Image processing method and apparatus, electronic device and storage medium

Abstract

Provided are an image processing method and apparatus, an electronic device and a storage medium. The image processing method comprises: acquiring one or more input images; performing first target detection on the one or more input images, so as to detect one or more potential threat targets in a region of interest in the one or more input images; in response to the detection of the one or more potential threat targets, cropping the one or more input images to obtain one or more potential threat region images corresponding to the one or more potential threat targets; performing second target detection on the one or more potential threat region images, so as to determine one or more sensitive-information regions in the one or more potential threat region images; and blurring the one or more sensitive-information regions in the one or more potential threat region images.

Description

Image processing methods, apparatuses, electronic devices and storage media

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese patent application No. 202510374237.4, filed on March 26, 2025, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to the field of image processing technology, and in particular to an image processing method, an image processing apparatus, an electronic device, and a computer-readable storage medium. Background Technology

[0004] The Sentinel Mode on two-wheeled vehicles aims to enhance vehicle safety, boost driver confidence, and promote the intelligent development of two-wheeled vehicles. Once activated, Sentinel Mode continuously monitors the vehicle and its surroundings using the vehicle's built-in camera. When abnormal behavior or potential threats are detected, Sentinel Mode automatically records the surrounding environment and sends an alert to the driver via a mobile app or SMS. Drivers can control the vehicle via their mobile phones to activate functions such as the horn and flashing lights, enabling them to respond promptly to potential risks.

[0005] The methods described in this section are not necessarily methods that had been previously conceived or adopted. Unless otherwise specified, no method described in this section should be assumed to be prior art simply because it is included in this section. Similarly, unless otherwise specified, the issues mentioned in this section should not be considered to be accepted in any prior art. Summary of the Invention

[0006] This disclosure provides an image processing method, electronic device, and storage medium to desensitize sensitive information in an image before uploading it to the cloud.

[0007] According to one aspect of this disclosure, an image processing method is provided, comprising: acquiring one or more input images; performing a first target detection on the one or more input images to detect one or more potential threat targets in regions of interest within the one or more input images; cropping the one or more input images in response to detecting one or more potential threat targets to obtain one or more potential threat region images corresponding to the one or more potential threat targets; performing a second target detection on the one or more potential threat region images to determine one or more sensitive information regions within the one or more potential threat region images; and blurring the one or more sensitive information regions within the one or more potential threat region images.

[0008] According to another aspect of this disclosure, an image processing apparatus is provided, comprising: an acquisition unit configured to acquire one or more input images; a potential threat region determination unit configured to perform a first target detection on the one or more input images to detect one or more potential threat targets in regions of interest within the one or more input images; an image cropping unit configured to crop the one or more input images in response to detecting one or more potential threat targets to obtain one or more potential threat region images corresponding to the one or more potential threat targets; a sensitive information region determination unit configured to perform a second target detection on the one or more potential threat region images to determine one or more sensitive information regions within the one or more potential threat region images; and a desensitization unit configured to perform a blurring operation on the one or more sensitive information regions within the one or more potential threat region images.

[0009] According to another aspect of this disclosure, an electronic circuit is provided, comprising: a circuit configured to perform the steps of the above-described method.

[0010] According to another aspect of this disclosure, an electronic device is provided. The electronic device includes: a processor; and a memory storing a program including instructions that, when executed by the processor, cause the processor to perform the methods described above.

[0011] According to another aspect of this disclosure, a non-transitory computer-readable storage medium storing a program is provided. The program includes instructions that, when executed by a processor of an electronic device, cause the electronic device to perform the methods described above.

[0012] According to another aspect of this disclosure, a computer program product is provided. This computer program product includes a computer program that, when executed by a processor, implements the above-described method.

[0013] According to another aspect of this disclosure, a two-wheeled vehicle is provided. The two-wheeled vehicle includes the aforementioned electronic equipment.

[0014] According to embodiments of this disclosure, sensitive information can be desensitized through secondary target detection using pure vision. This not only reduces the use of complex sensors and lowers costs, but also improves the generalization ability of the target detection model, increases the target detection accuracy, saves computing resources, and reduces device power consumption, thereby enabling large-scale promotion on two-wheeled vehicles.

[0015] These and other aspects of this disclosure will be apparent from the embodiments described below, and will be elucidated with reference to the embodiments described below. Attached Figure Description

[0016] The accompanying drawings exemplify embodiments and form part of the specification, serving together with the textual description to explain exemplary implementations of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0017] Figure 1 shows a flowchart of an exemplary process of an image processing method according to an embodiment of the present disclosure;

[0018] Figure 2 shows an exemplary block diagram of an image processing apparatus according to an embodiment of the present disclosure; and

[0019] Figure 3 is a block diagram illustrating an example of an electronic device according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0020] In this disclosure, unless otherwise stated, the use of terms such as "first," "second," etc., to describe various elements is not intended to limit the positional, temporal, or importance relationships of these elements; such terms are merely used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of that element, while in other cases, based on the context, they may refer to different instances.

[0021] The terminology used in the description of the various examples described in this disclosure is for the purpose of describing particular examples only and is not intended to be limiting. Unless the context explicitly indicates otherwise, an element may be one or more unless the number of elements is specifically limited. Furthermore, the term "and / or" as used in this disclosure covers any one of the listed items and all possible combinations thereof.

[0022] In related technologies, sentinel mode systems require some hardware devices based on complex sensors, and the detection methods are costly, making it difficult to promote and apply them on a large scale on two-wheeled vehicles. Some detection methods lack intelligent processing capabilities and cannot accurately detect abnormal behavior or potential threats, leading to misjudgments and false alarms by the sentinel system. Existing detection methods may perform differently in different regions and environments, have poor generalization ability, and cannot adapt to diverse application scenarios. Existing perception algorithms are usually highly complex and require a lot of computing resources, which not only increases the energy consumption of the device but may also affect the normal operation of other functions.

[0023] To address the aforementioned problems in related technologies, this disclosure provides a novel image processing method that desensitizes sensitive information through secondary target detection using pure vision. This not only reduces the use of complex sensors and lowers costs but also improves the generalization ability of the target detection model, enhances target detection accuracy, saves computational resources, and reduces device power consumption, thereby enabling large-scale deployment on two-wheeled vehicles. Embodiments of this disclosure are described in detail below with reference to the accompanying drawings.

[0024] Figure 1 shows a flowchart of an exemplary process of an image processing method according to an embodiment of the present disclosure.

[0025] In step S102, one or more input images may be acquired.

[0026] In step S104, a first target detection can be performed on one or more input images to detect one or more potential threat targets in the regions of interest of one or more input images.

[0027] In step S106, in response to detecting one or more potential threat targets, one or more input images may be cropped to obtain one or more potential threat area images corresponding to one or more potential threat targets.

[0028] In step S108, a second target detection can be performed on one or more potential threat area images to determine one or more sensitive information areas in the one or more potential threat area images.

[0029] In step S110, one or more sensitive information regions in one or more potential threat area images may be blurred.

[0030] The image processing method provided by the embodiments of this disclosure can desensitize sensitive information through secondary target detection based on pure vision. This not only reduces the use of complex sensors and lowers costs, but also reduces computational load and device power consumption by performing secondary target detection on potential threat areas after image cropping, thereby enabling large-scale promotion on two-wheeled vehicles.

[0031] The steps of method 100 are described in detail below.

[0032] In step S102, one or more input images may be acquired.

[0033] In some embodiments, one or more input images may be acquired via a camera on a two-wheeled vehicle.

[0034] In some embodiments, one or more input images are acquired via a front-view camera, a center-view camera, and a rear-view camera of the two-wheeled vehicle. In some examples, the front-view camera may be located at the front of the two-wheeled vehicle to capture the front of the vehicle; the center-view camera may be located on the dashboard of the two-wheeled vehicle to capture the state of the rider or seat; and the rear-view camera may be located at the rear of the vehicle to capture the rear of the two-wheeled vehicle.

[0035] In step S104, a first target detection is performed on one or more input images to detect one or more potential threat targets in the regions of interest of the one or more input images.

[0036] In some examples, in step S104, a first target detection can be performed on one or more input images, such as images acquired from a front-view camera, a center-view camera, and a rear-view camera. The first target detection targets one or more potential threat targets entering the region of interest (ROI) in the aforementioned images. In a sentry system, the ROI may include an area within a certain distance around a two-wheeled vehicle; people, vehicles, or other objects entering this area that could pose a danger can be identified as potential threat targets by the sentry system. The ROI can be determined from the images captured by the cameras based on a predetermined position and predetermined size, such as a region of predetermined size at the center of the image. In embodiments of this disclosure, a target detection model trained using deep learning methods can be used to perform the target detection task. Other suitable methods can also be used to perform the target detection task without departing from the principles of the embodiments of this disclosure. In some examples, further logical judgments can be made on people, vehicles, or other objects detected within the ROI based on predetermined rules, such as the duration of stay within the ROI and the speed of movement, to determine whether they are identified as potential threat targets.

[0037] In some embodiments, method 100 may further include updating the first object detection algorithm model by using the detection results of the first object detection on one or more input images as new samples for training the algorithm model. The detection results include object bounding boxes (bboxes) and labels. In this way, images acquired during use and labeled with object bounding boxes and labels can be used as training datasets for updating the training.

[0038] Therefore, by relying on a large amount of newly acquired image data, such as images of pedestrians or vehicles and other potential threat targets collected from various angles under various weather conditions such as daytime, nighttime, rainy days, and foggy days, the algorithm model for the first target detection is trained and updated, so that the algorithm model for the first target detection has better generalization ability and improves the target detection accuracy.

[0039] In step S106, in response to detecting one or more potential threat targets, one or more input images may be cropped to obtain one or more potential threat area images corresponding to one or more potential threat targets.

[0040] In some examples, in response to the detection of one or more potential threat targets within a certain range of a region of interest near a two-wheeled vehicle, image cropping algorithms, such as image edge cropping, image thresholding, and deep learning-based image cropping, can be used to crop one or more input images to obtain one or more potential threat region images corresponding to the one or more potential threat targets. The potential threat region image can be an image within the target detection bounding box of the potential threat target.

[0041] Therefore, by cropping one or more input images to obtain smaller images of one or more potential threat regions, the computational load of subsequent secondary target detection can be reduced, thereby reducing computational costs, improving computational efficiency, and reducing device power consumption, so as to realize its promotion on two-wheeled vehicles.

[0042] In some embodiments, method 100 may further include starting local video stream recording in response to detecting one or more potential threat targets.

[0043] This enables automatic image acquisition, and the acquired images can be used to train the target detection model, thereby further improving the generalization ability of the target detection model and increasing the target detection accuracy.

[0044] In step S108, a second target detection can be performed on one or more potential threat area images to identify one or more sensitive information regions within those images. Depending on the specific circumstances, the second target detection and the first target detection may use the same or different model structures.

[0045] In some examples, the potential threat target could be a person or a vehicle, where faces and license plates are sensitive information and require secondary target detection.

[0046] In some examples, the potential threat targets detected from images from the front-view and rear-view cameras of a two-wheeled vehicle can be people or vehicles, with sensitive information being faces and license plates. The potential threat targets detected from images from the center-view camera of a two-wheeled vehicle can be people, with sensitive information being faces.

[0047] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0048] In some embodiments, method 100 may further include using the detection results of a second target detection on one or more images of potential threat regions as new samples for training the second target detection algorithm model, thereby updating and training the algorithm model. The detection results include target bounding boxes and labels. In this way, images acquired during use and labeled with target bounding boxes and labels can be used as the training dataset for update training.

[0049] Therefore, by relying on a large amount of newly collected image data, such as images of faces or license plates from various angles under different weather conditions like daytime, nighttime, rainy days, and foggy days, the algorithm model for second object detection can be trained and updated, enabling the second object detection algorithm model to have better generalization ability and improve the accuracy of object detection.

[0050] In some embodiments, at least one of the algorithm model for the first object detection and the algorithm model for the second object detection can be a neural network-based object detection model, such as the YOLO series algorithms, the R-CNN series algorithms, etc.

[0051] In some embodiments, method 100 may further include determining the ReLU function as the activation function of the model of the neural network.

[0052] In some embodiments, method 100 may further include performing graph optimization operations on the convolutional layers (Conv), batch normalization layers (BN), and activation functions of the object detection model after the object detection model is established.

[0053] In some embodiments, method 100 may further include reparameterizing the object detection model to couple multiple operators of the corresponding convolutional layer, batch normalization layer and activation function of the object detection model.

[0054] In some examples, the object detection model described above is the YOLOv8 algorithm model.

[0055] Here, we will use the YOLOv8 algorithm model as an example for illustration:

[0056] 1) Because the Sigmoid function used in the original YOLOv8 algorithm model involves exponential operations and is relatively complex, the formula for the Sigmoid function is as follows:

[0057] Therefore, the Sigmoid function is replaced with the simpler ReLU function, the formula for which is as follows:

[0058] As can be seen from formula (2), the ReLU function only selects data without performing complex exponential operations, thereby reducing the computational complexity of the model and improving the inference performance of the model on the edge.

[0059] 2) In the YOLOv8 algorithm model, the Conv+BN+ReLU structure accounts for more than 80%. This structure requires memory access three times. The formula for the Conv layer is as follows: y=w*x+b Formula (3)

[0060] Where y is the result of the convolution calculation; w is the weight, which is the parameter learned from a large amount of data; x is the input data, which is the input image; and b is the bias term.

[0061] The BN layer formula is as follows:

[0062] In this context, the result y of the convolution in the Conv layer is the input x of the BN layer, and E(x) is the mean of x. y is the standard deviation of x, and γ and β are adjustment coefficients.

[0063] The ReLU formula is shown in formula (2). By combining formulas (2)-(4), the Conv+BN+ReLU structure diagram can be optimized into a single operator, the formula of which is as follows:

[0064] Where, μ B The mean of x in the corresponding BN layer, This corresponds to the standard deviation of x in the BN layer. Therefore, optimizing the Conv+BN+ReLU structure graph into a single operator requires only one memory access, thus improving model inference performance.

[0065] 3) Quantizing the model's weights and activation values ​​can significantly improve the performance of neural networks while sacrificing some accuracy. Quantization methods can include pre-training quantization (PTQ) and in-training quantization (QAT), and the specific method is not limited here.

[0066] In some examples, PTQ INT8 quantization can be used to map the weight parameters and activation values ​​in the original YOLOv8 algorithm model from FP32 to INT8, enabling direct INT8 computation in the hardware, thereby improving inference speed. The PTQ INT8 quantization formula is as follows:

[0067] Where r represents the original float32 value; Z represents the offset of the float32 value; S represents the scaling factor of float32; Round(·) represents the mathematical function for rounding to the nearest integer; and q represents a quantized integer value.

[0068] Therefore, by optimizing the target detection algorithm model, reducing the model size and computational load, the inference speed on the edge is improved, making it more suitable for promotion on two-wheeled vehicles.

[0069] In step S110, one or more sensitive information regions in one or more potential threat area images may be blurred.

[0070] In some examples, sensitive information areas (faces, license plates, etc.) in images to be uploaded to the cloud need to be blurred, for example, through pixel interference, data encryption, masking, etc., to achieve desensitization.

[0071] In some embodiments, method 100 may further include uploading one or more blurred input images to the cloud after blurring one or more sensitive information regions in one or more potential threat area images to trigger an alarm.

[0072] In some examples, de-identified input images can be uploaded to the cloud and trigger an alert to warn car owners of potential threats in the vicinity.

[0073] According to some embodiments of this disclosure, the image processing method may further include: preprocessing one or more images to convert one or more images into one or more images in a preset format.

[0074] One or more images acquired in step S102 may not be directly applicable to the object detection model as input. Therefore, these images can be converted to a preset format (e.g., converted to YUV format via the Video Input module, or converted to a format suitable for the neural network model) so that they can be applied to the object detection model to achieve fast and accurate detection.

[0075] According to embodiments of the present disclosure, an image processing apparatus is also provided. FIG2 shows an exemplary block diagram of an image processing apparatus according to an embodiment of the present disclosure. The image processing apparatus 200 may include an acquisition unit 210 configured to acquire one or more input images; a potential threat region determination unit 220 configured to perform a first target detection on the one or more input images to detect one or more potential threat targets in regions of interest in the one or more input images; an image cropping unit 230 configured to crop the one or more input images in response to detecting one or more potential threat targets to obtain one or more potential threat region images corresponding to the one or more potential threat targets; a sensitive information region determination unit 240 configured to perform a second target detection on the one or more potential threat region images to determine one or more sensitive information regions in the one or more potential threat region images; and a desensitization unit 250 configured to perform a blurring operation on the one or more sensitive information regions in the one or more potential threat region images.

[0076] Here, the operation of each unit of the image processing device is similar to the operation of steps S102 to S110 described above, and will not be repeated here.

[0077] According to another aspect of this disclosure, an electronic circuit is also provided, including a circuit configured to perform the steps of the above-described method.

[0078] According to another aspect of this disclosure, an electronic device is also provided, comprising: a processor; and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform the methods described above.

[0079] According to another aspect of this disclosure, a non-transitory computer-readable storage medium storing a program is also provided, the program including instructions that, when executed by a processor of an electronic device, cause the electronic device to perform the method described above.

[0080] According to another aspect of this disclosure, a computer program product is also provided, including a computer program that, when executed by a processor, implements the above-described method.

[0081] According to another aspect of this disclosure, a two-wheeled vehicle is also provided, including the aforementioned electronic equipment.

[0082] Referring to Figure 3, electronic device 300 will now be described as an example of a hardware device (electronic device) that can be applied to various aspects of this disclosure. Electronic device 300 can be any machine configured to perform processing and / or calculations, and can be, but is not limited to, a workstation, server, desktop computer, laptop computer, tablet computer, personal digital assistant, robot, smartphone, in-vehicle computer, or any combination thereof. The image processing method 100 described above can be implemented wholly or at least partially by electronic device 300 or similar devices or systems.

[0083] Electronic device 300 may include elements that are connected to or communicate with bus 302 (possibly via one or more interfaces). For example, electronic device 300 may include bus 302, one or more processors 304, one or more input devices 306, and one or more output devices 308. The one or more processors 304 may be any type of processor and may include, but are not limited to, one or more general-purpose processors and / or one or more dedicated processors (e.g., special-purpose chips). Input devices 306 may be any type of device capable of inputting information to electronic device 300 and may include, but are not limited to, a mouse, keyboard, touchscreen, microphone, and / or remote control. Output devices 308 may be any type of device capable of presenting information and may include, but are not limited to, a monitor, speaker, video / audio output terminal, vibrator, and / or printer. Electronic device 300 may also include a non-transitory storage device 310. The non-transitory storage device can be any storage device that is non-transitory and capable of storing data, including but not limited to disk drives, optical storage devices, solid-state storage, floppy disks, flexible disks, hard disks, magnetic tapes or any other magnetic media, optical discs or any other optical media, ROM (read-only memory), RAM (random access memory), cache memory and / or any other memory chip or cartridge, and / or any other medium from which a computer can read data, instructions, and / or code. The non-transitory storage device 310 can be detached from an interface. The non-transitory storage device 310 may have data / programs (including instructions) / code for implementing the methods and steps described above. Electronic device 300 may also include a communication device 312. The communication device 312 can be any type of device or system that enables communication with external devices and / or with a network, and may include, but is not limited to, modems, network interface cards, infrared communication devices, wireless communication devices and / or chipsets, such as Bluetooth™ devices, 802.11 devices, Wi-Fi devices, Wi-Max devices, cellular communication devices, and / or the like.

[0084] Electronic device 300 may also include working memory 314, which may be any type of working memory that can store programs (including instructions) and / or data useful for the operation of processor 304, and may include, but is not limited to, random access memory and / or read-only memory devices.

[0085] The software elements (programs) may reside in the working memory 314, including but not limited to the operating system 316, one or more application programs 318, drivers, and / or other data and code. Instructions for performing the methods and steps described above may be included in one or more application programs 318, and the image processing method 100 described above can be implemented by the processor 304 reading and executing the instructions of one or more application programs 318. More specifically, in the image processing method 100 described above, steps S102-S110 can be implemented, for example, by the processor 304 executing an application program 318 having instructions for steps S102-S110. Furthermore, other steps in the image processing method 100 described above can be implemented, for example, by the processor 304 executing an application program 318 having instructions for performing the corresponding steps. The executable code or source code of the instructions of the software elements (programs) may be stored in a non-transitory computer-readable storage medium (e.g., the storage device 310 described above) and may be stored in the working memory 314 during execution (possibly compiled and / or installed). The executable code or source code of the instructions of the software elements (programs) may also be downloaded from a remote location.

[0086] It should also be understood that various modifications can be made depending on specific requirements. For example, custom hardware can also be used, and / or specific elements can be implemented using hardware, software, firmware, middleware, microcode, hardware description languages, or any combination thereof. For example, some or all of the disclosed methods and apparatus can be implemented by programming hardware (e.g., programmable logic circuits including field-programmable gate arrays (FPGAs) and / or programmable logic arrays (PLAs)) using logic and algorithms according to this disclosure in assembly language or hardware programming languages ​​(such as Verilog, VHDL, C++).

[0087] It should also be understood that the aforementioned methods can be implemented using a server-client model. For example, the client can receive user input data and send it to the server. Alternatively, the client can receive user input data, perform a portion of the processing described in the aforementioned methods, and send the resulting data to the server. The server can receive data from the client, execute the aforementioned methods or a portion thereof, and return the execution result to the client. The client can receive the execution result from the server and, for example, present it to the user via an output device.

[0088] It should also be understood that the components of electronic device 300 can be distributed across a network. For example, some processing can be performed using one processor, while other processing can be performed simultaneously by another processor located far away from that processor. Other components of computing system 300 can also be distributed similarly. Thus, electronic device 300 can be interpreted as a distributed computing system that performs processing in multiple locations.

[0089] The following describes some exemplary aspects of this disclosure.

[0090] Aspect 1. An image processing method, comprising:

[0091] Get one or more input images;

[0092] Perform a first target detection on the one or more input images to detect one or more potential threat targets in the regions of interest of the one or more input images;

[0093] In response to detecting the one or more potential threat targets, the one or more input images are cropped to obtain one or more potential threat area images corresponding to the one or more potential threat targets;

[0094] Perform a second target detection on the one or more potential threat area images to determine one or more sensitive information regions within the one or more potential threat area images; and

[0095] The sensitive information regions in the one or more potential threat area images are blurred.

[0096] Aspect 2. The image processing method according to aspect 1 further includes:

[0097] The detection results of the first object detection on the one or more input images are used as new samples to train the algorithm model of the first object detection, thereby updating and training the algorithm model of the first object detection.

[0098] Aspect 3. The image processing method according to aspect 1 further includes:

[0099] The detection results of the second target detection on the images of the one or more potential threat areas are used as new samples to train the algorithm model of the second target detection, thereby updating and training the algorithm model of the second target detection.

[0100] Aspect 4. The image processing method according to any one of Aspects 1 to 3, wherein the one or more input images are acquired by a camera of a two-wheeled vehicle.

[0101] Aspect 5. The image processing method according to aspect 4, wherein the one or more input images are acquired by a front-view camera, a center-view camera, or a rear-view camera of a two-wheeled vehicle.

[0102] Aspect 6. The image processing method according to any one of Aspects 1 to 3 further includes:

[0103] In response to the detection of one or more potential threat targets, local video stream recording is initiated.

[0104] Aspect 7. The image processing method according to any one of Aspects 1 to 3, wherein at least one of the algorithm model for the first target detection and the algorithm model for the second target detection is a target detection model based on a neural network.

[0105] Aspect 8. The image processing method according to aspect 7, further comprising:

[0106] The ReLU function is determined as the activation function of the neural network model.

[0107] Aspect 9. The image processing method according to aspect 8, further comprising:

[0108] After establishing the target detection model, graph optimization operations are performed on the convolutional layers, batch normalization layers, and activation functions of the target detection model.

[0109] Aspect 10. The image processing method according to aspect 9, further comprising:

[0110] The target detection model is reparameterized to couple multiple operators of the target detection model corresponding to the convolutional layer, the batch normalization layer, and the activation function.

[0111] Aspect 11. The image processing method according to any one of Aspects 1 to 3 further includes:

[0112] After blurring one or more sensitive information areas in one or more potential threat area images, the blurred one or more input images are uploaded to the cloud to trigger an alarm.

[0113] Aspect 12. An image processing apparatus, comprising:

[0114] The acquisition unit is configured to acquire one or more input images;

[0115] A potential threat region determination unit is configured to perform a first target detection on the one or more input images to detect one or more potential threat targets in a region of interest in the one or more input images;

[0116] An image cropping unit is configured to crop the one or more input images in response to detecting the one or more potential threat targets to obtain one or more potential threat region images corresponding to the one or more potential threat targets;

[0117] A sensitive information region determination unit is configured to perform second target detection on the one or more potential threat region images to determine one or more sensitive information regions in the one or more potential threat region images; and

[0118] The desensitization unit is configured to blur one or more sensitive information regions in one or more potential threat region images.

[0119] Aspect 13. An electronic circuit, comprising:

[0120] A circuit configured to perform the steps of the method according to any one of aspects 1 to 11.

[0121] Aspect 14. An electronic device comprising:

[0122] Processor; and

[0123] A memory storing a program, the program comprising instructions that, when executed by the processor, cause the processor to perform the method according to any one of aspects 1 to 11.

[0124] Aspect 15. A non-transitory computer-readable storage medium storing a program, the program comprising instructions that, when executed by a processor of an electronic device, cause the electronic device to perform the method according to any one of Aspects 1 to 11.

[0125] Aspect 16. A computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the method according to any one of Aspects 1 to 11.

[0126] Aspect 17. A two-wheeled vehicle including the electronic equipment described in aspect 14.

[0127] While embodiments or examples of this disclosure have been described with reference to the accompanying drawings, it should be understood that the methods, systems, and devices described above are merely exemplary embodiments or examples, and the scope of the invention is not limited by these embodiments or examples, but only by the granted claims and their equivalents. Various elements in the embodiments or examples may be omitted or replaced by their equivalents. Furthermore, the steps may be performed in a different order than that described in this disclosure. Further, various elements in the embodiments or examples may be combined in various ways. Importantly, as the technology evolves, many elements described herein can be replaced by equivalents that appear after this disclosure.

Claims

1. An image processing method, comprising: Get one or more input images; Perform a first target detection on the one or more input images to detect one or more potential threat targets in the regions of interest of the one or more input images; In response to detecting the one or more potential threat targets, the one or more input images are cropped to obtain one or more potential threat area images corresponding to the one or more potential threat targets; A second target detection is performed on the one or more potential threat area images to determine one or more sensitive information areas in the one or more potential threat area images; as well as The sensitive information regions in the one or more potential threat region images are blurred.

2. The image processing method according to claim 1, further comprising: The detection results of the first object detection on the one or more input images are used as new samples to train the algorithm model of the first object detection, thereby updating and training the algorithm model of the first object detection.

3. The image processing method according to claim 1, further comprising: The detection results of the second target detection on the images of the one or more potential threat areas are used as new samples to train the algorithm model of the second target detection, thereby updating and training the algorithm model of the second target detection.

4. The image processing method according to any one of claims 1 to 3, wherein, The one or more input images are acquired through the camera of the two-wheeled vehicle.

5. The image processing method according to claim 4, wherein, The one or more input images are acquired through the front-view camera, center-view camera, or rear-view camera of the two-wheeled vehicle.

6. The image processing method according to any one of claims 1 to 3, further comprising: In response to the detection of one or more potential threat targets, local video stream recording is initiated.

7. The image processing method according to any one of claims 1 to 3, wherein, At least one of the algorithm models for the first object detection and the second object detection is a neural network-based object detection model.

8. The image processing method according to claim 7, wherein, Also includes: The ReLU function is determined as the activation function of the neural network model.

9. The image processing method according to claim 8, wherein, Also includes: After establishing the target detection model, graph optimization operations are performed on the convolutional layers, batch normalization layers, and activation functions of the target detection model.

10. The image processing method according to claim 9, wherein, Also includes: The target detection model is reparameterized to couple multiple operators of the target detection model corresponding to the convolutional layer, the batch normalization layer, and the activation function.

11. The image processing method according to any one of claims 1 to 3, further comprising: After blurring the one or more sensitive information areas in the one or more potential threat area images, the blurred one or more input images are uploaded to the cloud to trigger an alarm.

12. An image processing apparatus, comprising: The acquisition unit is configured to acquire one or more input images; A potential threat region determination unit is configured to perform a first target detection on the one or more input images to detect one or more potential threat targets in a region of interest in the one or more input images; An image cropping unit is configured to crop the one or more input images in response to detecting the one or more potential threat targets to obtain one or more potential threat region images corresponding to the one or more potential threat targets; A sensitive information region determination unit is configured to perform second target detection on the one or more potential threat region images to determine one or more sensitive information regions in the one or more potential threat region images; as well as The desensitization unit is configured to blur one or more sensitive information regions in one or more potential threat region images.

13. An electronic circuit, comprising: A circuit configured to perform the steps of the method according to any one of claims 1 to 11.

14. An electronic device comprising: processor; as well as A memory storing a program, the program comprising instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 11.

15. A non-transitory computer-readable storage medium storing a program, the program comprising instructions that, when executed by a processor of an electronic device, cause the electronic device to perform the method according to any one of claims 1 to 11.

16. A computer program product comprising a computer program, wherein, When the computer program is executed by a processor, it implements the method according to any one of claims 1 to 11.

17. A two-wheeled vehicle comprising the electronic equipment according to claim 14.

Images