customized, turned on or off according to the business scenario, and each threshold value can be flexibly adjusted according to the actual situation.

[0144] In one possible implementation of the embodiment of the present application, after determining the false positive rate according to the false positive determination result of each garbage classification violation event, the method further includes at least one of the following:

[0145] Store the false positive determination result of each garbage classification violation event and the false positive rate into the garbage classification violation event database;

[0146] Correct the garbage classification violation event identification algorithm based on the false positive rate.

[0147] Specifically, the garbage classification violation event identification algorithm identifies the garbage classification result, determines whether the garbage classification event is in violation, uploads the violation event and the event type, event time, event image and other information of the violation event to the garbage classification violation event database, and uploads the business platform for classified display. The user can set the screening conditions and check the violation event of a certain type within a specified date at any time to obtain the total number of violation events of this type, the event generated by each event, the corresponding event image and other information.

[0148] The application judges the false positive of the identification result of the garbage classification violation event identification algorithm, obtains the false positive rate of the violation event in the first time interval, and the false positive rate of the violation event of different event types in the first time interval, and uploads the obtained false positive rate to the garbage classification violation event database. At the same time, the false positive judgment result of the event is associated and displayed, including the event type, event time, event image, whether it is a false positive, whether it is an anomaly, and which anomaly it belongs to, etc. The form of visual display can be set according to actual needs, such as only displaying the garbage bin placement position of the event in a certain date and time, so as to quickly check the occurrence of such abnormal conditions, cluster a certain type of situation, and intuitively and efficiently find the rules of various violation events. At this link, manual judgment can be intervened at any time, and if there is obvious misjudgment, the corresponding label of the event can be corrected, and the detection program can be adjusted in time.

[0149] The application can have two actual application scenarios. The first application scenario can apply the image recognition false positive rate determination method provided by the application to the business platform after displaying the violation event, so as to detect the accuracy of the garbage classification violation event identification algorithm, save the cost of manual verification, improve the efficiency, and also correct the garbage classification violation event identification algorithm based on the generated false positive rate.

[0150] Another application scenario of the application is that after the image recognition false positive rate determination method provided by the application is mature and stable, it can be used as a filter and set before the garbage classification violation event identification algorithm generates various violation events and before the generated violation events are uploaded to the business platform. Before the garbage classification violation event is uploaded to the business platform, it can be judged in advance whether the violation event is a false positive. If a false positive occurs, the false positive event can be intercepted before being uploaded to the business platform for display, so as to reduce the number and probability of false positive events uploaded to the business platform and reduce the workload of the business platform.

[0151] Specifically, the above two application scenarios can be performed simultaneously, that is, the image recognition false positive rate determination method provided by the application is set before and after uploading to the business platform, which can reduce the number and probability of false positive events uploaded to the business platform, and verify the accuracy of the garbage classification violation event identification algorithm.

[0152] Figure 7 The implementation flowchart of the image recognition false positive rate determination method provided by a specific embodiment of the application is as follows: Figure 7As shown, the embodiment obtains the classification result of the garbage classification violation event recognition algorithm, sets positive and negative key detection factors, determines that it is a non-misreport when the event image of the violation event matches the key detection factor, and otherwise, it is classified as a misreport event. For the non-misreport event, a second time interval is set, and when the number of consecutive events of the same type in the second time interval exceeds the number threshold, it is determined that a misreport is generated, otherwise, a third time interval is set, and when the number of consecutive events of the same type in the third time interval exceeds the number threshold, similarity determination is performed, and misreport determination is performed on similar events. On the other hand, the classification result of the garbage classification violation event recognition algorithm obtained is set with an abnormal situation detection factor, and when the event image of the violation event matches the abnormal situation detection factor, the violation event is marked as abnormal. Finally, the misreport events and the abnormal events obtained are summarized, the misreport rate is calculated, and the output is output.

[0153] The above embodiment introduces a method for determining the misreport rate of image recognition from the perspective of method flow. The following embodiment introduces a device for determining the misreport rate of image recognition from the perspective of virtual module or virtual unit. For details, see the following embodiment.

[0154] The embodiment of the present application provides a device for determining the misreport rate of image recognition, as shown in the figure, which can include: Figure 8

[0155] The acquisition module 801 is configured to acquire event information corresponding to each of a plurality of garbage classification violation events in a first time interval, wherein the event information includes an event type and an event image.

[0156] The first determination module 802 is configured to determine a target key detection factor according to the event type of a first garbage classification violation event, wherein the first garbage classification violation event is any event in the plurality of garbage classification violation events.

[0157] The judgment module 803 is configured to determine whether the event image of the first garbage classification violation event matches the target key detection factor. If not, the misreport determination result is determined to be a misreport. If yes, the misreport determination result is determined to be a non-misreport.

[0158] The second determination module 804 is configured to determine the misreport rate according to the misreport determination result corresponding to each garbage classification violation event.

[0159] In a preferred example, the second determination module 804 can be further configured to: when determining the misreport rate according to the misreport determination result corresponding to each garbage classification violation event, specifically for: taking the garbage classification violation event determined as a non-misreport in the first time interval as a second garbage classification violation event.

[0160] The first time interval is divided into a plurality of second time intervals. ​

[0161] when it is detected that the number of second garbage classification violation events of the same event type in the target second time interval exceeds the preset number threshold, adjusting the false positive judgment result of the second garbage classification violation events of the same event type in the target second time interval according to the preset number threshold; wherein the target second time interval is any interval of a plurality of second time intervals;

[0162] determining the false positive rate according to the latest false positive judgment result of each garbage classification violation event.

[0163] In a preferred example, the application can be further configured to: the second determination module 804, when determining the false positive rate according to the latest false positive judgment result of each garbage classification violation event, is specifically used for: dividing the first time interval into a plurality of third time intervals, wherein the third time interval is greater than the second time interval;

[0164] when it is detected that the number of third garbage classification violation events of the same event type in the target third time interval exceeds the preset number threshold, adjusting the false positive judgment result of the third garbage classification violation events of the same event type in the target third time interval based on the event image of the third garbage classification violation event corresponding to the target third time interval and the preset number threshold; wherein the third garbage classification violation event is the violation event after adjusting the false positive judgment result of the second garbage classification violation event according to the preset number threshold, and the target third time interval is any interval of a plurality of third time intervals;

[0165] determining the false positive rate according to the latest false positive judgment result of each garbage classification violation event.

[0166] In a preferred example, the application can be further configured to: the second determination module 804, when adjusting the false positive judgment result of the third garbage classification violation events of the same event type in the target third time interval based on the event image of the third garbage classification violation event corresponding to the target third time interval and the preset number threshold, is specifically used for:

[0167] determining the number of consecutive similar events from the third garbage classification violation events of the same event type in the target third time interval based on the event image of the third garbage classification violation event corresponding to the target third time interval;

[0168] when it is detected that the number of consecutive similar events exceeds the preset number threshold, adjusting the false positive judgment result of the consecutive similar events according to the preset number threshold.

[0169] The application can be further configured in a preferred example as follows: the second determination module 804 is specifically used for:

[0170] performing feature extraction on the event images corresponding to the third garbage classification violation events of the same event type in succession in the target third time interval, to obtain image features corresponding to each event image;

[0171] determining the similarity of the event images of each two adjacent events in the plurality of event images corresponding to the third garbage classification violation events of the same event type in succession in the target third time interval based on the image features corresponding to each event image;

[0172] when the similarity of the event images of the two adjacent events exceeds a preset similarity threshold, determining that both of the two adjacent events are similar events;

[0173] determining the number of similar events in the third garbage classification violation events of the same event type in succession in the target third time interval based on the similar event determination result.

[0174] The application can be further configured in a preferred example as follows: the apparatus further comprises:

[0175] an abnormal factor detection module, configured to: acquire a plurality of abnormal situation detection factors; wherein the plurality of abnormal situation detection factors include at least one of the following: a number of garbage cans, a garbage can angle, an event image blur degree;

[0176] when it is detected that the event image of the first garbage classification violation event satisfies any abnormal situation detection factor, determining that the first garbage classification violation event is an abnormal event; otherwise, determining that the first garbage classification violation event is a non-abnormal event;

[0177] storing the abnormality determination result of the first garbage classification violation event into the garbage classification violation event database.

[0178] The application can be further configured in a preferred example as follows: the apparatus further comprises:

[0179] a correction module, configured to:

[0180] store the false positive determination result and the false positive rate of each garbage classification violation event into the garbage classification violation event database;

[0181] correct the garbage classification violation event identification algorithm based on the false positive rate.

[0182] Figure 9This is a schematic diagram of a module of an image recognition false alarm rate determination device according to a specific embodiment of this application, as shown below. Figure 9 As shown, it includes: an algorithm recognition result module, i.e., an acquisition module, used to obtain event information and an image acquisition device for acquiring event information; a key detection factor module, i.e., a first determination module and a judgment module, used to determine false alarms for violation events based on key detection factors; a repeatability detection module, used to determine false alarms based on the comprehensive time interval, quantity, and similarity of violation events; an anomaly factor detection module, used to set anomaly detection factors and detect anomalies in event images; and a data calculation and display module, i.e., a second determination module, used to statistically analyze and classify the false alarm rate, event information, and anomaly labeling results, with the labeling result indicating whether the event is a false alarm event or an abnormal event, and what kind of anomaly it is; and image archiving indicating that the event images corresponding to the obtained violation events and abnormal events are uploaded to the violation event database.

[0183] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the image recognition false alarm rate determination device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0184] This application provides an electronic device, such as... Figure 10 As shown, Figure 10 The illustrated electronic device 1000 includes a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the electronic device 1000 may also include a transceiver 1004. It should be noted that in practical applications, the transceiver 1004 is not limited to one type, and the structure of this electronic device 1000 does not constitute a limitation on the embodiments of this application.

[0185] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0186] The bus 1002 can include a path that transmits information between the above-described components. The bus 1002 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 1002 can be divided into an address bus, a data bus, a control bus, etc. For convenience of representation, Figure 10 Only one thick line is used to represent the bus in the middle, but it does not mean that there is only one bus or one type of bus.

[0187] The memory 1003 can be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0188] The memory 1003 is used to store application program codes for implementing the scheme of the present application, and is controlled to execute by the processor 1001. The processor 1001 is used to execute the application program codes stored in the memory 1003 to realize the content shown in the foregoing information processing method for industry-university-research cooperation embodiment.

[0189] The electronic device includes, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a PDA (Personal Digital Assistant), a PAD (Tablet Personal Computer), a PMP (Portable Multimedia Player), a car terminal (such as a car navigation terminal), etc., and a fixed terminal such as a digital TV, a desktop computer, etc. It can also be a server, etc. Figure 10 The electronic device shown is only an example, and should not bring any limitation to the function and use range of the embodiments of the present application.

[0190] The embodiment of the present application provides a computer readable storage medium, which stores a computer program, and when the computer program is executed on a computer, the computer can execute the corresponding content in the foregoing method embodiment.

[0191] It should be understood that, although each step in the flowchart of the accompanying drawings is shown in sequence according to the indication of the arrow, these steps are not necessarily executed in sequence according to the indication of the arrow. Unless explicitly stated herein, the execution of these steps is not strictly limited in sequence, and can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.

[0192] The above is only part of the embodiments of the present application, and it should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for determining a false positive rate of image recognition, characterized in that, The method comprises: obtaining event information corresponding to each of a plurality of garbage classification violation events in a first time interval, wherein the event information comprises an event type and an event image; determining a target key detection factor according to an event type of a first garbage classification violation event, wherein the first garbage classification violation event is any event in the plurality of garbage classification violation events; determining a false positive judgment result as a false positive if the event image of the first garbage classification violation event does not match the target key detection factor, and determining the false positive judgment result as a non-false positive if the event image matches the target key detection factor; determining a false positive rate according to false positive judgment results corresponding to each garbage classification violation event; the determining of the false positive rate according to the false positive judgment results corresponding to each garbage classification violation event comprises: taking a garbage classification violation event determined as a non-false positive in the first time interval as a second garbage classification violation event; dividing the first time interval into a plurality of second time intervals; when it is detected that the number of second garbage classification violation events of a same event type continuously in a target second time interval exceeds a preset number threshold, adjusting false positive judgment results of the second garbage classification violation events of the same event type continuously in the target second time interval according to the preset number threshold, wherein the target second time interval is any interval in the plurality of second time intervals; determining a false positive rate according to latest false positive judgment results corresponding to each garbage classification violation event; the determining of the false positive rate according to the latest false positive judgment results corresponding to each garbage classification violation event comprises: dividing the first time interval into a plurality of third time intervals, wherein the third time interval is larger than the second time interval; when it is detected that the number of third garbage classification violation events of a same event type continuously in a target third time interval exceeds the preset number threshold, adjusting false positive judgment results of the third garbage classification violation events of the same event type continuously in the target third time interval based on event images of the third garbage classification violation events corresponding to the target third time interval and the preset number threshold, wherein the third garbage classification violation event is a violation event after the false positive judgment results of the second garbage classification violation events are adjusted according to the preset number threshold, and the target third time interval is any interval in the plurality of third time intervals; determining a false positive rate according to latest false positive judgment results corresponding to each garbage classification violation event; the adjusting of the false positive judgment results of the third garbage classification violation events of the same event type continuously in the target third time interval based on the event images of the third garbage classification violation events corresponding to the target third time interval and the preset number threshold comprises: determining the number of continuous similar events from the third garbage classification violation events of the same event type continuously in the target third time interval based on the event images of the third garbage classification violation events corresponding to the target third time interval. adjusting the false positive judgment result of the continuous similar events according to the preset number threshold when it is detected that the number of the continuous similar events exceeds the preset number threshold; the number of continuous similar events is determined from the continuous third garbage classification violation events of the same event type in the target third time interval based on the event images of the third garbage classification violation events corresponding to the target third time interval; feature extraction is performed on the event images corresponding to the continuous third garbage classification violation events of the same event type in the target third time interval to obtain image features corresponding to each event image; the similarity of the event images of two adjacent events is determined based on the image features corresponding to each event image; when the similarity of the event images of the two adjacent events exceeds a preset similarity threshold, it is determined that both of the two adjacent events are similar events; the number of continuous similar events in the continuous third garbage classification violation events of the same event type in the target third time interval is determined based on the determination result of similar events.

2. The method of claim 1, wherein, Further comprising: obtaining a plurality of abnormal situation detection factors; wherein the plurality of abnormal situation detection factors include at least one of the following: number of garbage cans, angle of garbage can, degree of blurring of event images; when it is detected that the event image of the first garbage classification violation event satisfies any abnormal situation detection factor, it is determined that the first garbage classification violation event is an abnormal event; otherwise, it is determined that the first garbage classification violation event is a non-abnormal event; storing the abnormality determination result of the first garbage classification violation event into a garbage classification violation event database.

3. The method of claim 1, wherein, After determining the false positive rate based on the false positive judgment results corresponding to each garbage classification violation event, the method further comprises at least one of the following: storing the false positive judgment results of the respective garbage classification violation events and the false positive rate into a garbage classification violation event database; correcting the garbage classification violation event identification algorithm based on the false positive rate.

4. A device for determining the false alarm rate of image recognition, characterized in that, comprising: an acquisition module configured to acquire event information corresponding to a plurality of garbage classification violation events in a first time interval, wherein the event information includes event type and event image; a first determination module configured to determine a target key detection factor according to the event type of a first garbage classification violation event, wherein the first garbage classification violation event is any event in the plurality of garbage classification violation events; a judgment module configured to determine whether the event image of the first garbage classification violation event matches the target key detection factor; if not, the false positive judgment result is determined to be false positive; if yes, the false positive judgment result is determined to be non-false positive; a second determination module configured to determine a false positive rate based on the false positive judgment results corresponding to each garbage classification violation event; the second determination module, when executing the determination of the false positive rate based on the false positive judgment results corresponding to each garbage classification violation event, is specifically configured to: determine, as second garbage sorting violation events, the garbage sorting violation events determined as non-false alarms in the first time interval; divide the first time interval into a plurality of second time intervals; when it is detected that the number of second garbage sorting violation events of a same event type in a target second time interval exceeds a preset number threshold, adjust false alarm determination results of the second garbage sorting violation events of the same event type in the target second time interval according to the preset number threshold, wherein the target second time interval is any one of the plurality of second time intervals; determine a false alarm rate according to the latest false alarm determination results of the respective garbage sorting violation events; the second determination module, when determining the false alarm rate according to the latest false alarm determination results of the respective garbage sorting violation events, is specifically configured to: divide the first time interval into a plurality of third time intervals, wherein the third time interval is greater than the second time interval; when it is detected that the number of third garbage sorting violation events of a same event type in a target third time interval exceeds the preset number threshold, adjust false alarm determination results of the third garbage sorting violation events of the same event type in the target third time interval based on event images of the third garbage sorting violation events corresponding to the target third time interval and the preset number threshold, wherein the third garbage sorting violation event is a violation event obtained by adjusting the false alarm determination results of the second garbage sorting violation events according to the preset number threshold, and the target third time interval is any one of the plurality of third time intervals; determine a false alarm rate according to the latest false alarm determination results of the respective garbage sorting violation events; the second determination module, when adjusting the false alarm determination results of the third garbage sorting violation events of the same event type in the target third time interval based on the event images of the third garbage sorting violation events corresponding to the target third time interval and the preset number threshold, is specifically configured to: determine the number of consecutive similar events from the third garbage sorting violation events of the same event type in the target third time interval based on the event images of the third garbage sorting violation events corresponding to the target third time interval; when it is detected that the number of consecutive similar events exceeds the preset number threshold, adjust false alarm determination results of the consecutive similar events according to the preset number threshold; the second determination module, when determining the number of consecutive similar events from the third garbage sorting violation events of the same event type in the target third time interval based on the event images of the third garbage sorting violation events corresponding to the target third time interval, is specifically configured to: perform feature extraction on the event images corresponding to the third garbage sorting violation events of the same event type in the target third time interval to obtain image features corresponding to each event image; and determine a false alarm rate according to the latest false alarm determination results of the respective garbage sorting violation events. determine, based on the image features corresponding to each of the event images, a similarity between event images of two adjacent events in the plurality of event images corresponding to the third garbage classification violation events of the same event type in the target third time interval; determine that the two adjacent events are both similar events when the similarity between the event images of the two adjacent events exceeds a preset similarity threshold; determine, based on the determination result of the similar events, an event quantity of the similar events in the third garbage classification violation events of the same event type in the target third time interval.

5. An electronic device, comprising: It comprises: one or more processors; a memory; one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the image recognition false positive rate determination method according to any one of claims 1 to 3.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed in the computer, the computer executes the image recognition false positive rate determination method according to any one of claims 1 to 3.

Images