Image processing method and device, electronic equipment, storage medium and program product

Abstract

This application provides an image processing method, apparatus, electronic device, storage medium, and program product. The image processing method includes: acquiring an image to be matched; the image to be matched includes a surveillance camera to be updated and marker information for marking the position of the surveillance camera to be updated in the image to be matched; determining a feature point matching relationship between the image to be matched and surveillance camera images in a preset surveillance camera image library; the preset surveillance camera image library includes multiple surveillance camera images, each of which includes a surveillance camera and marker information for marking the position of the surveillance camera in the corresponding surveillance camera image, and the surveillance cameras included in different surveillance camera images are different; determining whether the surveillance camera to be updated is a newly added surveillance camera based on the feature point matching relationship, the marker information in the image to be matched, and the marker information of the surveillance camera images in the preset surveillance camera image library. This method is used to accurately determine newly added surveillance cameras.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and more specifically, to an image processing method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] Electronic camera information is a type of low-level geographic data with multiple applications. Traditional methods for collecting electronic camera information involve acquiring road images, analyzing the electronic cameras within those images, and determining whether each camera is newly added (i.e., whether it is the same as existing cameras). Only if it is a newly added camera will it be used.

[0003] When determining whether a traffic camera is a newly added camera, the traditional method is to perform a difference analysis on the position and orientation information (image acquisition position and orientation) between the newly detected camera image and the existing camera images to determine whether the newly detected camera is a newly added camera.

[0004] The technical problems with traditional technologies are as follows: when the difference is made based on location and direction information, a newly detected electronic eye may be mistaken for the same electronic eye if it is close to an existing electronic eye; and when the GPS (Global Positioning System) positioning deviation is too large, it will lead to inaccurate difference of location information, which will also lead to misjudgment of newly added electronic eyes.

[0005] Therefore, traditional technologies cannot accurately detect newly added electronic eyes. Summary of the Invention

[0006] The purpose of this application is to provide an image processing method, apparatus, electronic device, storage medium, and program product to achieve accurate judgment of newly added electronic eyes.

[0007] This application discloses TS1, an image processing method, comprising: acquiring an image to be matched; the image to be matched includes a speed camera to be updated and marker information for marking the position of the speed camera to be updated in the image to be matched; determining a feature point matching relationship between the image to be matched and speed camera images in a preset speed camera image library; the preset speed camera image library includes multiple speed camera images, each speed camera image including a speed camera and marker information for marking the position of the speed camera in the corresponding speed camera image, and different speed camera images include different speed cameras; determining whether the speed camera to be updated is a newly added speed camera based on the feature point matching relationship, the marker information in the image to be matched, and the marker information of the speed camera images in the preset speed camera image library.

[0008] In the above implementation process, the feature point matching relationship represents whether the feature points of the images match. When the feature points match, the objects at the corresponding locations of those feature points can be considered the same. The matched images all include marker information used to mark the locations of the electronic eyes in the images. Therefore, based on the feature point matching relationship and the marker information in the images, it is possible to determine whether the electronic eyes are the same, and thus whether the electronic eye is a newly added one.

[0009] Compared to existing technologies, the above implementation process no longer uses the image acquisition location and direction information to determine the new electronic eyes. Instead, it uses image-to-image matching to determine the new electronic eyes. Image-to-image matching does not use the image acquisition location and direction information, which can avoid the influence of GPS errors on the judgment results. Furthermore, image-to-image matching does not require distance differentiation, which can avoid misjudgment when the distance is relatively close. Therefore, this method can achieve accurate judgment of new electronic eyes.

[0010] TS2, the method as described in TS1, wherein the image to be matched corresponds to location information for characterizing the image acquisition location, each electronic eye image corresponds to location information for characterizing the image acquisition location, and determining the feature point matching relationship between the image to be matched and electronic eye images in a preset electronic eye image library includes: determining a target electronic eye image based on the location information of the image to be matched and the location information of electronic eye images in the preset electronic eye image library; determining the feature point matching relationship between the image to be matched and the target electronic eye image within a first preset distance range based on the distance between the image acquisition location of the image to be matched and the image acquisition location of the target electronic eye image.

[0011] In the above implementation process, the electronic eye images used for matching the image to be matched are selected electronic eye images, and the distance between the image acquisition position of the selected electronic eye images and the image acquisition position of the image to be matched is within a first preset distance range. In this way, the images to be matched can be selected, reducing the computational load and thus improving the efficiency of determining feature point matching relationships.

[0012] TS3, the method as described in TS1, wherein determining the feature point matching relationship between the image to be matched and electronic eye images in a preset electronic eye image library includes: determining feature points in the image to be matched, and determining feature points in electronic eye images in the preset electronic eye image library; determining global information of feature points in the image to be matched, and determining global information of feature points in electronic eye images in the preset electronic eye image library; the global information is used to characterize the relationship between feature points; based on the global information of feature points in the image to be matched and the global information of feature points in electronic eye images in the preset electronic eye image library, matching the feature points in the image to be matched and the feature points in electronic eye images in the preset electronic eye image library to determine the feature point matching relationship.

[0013] In the above implementation process, by determining the global information of feature points in the image, the feature point matching relationship is determined using the global information. Since the global information represents the relationship between feature points, it can reduce the matching of redundant information, improve the efficiency of determining the feature point matching relationship, and improve the accuracy of the finally determined feature point matching relationship.

[0014] TS4. The method as described in TS1, wherein determining whether the electronic eye to be updated is a newly added electronic eye based on the feature point matching relationship, the marker information in the image to be matched, and the marker information of electronic eye images in the preset electronic eye image library includes: determining a first feature point corresponding to the marker information in the image to be matched, and determining a second feature point corresponding to the electronic eye image in the preset electronic eye image library; determining whether the first feature point and the second feature point match based on the feature point matching relationship; wherein, if the first feature point and the second feature point match, it indicates that the electronic eye to be updated is not a newly added electronic eye; if the first feature point and the second feature point do not match, it indicates that the electronic eye to be updated is a newly added electronic eye.

[0015] In the above implementation process, the first feature point and the second feature point can be determined by the marking information of the electronic eye in the image. The first feature point and the second feature point can be understood as the feature points at the corresponding locations of the electronic eye. If the first feature point and the second feature point match, it can be determined that the corresponding electronic eye is the same and is not a newly added electronic eye; if the first feature point and the second feature point do not match, it can be determined that the corresponding electronic eye is different and is a newly added electronic eye. Therefore, through the above implementation process, the accurate identification of newly added electronic eyes can be achieved.

[0016] TS5, the method as described in TS1, wherein acquiring the image to be matched includes: acquiring an image dataset; the image dataset includes: original images collected by multiple vehicles within a preset time period; determining multiple images to be processed from the original images; each image to be processed includes: a speed camera to be processed and marking information for marking the position of the speed camera to be processed in the image to be processed; determining a target image corresponding to each vehicle from the multiple images to be processed corresponding to each vehicle; the speed cameras to be processed included in different target images corresponding to each vehicle are different; determining the image to be matched from the target images of different vehicles; the speed cameras to be updated included in different images to be matched are different.

[0017] In the above implementation process, raw images collected from multiple vehicles within a preset time period are processed to filter out images containing both the electronic eye detection information (EES) and marker information. First, target images are determined based on the EES images for each vehicle. Since the EES images for each vehicle are different, it is ensured that there are no duplicate EES images for the same vehicle; that is, the EES information for the same vehicle is unique. Then, images to be matched are determined based on the target images for different vehicles. Again, since the EES images to be updated are different, it is ensured that there are no duplicate EES images for the same vehicle; that is, the EES information for the same vehicle is unique. This implementation method ensures that the EES information for the images to be matched is unique, reducing redundant EES information.

[0018] TS6, the method as described in TS5, wherein determining the target image corresponding to each vehicle from multiple images to be processed corresponding to each vehicle includes: for any one of the multiple images to be processed corresponding to each vehicle, determining a judgment result for characterizing whether the electronic eye to be processed in the image to be processed is a repeating electronic eye; determining the target image corresponding to each vehicle based on the judgment results of each image to be processed; wherein the process of determining the judgment result includes: determining the feature point matching relationship between the image to be processed and other images to be processed of the vehicle; determining whether the electronic eye to be processed in the image to be processed is a repeating electronic eye based on the feature point matching relationship between the image to be processed and other images to be processed of the vehicle, the marking information in the image to be processed and the marking information in the other images to be processed of the vehicle.

[0019] In the above implementation process, the judgment results of duplicate electronic eye detection for each image to be processed are first determined, and then the judgment results of duplicate electronic eye detection for each image to be processed are combined to filter out the images corresponding to duplicate electronic eye detection. When determining the judgment result of duplicate electronic eye detection, a picture-to-picture matching method is also used. Referring to the description of the technical effect of the aforementioned picture-to-picture matching method, accurate judgment of duplicate electronic eye detection can be achieved.

[0020] TS7, the method as described in TS6, wherein the plurality of images to be processed each correspond to acquisition time information; determining the feature point matching relationship between the image to be processed and other images to be processed of the vehicle includes: determining the target image to be processed of the vehicle based on the acquisition time information corresponding to the plurality of images to be processed; determining the feature point matching relationship between the image to be processed and the target image to be processed of the vehicle by having a preset time order relationship between the acquisition time information corresponding to the target image to be processed and the acquisition time information corresponding to the image to be processed.

[0021] In the above implementation process, there is a temporal order relationship between the acquisition time information of the image matching the image to be processed and the acquisition time information of the image to be processed, so as to realize the tracking of the same electronic eye in continuously acquired images.

[0022] TS8, the method as described in TS5, wherein determining the image to be matched from target images of different vehicles includes: for any target image among the target images of different vehicles, determining a judgment result for characterizing whether the electronic eye to be processed in the target image is a repeating electronic eye; determining the image to be matched based on the judgment result of repeating electronic eyes in each target image; wherein the process of determining the judgment result includes: determining the feature point matching relationship between the target image and the target images of different vehicles; determining whether the electronic eye to be processed in the target image is a repeating electronic eye based on the feature point matching relationship between the target image to be matched and the target images of different vehicles, the marker information in the target image and the marker information in the target images of different vehicles.

[0023] In the above implementation process, the judgment results of duplicate electronic eye detection for each target image are first determined, and then the judgment results of duplicate electronic eye detection for each target image are combined to filter out the target images corresponding to duplicate electronic eye detection. When determining the judgment result of duplicate electronic eye detection, a picture-to-picture matching method is also used. Referring to the description of the technical effect of the aforementioned picture-to-picture matching method, accurate judgment of duplicate electronic eye detection can be achieved.

[0024] TS9, as described in TS8, wherein each target image corresponds to location information used to characterize the image acquisition location; determining the feature point matching relationship between the target image and target images of different vehicles includes: determining a designated target image among the target images of different vehicles based on the location information corresponding to each target image; the distance between the image acquisition location of the target image and the acquisition location of the designated target image is within a second preset distance range; and determining the feature point matching relationship between the target image and the designated target image.

[0025] In the above implementation process, the distance between the image acquisition position of the image matching the target image and the image acquisition position of the target image is within a second preset distance range. This method enables the filtering of images that match the target image, reducing computational load and thus improving the efficiency of determining feature point matching relationships.

[0026] TS10, the method as described in TS1, wherein the image processing method further includes: if the electronic eye to be updated is a newly added electronic eye, updating the preset electronic eye image library according to the image to be matched.

[0027] In the above implementation process, after determining that the electronic eye to be updated is a newly added electronic eye, the preset electronic eye image library can be updated. Under the premise of ensuring that the electronic eyes in the electronic eye image library are unique, the electronic eye information is updated in a timely manner.

[0028] TS11. An image processing apparatus, comprising: an acquisition module and a processing module; the acquisition module is configured to acquire an image to be matched; the image to be matched includes a speed camera to be updated and marker information for marking the position of the speed camera to be updated in the image to be matched; the processing module is configured to: determine a feature point matching relationship between the image to be matched and speed camera images in a preset speed camera image library; the preset speed camera image library includes multiple speed camera images, each speed camera image including a speed camera and marker information for marking the position of the speed camera in the corresponding speed camera image, the speed cameras included in different speed camera images being different; and determine whether the speed camera to be updated is a newly added speed camera based on the feature point matching relationship, the marker information in the image to be matched, and the marker information of the speed camera images in the preset speed camera image library.

[0029] TS12. An electronic device, comprising a processor, a memory, and a communication bus; the communication bus is used to enable communication between the processor and the memory; the processor is used to execute one or more programs stored in the memory to implement any of the above-mentioned image processing methods.

[0030] TS13. A computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement any of the above-described image processing methods.

[0031] TS14. A computer program product comprising a computer program that, when executed by a processor, implements any of the above-described image processing methods. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A flowchart of the image processing method provided in the embodiments of this application;

[0034] Figure 2 Example diagram of the image to be matched provided in the embodiments of this application;

[0035] Figure 3 This is a schematic diagram of the structure of the image processing apparatus provided in the embodiments of this application;

[0036] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0037] Icons: 300 - Image processing device; 310 - Acquisition module; 320 - Processing module; 400 - Electronic device; 410 - Processor; 420 - Memory; 430 - Communication bus. Detailed Implementation

[0038] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0039] The technical solution provided in this application can identify newly added electronic eyes (eyes that are frequently detected by traffic cameras), and based on this identification, can update existing electronic eye information. Therefore, the technical solution provided in this application can be applied to application scenarios involving electronic eye information.

[0040] In application scenarios involving electronic surveillance camera information, the process might simply involve updating that information. For example, a platform holding nationwide electronic surveillance camera information can provide services such as querying and subscribing to this information. This platform needs to maintain the nationwide electronic surveillance camera information in real time. Therefore, when it receives newly detected electronic surveillance camera information, it can use the technical solution provided in this application's embodiments to first determine whether the newly detected information is existing. If it is, the nationwide electronic surveillance camera information is not updated; otherwise, it is updated.

[0041] In application scenarios involving electronic eye information, it may simply be a matter of determining whether any new electronic eyes have been installed. Example 1: After electronic eyes are deployed, the monitoring party needs to check if deployment is complete. In this case, the technical solution provided in this application can be used to determine if any new electronic eyes have been installed on the road. If so, further manual verification is needed to confirm whether the new electronic eye is the required one. If not, it is assumed that the required deployment of electronic eyes has not been completed. Example 2: For electronic eyes on the road, there is usually a maintenance party for the electronic eye information. This maintenance party can use the technical solution provided in this application to monitor in real time whether any new electronic eyes have been installed, thereby determining the frequency and location of new installations, and whether they conform to existing road conditions.

[0042] In application scenarios involving electronic eye information, it may be necessary to update the electronic eye information first, and then apply the updated information. For example, electronic maps have become a common auxiliary means for passenger car drivers to quickly find routes and destinations, and the accuracy and timeliness of the map are very important for user experience. Electronic eyes on the road are important underlying geographic data in the map, and accurate and timely broadcasting of electronic eye information is of great help to users. Therefore, when a new electronic eye is added to the road, it needs to be updated to the map in a timely manner. So, for the backend maintenance party of the electronic map, when new electronic eye information is detected, the technical solution provided in the embodiments of this application can be used to first determine whether the new electronic eye is an existing electronic eye. If it is, it is updated to the map; if not, the map is not updated.

[0043] The above application scenarios are merely examples. In actual applications, the technical solutions provided in the embodiments of this application can be reasonably applied in combination with specific needs. Furthermore, the technical solutions provided in the embodiments of this application can also be applied to other feasible application scenarios.

[0044] Based on the above application scenarios, the hardware operating environment corresponding to the technical solutions provided in this application embodiment includes, but is not limited to: electronic eye database platform, electronic eye information monitoring platform, electronic eye information maintenance platform, electronic map platform, etc.

[0045] Please refer to the following. Figure 1 Here is a flowchart of an image processing method provided in an embodiment of this application. The image processing method includes:

[0046] Step 110: Obtain the image to be matched.

[0047] The image to be matched includes the electronic eye to be updated and marker information used to mark the location of the electronic eye to be updated in the image to be matched.

[0048] The image to be matched can be understood as a road image captured by a vehicle or other image acquisition device, which includes the electronic eyes to be updated. These electronic eyes can be understood as those detected through electronic eye detection on the road image.

[0049] The tagging information is used to mark the locations of the surveillance cameras that need to be updated. For example, please refer to... Figure 2 This is an example image of the image to be matched provided in an embodiment of this application. The image includes a road area, and above the road area are electronic eye A and electronic eye B. The positions of electronic eye A and electronic eye B in the image to be matched are marked by marker boxes, which can be understood as marking information.

[0050] Here Figure 2 This is just one example; in practical applications, tagging information can take other forms. And, Figure 2 The marker box in the image is square. In practical applications, the marker box can also be other shapes, such as square, circle, or other shapes that match the shape of the electronic eye, so as to more accurately mark the position of the electronic eye in the image.

[0051] As mentioned above, the image to be matched is a road image captured by a vehicle or other image acquisition device; other image acquisition devices, such as a PTZ camera set in a fixed position, can capture road images within a preset range.

[0052] It's understandable that vehicles, as mobile devices, might move from one road to another over a period of time, capturing road images over a relatively large area. Other non-mobile image acquisition devices, however, should only capture road images within a fixed area over a given period.

[0053] Therefore, in practical applications, the decision to acquire the image to be matched from mobile imaging devices such as vehicles or from other non-mobile image acquisition devices can be made based on the detection or update requirements of the electronic eye information. For example, if the detection or update requirements of the electronic eye information are urgent, or cover a large road area (such as the entire country), the image to be matched can be acquired from mobile imaging devices such as vehicles. If the detection or update requirements of the electronic eye information are not particularly urgent, or cover a smaller road area (such as a fixed area), the image to be matched can be acquired from non-mobile image acquisition devices.

[0054] It is understood that the aforementioned mobile image acquisition device is not limited to vehicles, but can also be other feasible devices.

[0055] However, regardless of whether the image acquisition device is mobile or stationary, the image processing method used for the road images acquired by the device is the same. Therefore, the following section will use a vehicle as an example to illustrate the process of acquiring the image to be matched.

[0056] As an optional implementation, step 110 includes: acquiring an image dataset; the image dataset includes: original images collected by multiple vehicles within a preset time period; determining multiple images to be processed from the original images; each image to be processed includes: a speed camera to be processed and marking information for marking the position of the speed camera to be processed in the image to be processed; determining target images corresponding to each vehicle from the multiple images to be processed corresponding to each vehicle; the speed cameras to be processed included in different target images corresponding to each vehicle are different; determining matching images from the target images of different vehicles; the speed cameras to be updated included in different matching images are different.

[0057] Vehicles typically have onboard image sensors (e.g., the camera in a dashcam) that capture images of the road. These image sensors are also configured with an image capture frequency, for example, capturing road images every 5 minutes.

[0058] In addition, the vehicle-mounted image sensor also includes GPS information when acquiring road images. Therefore, the original images mentioned above also correspond to GPS information.

[0059] Furthermore, when the vehicle-mounted image sensor acquires road images, it also saves the acquisition time of the image. Therefore, the original image mentioned above also has acquisition time information.

[0060] The road image acquisition method based on vehicles involves the vehicle acquiring road images and then uploading them to an image collection platform. This image collection platform can be either the execution end of the image processing method or a vehicle management platform. If it is the execution end of the image processing method, the road images can be obtained directly from the vehicle. If it is a vehicle management platform, the execution end of the image processing method can request road images from the vehicle management platform.

[0061] The road range corresponding to the image dataset can be the national road network, or it can be the road network of a specific region, such as the road network of a city, or the road network of a specific highway. In practical applications, different implementation methods can be used depending on the specific needs.

[0062] Furthermore, in practical applications, each original image should have corresponding identification information. This identification information is used to mark the image acquisition party corresponding to each original image, so that original images acquired by different vehicles can be distinguished during image processing.

[0063] The preset time period can be understood as the collection cycle of road images. This preset time period can be set based on the detection or update needs of traffic camera information, as well as the road area corresponding to the road images. For example, if the detection or update needs of traffic cameras are urgent, or the road area corresponding to the road images is large, the preset time period can be set shorter, such as 1 day, corresponding to acquiring the raw images collected by vehicles each day. If the detection or update needs of traffic cameras are not urgent, or the road area corresponding to the road images is small, the preset time period can be set longer, such as 3 days, corresponding to acquiring the raw images collected by vehicles within those 3 days every 3 days.

[0064] It's understandable that vehicles travel on different roads every day, some of which may have traffic cameras (electronic eyes) and some may not. Therefore, when vehicles collect road images, they may or may not capture images of traffic cameras. Consequently, among the raw images collected by multiple vehicles within a preset time period, some raw images may include images of traffic cameras, while others may not.

[0065] Furthermore, after obtaining the image dataset, electronic eye detection can be performed on each original image to filter out images containing electronic eyes; and the locations of electronic eyes in the images containing electronic eyes can be marked to obtain the images to be processed.

[0066] As an optional implementation, multiple images to be processed are determined from the original image, including: inputting the original image into a pre-trained electronic eye detection model to obtain the images to be processed output by the electronic eye detection model.

[0067] In this implementation, a pre-trained electronic eye detection model is provided. This model can not only detect whether an electronic eye is present in the original image, but also directly mark the location of the detected electronic eye. In other words, the image output by the model is the image to be processed.

[0068] The pre-trained electronic eye detection model can be an artificial intelligence model such as a neural network model or a random forest model, and there is no limitation here.

[0069] The pre-trained electronic eye detection model can be trained using a training dataset. This training dataset includes multiple sample images containing electronic eyes, in which the electronic eyes are labeled using the labeling information described in the foregoing embodiments.

[0070] During training, the training dataset is input into the initial electronic eye detection model for training, resulting in a trained electronic eye detection model.

[0071] In some embodiments, the training dataset can be divided into two parts: one part as training samples and the other part as test samples, with the number of training samples being greater than the number of test samples. First, the initial electronic eye detection model is trained using the training samples. Then, the accuracy of the trained electronic eye detection model is tested using the test samples. Finally, the accuracy test results are used to optimize the trained electronic eye detection model to obtain the final trained electronic eye detection model.

[0072] In some embodiments, other methods for improving model accuracy can also be used to improve model accuracy, such as: preset training times, training methods based on adversarial examples, etc., which are not limited here.

[0073] As another optional implementation, the above-mentioned electronic eye detection model may include two models: one model is the detection model of the electronic eye, and the other model is the labeling model of the electronic eye location.

[0074] In this implementation, the first model is used to classify the original image into images that include and do not include electronic eyes; then, the second model is used to label the locations of electronic eyes in the images that include electronic eyes, thus obtaining the image to be processed.

[0075] Correspondingly, the training methods for the detection model and the annotation model can refer to the implementation method of the training process of the aforementioned electronic eye detection model. The only difference is that the method of setting the training samples will not be elaborated here.

[0076] In practical applications, the two optional implementation methods described above can be selected based on actual needs. For example, if the number of original images is particularly large, the second implementation method can be used; if the number of original images is within a normal range, the first implementation method can be used. This ensures the accuracy of the final image to be processed.

[0077] It is understandable that the number of electronic eyes to be processed in the image to be processed is unlimited; it can be one or more.

[0078] In practical applications, based on their purpose, electronic eyes can be divided into two types: civilian electronic eyes and police electronic eyes. In some embodiments, only information from civilian electronic eyes may be needed, while in other embodiments, only information from police electronic eyes may be needed; of course, information from both types of electronic eyes may also be required.

[0079] If the information requirement of the electronic eye is that both types of electronic eye information are required, then after obtaining the image to be processed, subsequent processing can be performed directly based on the image to be processed.

[0080] If the requirement for electronic eye information is only for one type of electronic eye, then after obtaining the image to be processed, the type of electronic eye in the image to be processed is first identified, the image to be processed is classified according to the identified type of electronic eye, and then the image to be processed corresponding to the required electronic eye information is used as the data basis for subsequent image processing.

[0081] As an optional implementation, classifying the image to be processed includes: inputting the image to be processed into a pre-trained image classification model to obtain the image to be processed corresponding to the specified electronic eye type output by the image classification model.

[0082] In this implementation, a pre-trained image classification model can be trained using a training dataset. This training dataset includes multiple sample images of electronic eyes (speed cameras), in which the type of electronic eye is labeled using type labeling information. For example, the type labeling information is used to label police-use electronic eyes using a first labeling method, and non-police-use electronic eyes using a second labeling method.

[0083] When marking the type of electronic eye, it can be marked by the surrounding environment of the location of the electronic eye. For example, police electronic eyes are generally surrounded by specific warning signs; while civilian electronic eyes are not surrounded by specific warning signs, and their placement is more arbitrary.

[0084] During training, the training dataset is input into the initial image classification model for training, resulting in a well-trained image classification model.

[0085] In some embodiments, the training dataset can be divided into two parts: one part as training samples and the other part as test samples, with the number of training samples being greater than the number of test samples. First, the initial image classification model is trained using the training samples. Then, the accuracy of the trained image classification model is tested using the test samples. Finally, the accuracy test results are used to optimize the trained image classification model to obtain the final trained image classification model.

[0086] In some embodiments, other methods for improving model accuracy can also be used to improve model accuracy, such as: preset training times, training methods based on adversarial examples, etc., which are not limited here.

[0087] In some embodiments, the image classification model described above can directly output the image to be processed, including the required type of electronic eye. In this implementation, the label settings in the training dataset corresponding to the image classification model should directly characterize the required type of electronic eye.

[0088] In some embodiments, the image classification model described above may only output images to be processed labeled with the type of electronic eye, and then filter out images containing the required type of electronic eye based on the label information. In this implementation, the labels in the training dataset corresponding to the image classification model only need to represent the type of electronic eye.

[0089] In practical applications, the two optional implementation methods described above can be selected based on actual needs. For example, if the number of images to be processed is particularly large, the first implementation method can be used. If the number of images to be processed is within a normal range, the second implementation method can be used. In this way, the accuracy of the final determined images to be processed, including those of the specified electronic eye type, can be guaranteed.

[0090] It is understood that the above implementation method takes police and non-police electronic eyes as examples. In actual application, the images to be processed, including other types of electronic eyes, can be screened according to the needs of specific application scenarios and the above implementation method.

[0091] Regardless of whether the image to be processed has undergone further processing, for the final image to be processed, the target image corresponding to each vehicle is determined from the multiple images to be processed corresponding to each vehicle.

[0092] The process of determining the target image can be understood as identifying the unique identifier of the electronic eye corresponding to the image captured by the electronic eye for each vehicle. For example, for vehicle one, there are three images to be processed. The electronic eye included in images one and two is the same, while the electronic eye included in image three is a different electronic eye. After this processing step, the target image corresponding to vehicle one should include either image one or image two, as well as image three. Furthermore, the electronic eyes included in the different target images corresponding to each vehicle are different, i.e., they have unique identifiers.

[0093] Furthermore, since the electronic eyes in the same image are located in different positions, it is unlikely that they belong to the same electronic eye. Therefore, it is assumed that the electronic eyes included in the same target image are not the same.

[0094] It is understandable that although the electronic eyes to be processed are different in different target images corresponding to the same vehicle, the final image to be matched corresponds to multiple vehicles. That is, after integrating the target images corresponding to multiple vehicles, there is a possibility that different target images contain the same electronic eyes.

[0095] Therefore, after determining the target image corresponding to each vehicle, the image to be matched is determined based on the target images of multiple vehicles, so that the electronic eyes to be updated are different in different images to be matched.

[0096] For example, suppose the target images corresponding to vehicle A include target image A and target image B, and the target images corresponding to vehicle B include target image C and target image D. Since both vehicle A and vehicle B have traveled the same road segment, the speed camera included in target image A and the speed camera included in target image B is the same speed camera. Therefore, based on the target images corresponding to vehicle A and vehicle B, the final images to be matched include: target image A or target image C, and target images B and D.

[0097] Furthermore, in the examples above, if the two images contain the same electronic eye, the final image to be retained should be the one that includes more electronic eye information, and that this more information is different from the electronic eyes in the other images. Of course, if both images contain only the same electronic eye, then either image can be arbitrarily retained as the image to be matched.

[0098] In the above implementation process, raw images collected from multiple vehicles within a preset time period are processed to filter out images containing both the electronic eye detection information (EES) and marker information. First, target images are determined based on the EES images for each vehicle. Since the EES images for each vehicle are different, it is ensured that there are no duplicate EES images for the same vehicle; that is, the EES information for the same vehicle is unique. Then, images to be matched are determined based on the target images from multiple vehicles. Again, since the EES images to be matched are different, it is ensured that there are no duplicate EES images for different vehicles; that is, the EES information for different vehicles is unique. This implementation method ensures that the EES information for the images to be matched is unique, reducing redundant EES information.

[0099] As an optional implementation, the target image corresponding to each vehicle is determined from multiple images to be processed corresponding to each vehicle, including: for any one of the multiple images to be processed corresponding to each vehicle, determining a judgment result used to characterize whether the electronic eye to be processed in the image to be processed is a repeating electronic eye; and determining the target image corresponding to each vehicle based on the judgment result of each image to be processed.

[0100] As an optional implementation, the determination process of the judgment result includes: determining the feature point matching relationship between the image to be processed and other images to be processed of the vehicle; and determining whether the electronic eye to be processed in the image to be processed is a duplicate electronic eye based on the feature point matching relationship between the image to be processed and other images to be processed of the vehicle, the marking information in the image to be processed and the marking information in the other images to be processed of the vehicle.

[0101] In this implementation, the image to be processed can be matched with other images of the vehicle to determine the feature point matching relationship.

[0102] In practical applications, in order to improve the efficiency of feature point matching, the image to be processed can also be matched with the specific image to be processed for that vehicle.

[0103] As mentioned in the above embodiments, the original images captured by the vehicle correspond to capture time information; therefore, the images to be processed also correspond to capture time information. In this embodiment, the capture time information can be used to determine the features of the images to be processed.

[0104] As an optional implementation, determining the feature point matching relationship between the image to be processed and other images of the vehicle to be processed includes: determining the target image to be processed of the vehicle based on the acquisition time information corresponding to multiple images to be processed; and determining the feature point matching relationship between the image to be processed and the target image to be processed of the vehicle by having a preset time order relationship between the acquisition time information corresponding to the target image to be processed and the acquisition time information corresponding to the image to be processed.

[0105] In practical applications, multiple images to be processed can be sorted in chronological order based on the acquisition time information of the vehicle, that is, the image acquired earlier is in front, and the image acquired later is in the back.

[0106] Correspondingly, based on the sorted multiple images to be processed, feature matching is performed on adjacent images. For example, if the sorted multiple images to be processed are: image A, image B, and image C; then, feature matching is first performed between image A and image B to determine the result of repeated electronic surveillance. Then, feature matching is performed between image C and image B to determine the result of repeated electronic surveillance.

[0107] In the above implementation process, there is a temporal order relationship between the acquisition time information of the image matching the image to be processed and the acquisition time information of the image to be processed, so as to realize the tracking of the same electronic eye in continuously acquired images.

[0108] Regardless of which implementation method is used, when determining the feature point matching relationship between the image to be processed of the vehicle and other images to be processed, the traditional feature point matching relationship can be used.

[0109] Compared to traditional feature point matching methods, in order to improve the accuracy of feature point matching relationships and the efficiency of the matching process, as an optional implementation method, the feature point matching relationship between the image to be processed and other images of the vehicle to be processed is determined. This includes: determining feature points in the image to be processed and determining feature points in the images that are matched with the image to be processed; then determining global information of feature points in the image to be processed and determining global information of feature points in the images that are matched with the image to be processed; finally, using the global information to match feature points in the image to be processed and the images that are matched with the image to be processed to determine the feature point matching relationship.

[0110] In the above implementation process, by determining the global information of feature points in the image, the feature point matching relationship is determined using the global information. Since the global information represents the relationship between feature points, it can reduce the matching of redundant information, improve the efficiency of determining the feature point matching relationship, and improve the accuracy of the finally determined feature point matching relationship.

[0111] Based on feature point matching relationships and the labeling information in the image to be processed and the image being matched, assuming the image to be processed is image A, which includes electronic eye A and its labeling information, and the image being matched is image B, which includes electronic eye B and its labeling information, determining whether the electronic eye to be processed is a duplicate involves: identifying the feature points at the location of the labeling information of electronic eye A in image A, and identifying the feature points at the location of the labeling information of electronic eye B in image B; then, searching the feature matching relationships to see if the feature points at the location of the labeling information of electronic eye A match the feature points at the location of the labeling information of electronic eye B. If they match, then electronic eye A is a duplicate. Of course, electronic eye B is also a duplicate, but this determination is based on the result for duplicate electronic eyes in image A. Since the results of duplicate electronic eye determinations from multiple images will be combined for filtering later, only one duplicate electronic eye will be retained, thus not affecting the final filtering result.

[0112] The above implementation only describes the case where the image to be processed contains only one electronic eye. When the image contains multiple electronic eyes, the determination of duplicate electronic eyes can also be achieved. For example, based on the above process description, assuming image A also contains electronic eye C and image B also contains electronic eye D, it is necessary to determine the feature points at the locations of electronic eyes C and D respectively. When determining duplicate electronic eyes, for electronic eye A, it is necessary to determine whether electronic eye A is a duplicate of electronic eyes B and D, and similarly, for electronic eye C, it is necessary to determine whether electronic eye C is a duplicate of electronic eyes B and D.

[0113] After obtaining the judgment results for each vehicle's corresponding image to be processed, the judgment results can be integrated to determine the target image for each vehicle. Various implementation methods can be used for integration. In general, the integration principle is: based on the images to be processed corresponding to the duplicate electronic eyes, only one image is retained as the target image, which serves as the unique identifier for that duplicate electronic eye.

[0114] The specific integration scenarios have been illustrated in the preceding embodiments and will not be repeated here.

[0115] In the above implementation process, the judgment results of duplicate electronic eye detection for each image to be processed are first determined, and then the judgment results of duplicate electronic eye detection for each image to be processed are combined to filter out the images corresponding to duplicate electronic eye detection. When determining the judgment results of duplicate electronic eye detection, an image-to-image matching method is used to determine them, which can achieve accurate judgment of duplicate electronic eye detection.

[0116] The above implementation method has clarified how to determine the target image from multiple images to be processed corresponding to each vehicle. Next, we will discuss how to determine the image to be matched from target images corresponding to multiple vehicles.

[0117] As an optional implementation, for any target image among target images of different vehicles, a judgment result is determined to characterize whether the electronic eye to be processed in the target image is a duplicate electronic eye; the image to be matched is determined based on the judgment result of duplicate electronic eyes in each target image; wherein, the process of determining the judgment result includes: determining the feature point matching relationship between the target image and the target images of different vehicles; determining whether the electronic eye to be processed in the target image is a duplicate electronic eye based on the feature point matching relationship between the target image to be matched and the target images of different vehicles, the marking information in the target image and the marking information in the target images of different vehicles.

[0118] In this implementation, the method for determining the image to be matched is based on the same principle as the method for determining the target image, except that the image to be processed is changed to the target image, and the matching image corresponding to the image to be processed is changed to another target image. Furthermore, the aforementioned determination process is only for road images collected by one vehicle, while the determination process here is for road images collected by different vehicles.

[0119] Therefore, for the sake of brevity, the implementation methods of some steps in this embodiment can be referred to the description in the foregoing embodiments. The process of determining the feature point matching relationship and the method of judging repeated electronic eyes will not be described again here.

[0120] In this implementation, the target image of the current vehicle can be feature-matched with the target images of all other vehicles.

[0121] However, in practical applications, the road areas corresponding to the road images collected by some vehicles may not share a common region. For example, the road area may be very large; or the roads may be far apart, such as one in the south of the city and another in the north. In such cases, the traffic cameras captured in these two locations are unlikely to be duplicates. Therefore, to improve the efficiency of determining the images to be matched, the target images for matching the current vehicle's target image can be filtered.

[0122] As mentioned in the foregoing embodiments, the road images collected by the vehicle also correspond to GPS information, which can be understood as location information used to characterize the image acquisition location. Based on this location information, as an optional implementation, determining the feature point matching relationship between the target image and target images of different vehicles includes: determining a designated target image among the target images of different vehicles according to the location information corresponding to each target image; the distance between the image acquisition location of the target image and the acquisition location of the designated target image is within a second preset distance range; and determining the feature point matching relationship between the target image and the designated target image.

[0123] The second preset distance range can be understood as the distance range that may be repeated by electronic eyes, such as 500m. In practical applications, this preset distance range can be set according to different road conditions.

[0124] When determining a specific target image, the distance between the image acquisition locations is first calculated based on the location information of the current target image and the location information of target images of different vehicles. Then, it is determined whether the distance value is within a second preset distance range. If it is, the corresponding target image is the specified target image; otherwise, the corresponding target image is not the specified target image.

[0125] In the above implementation process, the distance between the image acquisition position of the image matching the target image and the image acquisition position of the target image is within a second preset distance range. This method enables the filtering of images that match the target image, reducing computational load and thus improving the efficiency of determining feature point matching relationships.

[0126] Furthermore, in practical applications, by reasonably presetting the second preset distance range, it is possible to track the same electronic eye in images within a relatively large distance range.

[0127] Through the description of the implementation method of step 110 above, it can be understood that each electronic eye to be updated included in different matching images has a unique identifier, that is, each electronic eye to be updated corresponding to multiple matching images is unique; and each matching image includes marking information for marking the position of the electronic eye to be updated.

[0128] At this point, step 120 can be executed: determine the feature point matching relationship between the image to be matched and the electronic eye images in the preset electronic eye image library.

[0129] The preset electronic eye image library includes multiple electronic eye images. Each electronic eye image includes an electronic eye and marking information used to mark the position of the electronic eye in the corresponding electronic eye image. Different electronic eye images include different electronic eyes.

[0130] It is understood that in the preset electronic eye image library, all electronic eye images correspond to multiple unique electronic eyes, meaning that each electronic eye in the preset electronic eye image library also has a unique identifier. Furthermore, when setting up this preset electronic eye image library, some initial electronic eye images can be preset first, and then each subsequent update is performed according to the technical solution provided in the embodiments of this application, thereby ensuring that all electronic eyes in the image library have a unique identifier.

[0131] In addition, each electronic eye image also corresponds to GPS information, which can be used to characterize the image acquisition location of the electronic eye image.

[0132] Furthermore, in step 110, the number of images to be matched may be one or more. If there are multiple images, each image can be processed separately according to the processing method in step 120.

[0133] In step 120, the image to be matched can be matched with each electronic eye image in a preset electronic eye image library to determine the respective feature matching relationships. However, in practical applications, in order to improve the efficiency of feature point matching, the image to be matched can be matched only with specific electronic eye images.

[0134] It is understandable that the electronic eye images in the preset electronic eye image library correspond to different road ranges. The road range corresponding to the current image to be matched may only be related to other road ranges. For those road ranges that are not related, feature point matching is not very meaningful.

[0135] Therefore, combining the location information corresponding to the image to be matched and the location information corresponding to each electronic eye image, as an optional implementation method, step 120 includes: determining the target electronic eye image based on the location information of the image to be matched and the location information of electronic eye images in a preset electronic eye image library; determining the feature point matching relationship between the image to be matched and the target electronic eye image within a first preset distance range based on the distance between the image acquisition position of the image to be matched and the image acquisition position of the target electronic eye image.

[0136] In this implementation, the first preset distance range can be set in combination with the required distance range. For example, if you want to determine whether the same electronic eye is included in the image data within a distance range of one kilometer, then the first preset distance range is one kilometer.

[0137] When determining the target electronic eye image, the distance between the image acquisition positions of the two images is first calculated based on the position information of the image to be matched and the position information of the electronic eye image. Then, it is determined whether the calculated distance is within the first preset distance range. If it is, the corresponding electronic eye image is the target electronic eye image; otherwise, the corresponding electronic eye image is not the target electronic eye image.

[0138] In the above implementation process, the electronic eye images used for matching with the image to be matched are selected electronic eye images, and the distance between the image acquisition position of the selected electronic eye images and the image acquisition position of the image to be matched is within a first preset distance range. This approach reduces the computational load during feature point matching and improves the efficiency of determining feature point matching relationships.

[0139] Furthermore, in practical applications, by reasonably presetting the first preset distance range, it is possible to match image data within a larger distance range, thereby determining whether there are any new electronic eyes within the larger distance range.

[0140] Whether feature matching is performed based on a target electronic eye image or on all electronic eye images, as an optional implementation method, the feature matching process includes: determining feature points in the image to be matched, and determining feature points in electronic eye images (target electronic eye image or all electronic eye images) in a preset electronic eye image library; determining global information of feature points in the image to be matched, and determining global information of feature points in electronic eye images in the preset electronic eye image library; the global information is used to characterize the relationship between feature points; based on the global information of feature points in the image to be matched and the global information of feature points in electronic eye images in the preset electronic eye image library, matching the feature points in the image to be matched with the feature points in electronic eye images in the preset electronic eye image library to determine the feature point matching relationship.

[0141] In the foregoing embodiments, the implementation method of feature point matching in this embodiment has been described. Therefore, for the sake of brevity, the implementation method of this part refers to the foregoing embodiments and will not be described again here.

[0142] In the above implementation process, by determining the global information of feature points in the image, the feature point matching relationship is determined using the global information. Since the global information represents the relationship between feature points, it can reduce the matching of redundant information, improve the efficiency of determining the feature point matching relationship, and improve the accuracy of the finally determined feature point matching relationship.

[0143] After determining the feature point matching relationship in step 120, step 130 is executed: based on the feature point matching relationship, the marking information in the image to be matched, and the marking information of the electronic eye images in the preset electronic eye image library, it is determined whether the electronic eye to be updated is a newly added electronic eye.

[0144] As an optional implementation, step 130 includes: determining a first feature point corresponding to the marker information in the image to be matched, and determining a second feature point corresponding to an electronic eye image in a preset electronic eye image library; determining whether the first feature point and the second feature point match according to the feature point matching relationship; wherein, if the first feature point and the second feature point match, it indicates that the electronic eye to be updated is not a newly added electronic eye; if the first feature point and the second feature point do not match, it indicates that the electronic eye to be updated is a newly added electronic eye.

[0145] In the foregoing embodiments, determining whether a new electronic eye is a new one is equivalent to determining whether a duplicate electronic eye is a duplicate one in the foregoing embodiments. Therefore, for the sake of brevity, the implementation method of this part refers to the description of the determination process of duplicate electronic eyes in the foregoing embodiments, and will not be repeated here.

[0146] In the above implementation process, the first feature point and the second feature point can be determined by the marking information of the electronic eye in the image. The first feature point and the second feature point can be understood as the feature points at the corresponding locations of the electronic eye. If the first feature point and the second feature point match, it can be determined that the corresponding electronic eye is the same and is not a newly added electronic eye; if the first feature point and the second feature point do not match, it can be determined that the corresponding electronic eye is different and is a newly added electronic eye. Therefore, through the above implementation process, the accurate identification of newly added electronic eyes can be achieved.

[0147] After step 130 is completed, it is possible to determine whether the electronic eye to be updated is a newly added electronic eye. As can be seen from the above descriptions of various implementation methods, feature point matching represents whether feature points between images match. When feature points match, the objects at the corresponding locations of those feature points can be considered identical. Matched images all include marker information used to mark the location of the electronic eye in the image. Therefore, based on the feature point matching relationship and the marker information in the image, it is possible to determine whether the electronic eyes are identical, and thus whether the electronic eye is a newly added one.

[0148] Compared to existing technologies, the above implementation process no longer uses the image acquisition location and direction information to determine the new electronic eyes. Instead, it uses image-to-image matching to determine the new electronic eyes. Image-to-image matching does not use the image acquisition location and direction information, which can avoid the influence of GPS errors on the judgment results. Furthermore, image-to-image matching does not require distance differentiation, which can avoid misjudgment when the distance is relatively close. Therefore, this method can achieve accurate judgment of new electronic eyes.

[0149] After step 130 is completed, the result of whether the electronic eye to be updated is a newly added electronic eye can be obtained. Combining the description of application scenarios in the previous embodiment, different applications can be applied to the newly added electronic eye based on the result of the judgment.

[0150] As a first optional application method, the image processing method also includes: if the electronic eye to be updated is a newly added electronic eye, updating the preset electronic eye image library according to the image to be matched.

[0151] In the pre-set electronic eye image library, each electronic eye image can be classified according to the corresponding image acquisition device and the corresponding road range.

[0152] Therefore, in this application method, the image to be matched can be stored under the category of the corresponding image acquisition device (e.g., the corresponding vehicle identification) or under the category of the corresponding road range.

[0153] Of course, each electronic eye image can also be unclassified and only have a corresponding image identifier. In this case, during the update, the image identifier of the image to be matched is determined based on the existing image identifier, and then the image with the set identifier is updated to the preset electronic eye image library.

[0154] In practical applications, other methods for updating the electronic eye database can be used based on the image to be matched and the electronic eyes to be updated included therein, which are not limited here.

[0155] In the above implementation process, after determining that the electronic eye to be updated is a newly added electronic eye, the preset electronic eye image library can be updated. Under the premise of ensuring that the electronic eyes in the electronic eye image library are unique, the electronic eye information is updated in a timely manner.

[0156] As a second application method, the location of the newly added electronic eyes is determined and the new electronic eyes are updated in the map.

[0157] In this application method, the actual location of the newly added electronic eye can be determined by combining the location of the new electronic eye in the image to be matched (which can be determined by the identification information) and the location information of the image to be matched. Then, according to the actual location of the new electronic eye, the identification of the new electronic eye is added to the corresponding location on the map to achieve map updating.

[0158] Of course, in practical applications, other map update methods can also be used based on newly added electronic eyes and images to be matched, which are not limited here.

[0159] The two application methods mentioned above are only examples. In practical applications, the newly added electronic eyes can be used in more ways in different application scenarios, which will not be listed here.

[0160] Based on the same inventive concept, this application also provides an image processing apparatus 300. Please refer to [link / reference]. Figure 3 As shown, Figure 3 It shows the use of Figure 1 The image processing apparatus of the method shown. It should be understood that the specific functions of the image processing apparatus 300 can be found in the description above; to avoid repetition, detailed descriptions are appropriately omitted here. The image processing apparatus 300 includes at least one software function module that can be stored in memory or embedded in the operating system of the image processing apparatus 300 in the form of software or firmware. Specifically:

[0161] See Figure 3 As shown, the image processing apparatus 300 includes: an acquisition module 310 and a processing module 320. Wherein:

[0162] The acquisition module 310 is used to acquire the image to be matched. The image to be matched includes the electronic eye to be updated and marking information used to mark the position of the electronic eye to be updated in the image to be matched.

[0163] The processing module 320 is configured to: determine the feature point matching relationship between the image to be matched and the electronic eye images in a preset electronic eye image library; the preset electronic eye image library includes multiple electronic eye images, each of which includes an electronic eye and marking information for marking the position of the electronic eye in the corresponding electronic eye image, and different electronic eye images include different electronic eyes; and determine whether the electronic eye to be updated is a newly added electronic eye based on the feature point matching relationship, the marking information in the image to be matched, and the marking information of the electronic eye images in the preset electronic eye image library.

[0164] In this embodiment, the image to be matched corresponds to location information characterizing the image acquisition location, and each electronic eye image corresponds to location information characterizing the image acquisition location. Correspondingly, the processing module 320 is specifically configured to: determine a target electronic eye image based on the location information of the image to be matched and the location information of electronic eye images in a preset electronic eye image library; and determine the feature point matching relationship between the image to be matched and the target electronic eye image within a first preset distance range, based on the distance between the image acquisition location of the image to be matched and the image acquisition location of the target electronic eye image.

[0165] In this embodiment of the application, the processing module 320 is further configured to: determine feature points in the image to be matched, and determine feature points in electronic eye images in the preset electronic eye image library; determine global information of feature points in the image to be matched, and determine global information of feature points in electronic eye images in the preset electronic eye image library; the global information is used to characterize the relationship between feature points; based on the global information of feature points in the image to be matched and the global information of feature points in electronic eye images in the preset electronic eye image library, match the feature points in the image to be matched and the feature points in electronic eye images in the preset electronic eye image library to determine the feature point matching relationship.

[0166] In this embodiment of the application, the processing module 320 is further configured to: determine a first feature point corresponding to the marker information in the image to be matched, and determine a second feature point corresponding to the electronic eye image in the preset electronic eye image library; determine whether the first feature point and the second feature point match according to the feature point matching relationship; wherein, if the first feature point and the second feature point match, it indicates that the electronic eye to be updated is not a newly added electronic eye; if the first feature point and the second feature point do not match, it indicates that the electronic eye to be updated is a newly added electronic eye.

[0167] In this embodiment, the acquisition module 310 is further configured to: acquire an image dataset; the image dataset includes: original images collected by multiple vehicles within a preset time period; determine multiple images to be processed from the original images; each image to be processed includes: a speed camera to be processed and marking information for marking the position of the speed camera to be processed in the image to be processed. The processing module 320 is further configured to: determine target images corresponding to each vehicle from the multiple images to be processed corresponding to each vehicle; the speed cameras to be processed included in different target images corresponding to each vehicle are different; determine the matching images from the target images of different vehicles; the speed cameras to be updated included in different matching images are different.

[0168] In this embodiment of the application, the processing module 320 is further configured to: for any one of the multiple images to be processed corresponding to each vehicle, determine a judgment result that characterizes whether the electronic eye to be processed in the image to be processed is a repeating electronic eye; determine the target image corresponding to each vehicle based on the judgment results of each image to be processed; wherein, the process of the processing module 320 determining the judgment result includes: determining the feature point matching relationship between the image to be processed and other images to be processed of the vehicle; determining whether the electronic eye to be processed in the image to be processed is a repeating electronic eye based on the feature point matching relationship between the image to be processed and other images to be processed of the vehicle, the marking information in the image to be processed and the marking information in the other images to be processed of the vehicle.

[0169] In this embodiment of the application, the plurality of images to be processed each correspond to acquisition time information; the processing module 320 is specifically used to: determine the target image to be processed of the vehicle based on the acquisition time information corresponding to the plurality of images to be processed; and determine the feature point matching relationship between the image to be processed and the target image to be processed of the vehicle by having a preset time order relationship between the acquisition time information corresponding to the target image to be processed and the acquisition time information corresponding to the image to be processed.

[0170] In this embodiment of the application, the processing module 320 is further configured to: determine, for any target image among target images of different vehicles, a judgment result characterizing whether the electronic eye to be processed in the target image is a duplicate electronic eye; and determine the image to be matched based on the judgment results of duplicate electronic eyes in each target image. The process by which the processing module 320 determines the judgment result includes: determining the feature point matching relationship between the target image and target images of different vehicles; and determining whether the electronic eye to be processed in the target image is a duplicate electronic eye based on the feature point matching relationship between the target image and target images of different vehicles, the marker information in the target image, and the marker information in the target images of different vehicles.

[0171] In this embodiment of the application, each target image corresponds to location information used to characterize the image acquisition location; the processing module 320 is further configured to: determine a designated target image among the target images of different vehicles based on the location information corresponding to each target image; the distance between the image acquisition location of the target image and the acquisition location of the designated target image is within a second preset distance range; and determine the feature point matching relationship between the target image and the designated target image.

[0172] In this embodiment of the application, the processing module 320 is further configured to: if the electronic eye to be updated is a newly added electronic eye, update the preset electronic eye image library according to the image to be matched.

[0173] It should be understood that, for the sake of brevity, the content described in the method embodiments will not be repeated in this embodiment.

[0174] Please refer to Figure 4 This application also provides an electronic device 400, which can serve as the hardware operating environment for the aforementioned image processing method. It includes a processor 410, a memory 420, and a communication bus 430. Wherein:

[0175] The communication bus 430 is used to realize the connection and communication between the processor 410 and the memory 420.

[0176] The processor 410 is used to execute one or more programs stored in the memory 420 to implement the image processing method described in the above embodiments.

[0177] Understandable. Figure 4 The structure shown is for illustrative purposes only; the electronic device may also include components that are more advanced than those shown. Figure 4 The more or fewer components shown, or having the same Figure 4 The different configurations shown.

[0178] It should also be noted that the electronic device 400 provided in this embodiment can be implemented as an electronic device with data processing function, such as a server or host.

[0179] This embodiment also provides a computer-readable storage medium, such as a floppy disk, optical disk, hard disk, flash memory, USB flash drive, SD (Secure Digital Memory Card), MMC (Multimedia Card), etc., in which one or more programs implementing the above steps are stored. These one or more programs can be executed by one or more processors to implement the image processing method described in the above embodiments. Further details will not be elaborated here.

[0180] This embodiment also provides a computer program product, including a computer program that, when executed by a processor, implements the image processing method described in the above embodiments. Further details will not be elaborated upon here.

[0181] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0182] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0183] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0184] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0185] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An image processing method, characterized in that, include: Obtain the image to be matched; the image to be matched includes the electronic eye to be updated and marking information for marking the position of the electronic eye to be updated in the image to be matched; The feature point matching relationship between the image to be matched and the electronic eye images in the preset electronic eye image library is determined; the preset electronic eye image library includes multiple electronic eye images, each of which includes an electronic eye and marking information for marking the position of the electronic eye in the corresponding electronic eye image, and different electronic eye images include different electronic eyes; wherein, the feature point matching relationship is determined by graph-to-graph matching. Based on the feature point matching relationship, the marking information in the image to be matched, and the marking information of the electronic eye images in the preset electronic eye image library, it is determined whether the electronic eye to be updated is a newly added electronic eye; The step of determining whether the electronic eye to be updated is a newly added electronic eye based on the feature point matching relationship, the marker information in the image to be matched, and the marker information of electronic eye images in the preset electronic eye image library includes: Determine the first feature point corresponding to the marker information in the image to be matched, and determine the second feature point corresponding to the electronic eye image in the preset electronic eye image library; Determine whether the first feature point and the second feature point match based on the feature point matching relationship; If the first feature point and the second feature point match, it indicates that the electronic eye to be updated is not a newly added electronic eye. If the first feature point and the second feature point do not match, it indicates that the electronic eye to be updated is a newly added electronic eye; The step of determining the feature point matching relationship between the image to be matched and electronic eye images in a preset electronic eye image database includes: The feature points in the image to be matched are determined, and the feature points in the electronic eye images in the preset electronic eye image library are also determined; Global information of feature points in the image to be matched is determined, as well as global information of feature points in electronic eye images in the preset electronic eye image library; the global information is used to characterize the relationship between feature points. Based on the global information of feature points in the image to be matched and the global information of feature points in the electronic eye images in the preset electronic eye image library, the feature points in the image to be matched and the feature points in the electronic eye images in the preset electronic eye image library are matched to determine the feature point matching relationship.

2. The image processing method according to claim 1, characterized in that, The image to be matched corresponds to location information representing the image acquisition location, and each electronic eye image corresponds to location information representing the image acquisition location. Determining the feature point matching relationship between the image to be matched and electronic eye images in a preset electronic eye image library includes: The target electronic eye image is determined based on the location information of the image to be matched and the location information of electronic eye images in a preset electronic eye image library; the distance between the image acquisition location of the image to be matched and the image acquisition location of the target electronic eye image is within a first preset distance range; Determine the feature point matching relationship between the image to be matched and the target electronic eye image.

3. The image processing method according to claim 1, characterized in that, The process of obtaining the image to be matched includes: Obtain an image dataset; the image dataset includes: original images collected by multiple vehicles within a preset time period; Multiple images to be processed are determined from the original image; each image to be processed includes: an electronic eye to be processed and marking information for marking the position of the electronic eye to be processed in the image to be processed; The target image corresponding to each vehicle is determined from multiple images to be processed for each vehicle; the different target images corresponding to each vehicle contain different electronic eyes to be processed; The image to be matched is determined from target images of different vehicles; different images to be matched include different electronic eyes to be updated.

4. The image processing method according to claim 3, characterized in that, The step of determining the target image corresponding to each vehicle from multiple images to be processed for each vehicle includes: For any one of the multiple images to be processed corresponding to each vehicle, determine the judgment result used to characterize whether the electronic eye to be processed in the image is a duplicate electronic eye; The target image corresponding to each vehicle is determined based on the judgment results of each image to be processed; The process of determining the judgment result includes: Determine the feature point matching relationship between the image to be processed and other images of the vehicle to be processed; Based on the feature point matching relationship between the image to be processed and other images to be processed of the vehicle, the marking information in the image to be processed and the marking information in other images to be processed of the vehicle, it is determined whether the electronic eye to be processed in the image to be processed is a duplicate electronic eye.

5. The image processing method according to claim 4, characterized in that, The multiple images to be processed each correspond to the acquisition time information; Determining the feature point matching relationship between the image to be processed and other images of the vehicle to be processed includes: The target image to be processed for the vehicle is determined based on the acquisition time information corresponding to the multiple images to be processed; there is a preset time order relationship between the acquisition time information corresponding to the target image to be processed and the acquisition time information corresponding to the image to be processed. Determine the feature point matching relationship between the image to be processed and the target image to be processed of the vehicle.

6. The image processing method according to claim 3, characterized in that, Determining the image to be matched from target images of different vehicles includes: For any target image among target images of different vehicles, determine the judgment result used to characterize whether the electronic eye to be processed in the target image is a repeating electronic eye; The image to be matched is determined based on the judgment results of the repeated electronic eyes of each target image; The process of determining the judgment result includes: Determine the feature point matching relationship between the target image and target images of different vehicles; Based on the feature point matching relationship between the target image and target images of different vehicles, the marking information in the target image and the marking information in target images of different vehicles, it is determined whether the electronic eye to be processed in the target image is a duplicate electronic eye.

7. The image processing method according to claim 6, characterized in that, Each target image corresponds to location information used to characterize the image acquisition location; Determining the feature point matching relationship between the target image and target images of different vehicles includes: The designated target image in the target images of different vehicles is determined based on the location information corresponding to each target image; the distance between the image acquisition location of the target image and the acquisition location of the designated target image is within a second preset distance range; Determine the feature point matching relationship between the target image and the specified target image.

8. The image processing method according to claim 1, characterized in that, The image processing method further includes: If the electronic eye to be updated is a newly added electronic eye, the preset electronic eye image library is updated according to the image to be matched.

9. An image processing apparatus, characterized in that, include: Acquisition module and processing module; The acquisition module is used to acquire an image to be matched; the image to be matched includes an electronic eye to be updated and marking information for marking the position of the electronic eye to be updated in the image to be matched; The processing module is used for: The feature point matching relationship between the image to be matched and the electronic eye images in the preset electronic eye image library is determined; wherein, the feature point matching relationship is determined by graph-to-graph matching; the preset electronic eye image library includes multiple electronic eye images, each of which includes an electronic eye and marking information for marking the position of the electronic eye in the corresponding electronic eye image, and different electronic eye images include different electronic eyes; Based on the feature point matching relationship, the marking information in the image to be matched, and the marking information of the electronic eye images in the preset electronic eye image library, it is determined whether the electronic eye to be updated is a newly added electronic eye; The step of determining whether the electronic eye to be updated is a newly added electronic eye based on the feature point matching relationship, the marker information in the image to be matched, and the marker information of electronic eye images in the preset electronic eye image library includes: Determine the first feature point corresponding to the marker information in the image to be matched, and determine the second feature point corresponding to the electronic eye image in the preset electronic eye image library; Determine whether the first feature point and the second feature point match based on the feature point matching relationship; If the first feature point and the second feature point match, it indicates that the electronic eye to be updated is not a newly added electronic eye. If the first feature point and the second feature point do not match, it indicates that the electronic eye to be updated is a newly added electronic eye; The step of determining the feature point matching relationship between the image to be matched and electronic eye images in a preset electronic eye image database includes: The feature points in the image to be matched are determined, and the feature points in the electronic eye images in the preset electronic eye image library are also determined; Global information of feature points in the image to be matched is determined, as well as global information of feature points in electronic eye images in the preset electronic eye image library; the global information is used to characterize the relationship between feature points. Based on the global information of feature points in the image to be matched and the global information of feature points in the electronic eye images in the preset electronic eye image library, the feature points in the image to be matched and the feature points in the electronic eye images in the preset electronic eye image library are matched to determine the feature point matching relationship.

10. An electronic device, characterized in that, It includes a processor, a memory, and a communication bus; the communication bus is used to enable communication between the processor and the memory; the processor is used to execute one or more programs stored in the memory to implement the image processing method as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the image processing method as described in any one of claims 1-8.

12. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the image processing method as described in any one of claims 1-8.

Images