Image processing method, device and image processing system based on NVR

Abstract

The present application discloses an image processing method, device and an image processing system based on NVR, which relate to the field of image processing technology. The image processing method comprises establishing a first association relationship between scene conditions and resolution information, obtaining scene conditions and resolution information of a current image, and adjusting the resolution corresponding to the current image based on the scene conditions and the resolution information in combination with the first association relationship. The image processing device can output a more targeted image according to the scene conditions, without replacing a camera with a set resolution, and performing resolution enhancement processing on the image acquired by the camera to obtain a clearer image that is more suitable for image recognition processing, thereby saving municipal security costs. The image processing system based on NVR can obtain more targeted images of different resolutions using an NVR device with a set resolution, and has a wider range of applications.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and more specifically, to an image processing method, device and NVR-based image processing system. Background Art

[0002] NVR, or network video recorder, is the storage and forwarding part of the network video surveillance system. NVR works with video encoders or network cameras to complete video recording, storage and forwarding functions.

[0003] Since surveillance cameras in many areas were installed earlier, their resolution is low and they cannot provide the higher-resolution images required by today's rapidly developing facial recognition and image recognition algorithms. If the early low-resolution cameras are replaced on a large scale, it will lead to a sharp increase in municipal security costs. Summary of the invention

[0004] In view of the problem that the resolution of images acquired by existing cameras in actual applications is low and cannot meet the needs of recognition algorithms, the first purpose of this application is to propose an image processing method, which can process the set resolution images acquired by existing cameras according to different scene conditions so that they can meet the usage requirements of different scene conditions; based on the above image processing method, the second purpose of this application is to provide an image processing device, which can output more targeted images according to scene conditions; based on the above image processing method, the third purpose of this application is to protect an image processing system based on NVR.

[0005] The specific plan is as follows:

[0006] An image processing method, comprising:

[0007] Establishing a first association relationship between scene conditions and resolution information;

[0008] Get scene conditions and resolution information of the current image;

[0009] Based on the scene condition and the resolution information, and in combination with the first association relationship, the resolution corresponding to the current image is adjusted.

[0010] By adopting the above technical solution, different usage scenarios are divided using the first association relationship, different scene conditions are matched with the most suitable image resolution under the scene conditions, and the image acquired under the current scene conditions is processed using image processing technology to adjust it to a suitable image resolution. Compared with directly processing the acquired image using various image recognition technologies, the image after resolution processing can better meet the requirements of image recognition and other technologies, and is also convenient for subsequent operations on the image.

[0011] Preferably, the acquiring scene conditions includes:

[0012] Setting and storing trigger feature information obtainable via an image and a second association relationship between the trigger feature information and each of the scene conditions; identifying the trigger feature information based on current image information;

[0013] Based on the recognition result and the second association relationship, the scene condition is obtained.

[0014] By adopting the above technical solution, the image is divided into different scene conditions using different trigger identifiers of the image, which provides judgment conditions for dividing the scene conditions and facilitates different processing of the image using the feature information in the image to adapt to different algorithms.

[0015] Preferably, the trigger feature information includes:

[0016] Facial feature information, license plate feature information, and dangerous object feature information in the image information; and

[0017] The target object movement speed information is obtained based on the image information processed by a set algorithm.

[0018] By adopting the above technical solution, the presence of facial feature information in the image indicates that the image will be used for face recognition, the license plate feature information in the image indicates that the image will be used for license plate recognition, and the target object moving speed information indicates that the image will be used to identify objects thrown from high places or whether the vehicle is speeding. The above feature information is used as a judgment condition and the image is subjected to different image processing to facilitate matching with different recognition algorithms.

[0019] Preferably, the identifying trigger feature information based on the current image information includes:

[0020] Acquire facial feature information, license plate feature information, and dangerous object feature information based on single-frame image recognition;

[0021] The target object's moving speed information is obtained based on adjacent frame images combined with the set algorithm.

[0022] By adopting the above technical solution, facial contour features, license plate contour and color features, and dangerous object contour features are captured through the features in a single frame image. The moving speed of the target object is obtained by comparing the pixels of adjacent frame images and combining the algorithm.

[0023] Preferably, the scene acquisition conditions include:

[0024] Setting and storing automatic trigger information obtainable through image analysis or device input and a third association relationship between the information and each of the scene conditions;

[0025] Automatically trigger information based on the current image recognition;

[0026] Based on the recognition result and the third association relationship, the scene condition is obtained.

[0027] By adopting the above technical solution, the automatic trigger information attached to or input into the image is used as the basis for dividing the scene conditions. When the automatic trigger information is met, the image resolution is automatically processed in a targeted manner, and the output resolution is more suitable for its application scenario. Setting the third association relationship increases the basis for dividing the scene conditions, making the output image more targeted.

[0028] Preferably, the automatic trigger information also includes time information, geographical location information, environmental information associated with the current image, and memory size information of the image storage location.

[0029] By adopting the above technical solution, when the image time information, geographic location information, and environmental information meet the conditions, such as when it is at night, located in a remote road section, or when the environmental visibility is reduced, the image also needs to be enhanced to facilitate matching of the recognition algorithm.

[0030] Preferably, the automatic triggering information based on the current image recognition includes obtaining the memory size information of the image storage location based on the memory detection algorithm, and the network bandwidth corresponding to the image transmission.

[0031] By adopting the above technical solution, the detection algorithm is used to obtain the memory size information of the image storage location and the network bandwidth corresponding to the image transmission.

[0032] Preferably, a target reference object is selected in the image, and the application scene is divided into a static scene and a dynamic scene according to whether the target reference object moves;

[0033] Establishing basic image resolution information, first resolution information corresponding to static scenes, and second resolution information corresponding to dynamic scenes;

[0034] Obtaining memory size information of the image storage location, if the memory size satisfies the storage of the set duration image based on the basic image resolution information, storing the acquired image at the basic image resolution;

[0035] If the application scenario is a dynamic scene, the resolution corresponding to the current image is adjusted to the second resolution, and the resolution of the static scene image in the subsequent duration is adjusted to the first resolution, so that the memory size meets the requirement of storing images of the set duration.

[0036] By adopting the above technical solution, static scene images are stored at the basic image resolution, dynamic scenes are stored at the second resolution, and in combination with the memory, the static scenes after the dynamic scenes are stored at the first resolution, so that the memory required for the image information meets the actual memory, and the memory required for the entire image information is always adapted to the actual memory of the hardware, avoiding the situation where the hardware device is unable to store the entire image information due to the enhanced dynamic scene resolution.

[0037] Preferably, adjusting the resolution corresponding to the current image information to the second resolution includes:

[0038] Divide the dynamic scene image into key frames and feature frames based on scene conditions;

[0039] If the current image is a key frame, storing the acquired image information at the second resolution;

[0040] Based on the memory size information of the image storage location and the required memory for storing the current image of the set duration, the existing feature frame is stored at the basic resolution or the first resolution.

[0041] By adopting the above technical solution and combining it with the characteristic information in the trigger mark, only the key frames are enhanced in resolution. For example, when the image has the most frontal faces, the most dynamic changes, or other set references appear, the other image frames with identification information remain unchanged or are stored with reduced resolution. They can be enhanced in other ways later. As a result, the content of the video storage will be more accurate, and the desired key image information can be quickly obtained when it is retrieved for review. It is also more conducive to the application of monitoring scenes. Compared with storing dynamic images in the entire segment, it takes up less memory and is convenient for review.

[0042] An image processing device, used to implement the above-mentioned image processing method, comprising:

[0043] A storage module for storing a first association relationship between scene conditions and resolution information;

[0044] An information acquisition module for acquiring scene conditions and current image resolution information; and

[0045] A resolution adjustment module for adjusting the resolution corresponding to the current image according to the scene condition and resolution information combined with the first association relationship; wherein the information acquisition module acquires the scene condition and resolution information of the current image, and the resolution adjustment module adjusts the resolution of the current image based on the first association relationship stored in the storage module.

[0046] By adopting the above technical solution, the first association relationship stored in the storage module is called to determine the scene condition of the current image, the information acquisition module matches the resolution information according to the scene condition of the current image to obtain the most suitable resolution information for the current image, and the resolution adjustment module adjusts the current image to the most suitable resolution to complete the processing of the image.

[0047] Preferably, the information acquisition module includes an image feature capture module for acquiring trigger feature information and an image information analysis module connected to external device data, and acquires trigger feature information and automatic trigger information in real time;

[0048] The storage module also stores a second association relationship between the trigger feature information and the scene condition and a third association relationship between the automatic trigger information and the scene condition.

[0049] By adopting the above technical solution, the image feature capture module captures the features of the image to obtain the trigger feature information on the image, and the external device transmits the image's environmental information, time, location and other related information to the image information analysis module. The image information analysis module analyzes the relevant information to obtain automatic trigger information, and by calling the association relationship in the storage module, the relationship between the trigger feature information, automatic trigger information and the scene conditions is used to determine the current scene conditions.

[0050] Preferably, an NVR-based image processing system comprises an NVR device with a set resolution and an image output device; the NVR device comprises or is connected to the image processing device as described above;

[0051] The image output terminal of the NVR device is data-connected to the data input terminal of the information acquisition module, and the image input terminal of the image output device is data-connected to the image output terminal of the resolution adjustment module.

[0052] By adopting the above technical solution, the image processing device performs targeted processing on the images acquired by the NVR device, so that the images that can be output by the NVR device, which can only acquire a fixed resolution, can be images with multiple resolutions suitable for different scene conditions, balancing the required memory and the accuracy of subsequent image recognition, thereby enhancing the scope of application of the NVR-based image processing system.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] (1) By using image processing technology to process the image acquired under the current scene conditions, the image is adjusted to a suitable image resolution. Compared with directly processing the acquired image using various image recognition technologies, the image after resolution processing can better meet the requirements of image recognition and other technologies, and is also convenient for subsequent image processing;

[0055] (2) Through the trigger feature information on the image and the automatic trigger information obtained by analyzing the image or input, the scene conditions applicable to the image are divided in detail for various scene conditions;

[0056] (3) By dividing the image into dynamic scenes, static scenes, key frames and feature frames, it is ensured that the image can be stored completely when the memory of the image storage location is limited. At the same time, the key information is enhanced to facilitate the later retrieval and review of the key information. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flowchart of the application;

[0058] Figure 2 A flowchart for obtaining scene conditions for this application;

[0059] Figure 3 Another flow chart for obtaining scene conditions for this application;

[0060] Figure 4 is a schematic diagram of an image processing device of the present application.

[0061] Figure numerals: 1. Storage module; 2. Information acquisition module; 3. Resolution adjustment module. DETAILED DESCRIPTION

[0062] The present application is further described in detail below with reference to embodiments and figures, but the implementation methods of the present application are not limited thereto.

[0063] Embodiment 1

[0064] An image processing method such as Figure 1 As shown, including:

[0065] S100: Establish a first association relationship between scene conditions and resolution information.

[0066] S200: Obtain scene conditions and resolution information of the current image.

[0067] S300: Based on the scene condition and the resolution information, and in combination with the first association relationship, adjust the resolution corresponding to the current image.

[0068] Through the above steps, there is no need to replace the camera with a set resolution. When objects such as cigarette butts and garbage thrown from high altitude are identified, the scene condition is a high-altitude throwing scene. The image acquired by the camera is resolution enhanced, and the current image is enhanced to a 1080p high-definition image to obtain a clearer image that is more suitable for high-altitude throwing scene recognition, which is convenient for identifying the target object and where it was thrown from. There is no need to replace the high-resolution camera, saving municipal security costs.

[0069] Embodiment 2

[0070] An image processing method, comprising:

[0071] S100: Establish a first association relationship between scene conditions and resolution information.

[0072] Specifically, different image information has different usage scenarios. Different scene conditions are divided according to the scenarios in which the images are used and are used to classify the images. In the implementation manner of the present application, the scene conditions include face recognition scenarios, vehicle recognition scenarios and dangerous scenarios. According to the existing face recognition database, the most suitable resolution for images used in combination with face recognition algorithms is 1080p. Images with the above resolution have the highest face recognition accuracy after considering the size of the memory occupied. Similarly, when the image resolution for vehicle recognition scene conditions is 640p, the memory occupied and the recognition accuracy can be balanced. The most suitable resolution for dangerous scenes is 720p.

[0073] S200: Obtain scene conditions and resolution information of the current image.

[0074] Specifically, how to determine the scene condition applicable to the current image through the above first association relationship? Figure 2-3 shown.

[0075] S210, setting and storing the trigger feature information obtainable via the image and the second association relationship between the trigger feature information and each of the scene conditions.

[0076] Specifically, the trigger feature information includes facial feature information, license plate feature information, and dangerous object feature information obtained based on single-frame image recognition. Facial feature information is the face in the image, which is realized by a face recognition algorithm. The difference from the algorithm for face matching is that it only needs to capture and determine whether there is a face in the image, and there is no need to match the extracted facial features with the facial features in the database. Similarly, license plate feature information is realized by a license plate recognition algorithm. It only needs to capture the presence of a license plate-like object containing numbers in the image, that is, to determine that there is a license plate in the image, and there is no need to accurately identify the numbers and letters in the license plate. The dangerous object feature information is matched with the dangerous object contour in the database to identify whether there is a dangerous object in the current image. In the implementation of the present application, the above-mentioned dangerous object information is preferably tool contour information and explosive contour information.

[0077] The trigger feature information also includes information on the target object's movement speed obtained based on adjacent frame images combined with a set algorithm. When a moving object is found in the image after comparing adjacent frame images, the object is selected as the target object. The target object's movement speed information can be obtained by combining the number of frames per second of the current image, the number of pixels moved by the target object in each frame image, and the proportional relationship between the object and the actual object in the angle image.

[0078] The scene conditions also include a high-altitude object dropping scene. In the high-altitude object dropping scene, the target object moving speed information is applicable alone. When it is identified that the target object's moving speed in the vertical direction meets the falling speed of high-altitude object dropping, it is determined that the current image is suitable for the high-altitude object dropping scene, and the most suitable resolution for the high-altitude object dropping scene is 640p.

[0079] In another embodiment, the target object movement information can also be combined with the license plate feature information. When the scene condition is a vehicle speeding scene, the target vehicle is locked according to the vehicle's moving speed information, and then the image resolution containing the target vehicle is enhanced to 640p suitable for license plate recognition.

[0080] To facilitate understanding of the second association relationship in this step, the table below lists several commonly used scenario conditions for using trigger feature information alone or in combination.

[0081]

[0082]

[0083] S220: Identify trigger feature information based on the current image information.

[0084] The current single frame image is judged using the single frame image association facial feature capture algorithm, license plate feature capture algorithm, and dangerous object contour feature capture algorithm, and the adjacent frame image association target object movement speed algorithm is used to determine whether there is a moving target in the current image.

[0085] S230: Based on the recognition result and the second association relationship, obtain the scene condition.

[0086] Specifically, in combination with the above table, the trigger feature information is matched with the scene conditions.

[0087] S300: Based on the scene condition and the resolution information, and in combination with the first association relationship, adjust the resolution corresponding to the current image.

[0088] According to the scene conditions of image matching, the resolution of the image is adjusted to make it better match the subsequent recognition algorithm.

[0089] The above steps can classify and match the scene conditions of the image according to the trigger feature information directly or indirectly obtained from the image. In order to further optimize the image processing method and enhance the scope of application of the image processing method, obtaining the scene conditions can also include the following steps:

[0090] S211. Setting and storing automatic trigger information obtainable through image analysis or device input and a third association relationship between the automatic trigger information and each of the scene conditions.

[0091] Specifically, the automatic trigger information can be obtained by parsing the time or geographic location mark of the image acquisition device itself, or an automatic trigger information acquisition device can be connected to the image acquisition device, such as installing an environmental monitoring device on the camera to obtain real-time weather information, visibility information, haze information, etc. near the camera, so as to facilitate the acquisition of environmental conditions with poor visibility such as rainy and snowy weather, heavy fog or severe haze, and scene conditions that require automatically adjusting the current image resolution to a higher level.

[0092] The automatic trigger information includes time information, geographic location information, environmental information and memory size information of the image storage location associated with the current image. When the acquired time information shows that the image is in a night environment, the scene condition of the image belongs to a night recognition scene. When the acquired geographic location information shows that the image is in an accident-prone area, the scene condition of the image belongs to an accident-prone area recognition scene. When the acquired time information shows that the image is in a harsh environment, the scene condition of the image belongs to a harsh environment recognition scene. The above scene conditions all require resolution enhancement processing of the image to facilitate subsequent recognition or other applications.

[0093] When the acquired memory size information of the image storage location shows that the memory of the image storage location is smaller than the memory required for current image storage, the image needs to be processed with reduced resolution to ensure that the image can be stored completely.

[0094] S221, identifying automatic trigger information in the current image.

[0095] The memory size information of the image storage location is obtained through the memory detection algorithm, such as using 0 and 1 to traverse each bit of the image storage location. Similarly, the built-in bandwidth detection tool is used to detect the network bandwidth corresponding to the image transmission to ensure the integrity of image storage and image transmission.

[0096] S231. Based on the recognition result and the third association relationship, obtain the scene condition.

[0097] Combined with the automatic trigger information obtained through image analysis or device input, the current image is divided into different scene conditions, so as to automatically enhance or weaken the image in a targeted manner.

[0098] Since the image storage location and memory are limited, if all images during the night time or all images containing faces are enhanced without discrimination, subsequent key information will not be stored. Therefore, even if the scene conditions that require resolution enhancement are met, it is necessary to further distinguish the scenes in the image, including the following steps.

[0099] S204: Select a target reference object in the image, and divide the application scene into a static scene and a dynamic scene according to whether the target reference object moves.

[0100] By comparing adjacent frame images, we can determine whether the target reference object in the image has moved, and divide the application scenarios into static scenes and dynamic scenes. For example, if the time information shows that it is currently night time, when the memory is insufficient, it does not make much sense to enhance the image resolution of static night recognition scenes, that is, night images without any object movement.

[0101] S205: Establish basic image resolution information, first resolution information corresponding to static scenes, and second resolution information corresponding to dynamic scenes.

[0102] In the implementation manner of the present application, the basic resolution of the image is set to 480p, the first resolution corresponding to the static scene is set to 240p, and the second resolution corresponding to the dynamic scene is set to 720p.

[0103] S206, obtaining memory size information of the image storage location, and if the memory size satisfies the storage of the set duration image based on the basic image resolution information, storing the obtained image at the basic image resolution.

[0104] If the memory size is sufficient to store images of the set duration based on the basic image resolution information, static scene images are stored or transmitted at the basic resolution of 480p without resolution processing.

[0105] S207: If the application scenario is a dynamic scenario, the resolution corresponding to the current image is adjusted to the second resolution, and the resolution of the static scene image in the subsequent duration is adjusted to the first resolution, so that the memory size meets the requirement of storing images of the set duration.

[0106] If after comparing adjacent frames, the target object is selected and the scene is determined to be a dynamic scene, the dynamic scene will be stored or transmitted at the second resolution of 720p. Since the resolution of the dynamic image is improved, the memory occupied by the image segment increases, and the resolution of the static scenes in subsequent images is reduced to achieve a dynamic balance in the memory occupied by the image segment.

[0107] If distinguishing between dynamic scenes and static scenes still cannot satisfy the memory of the storage location, the dynamic scenes are further distinguished according to the following steps.

[0108] Dynamic scene images are divided into key frames and feature frames based on scene conditions. For example, combined with facial feature information, images containing facial movement are divided into key frames with frontal faces and feature frames with only faces. Key frames are more suitable for face matching and recognition, and also strengthen key information when reviewing images later, which facilitates the use of image processing methods in monitoring scenarios.

[0109] If the current image is a key frame, the acquired image information is stored at the second resolution, that is, the key frame is resolution enhanced. Based on the memory size information of the image storage location and the memory required to store the current image for the set duration, if the memory is sufficient, the feature frame will be stored or transmitted at the basic resolution; if the memory is insufficient, the feature frame will be stored or transmitted at the first resolution of 240p.

[0110] An image processing device, such as Figure 4 As shown, the image processing method described above includes a storage module 1 for storing a first association relationship between scene conditions and resolution information, an information acquisition module 2 for acquiring scene conditions and current image resolution information, and a resolution adjustment module 3 for adjusting the resolution corresponding to the current image. The information acquisition module 2 acquires the scene conditions and resolution information of the current image, and the resolution adjustment module 3 adjusts the resolution of the current image based on the first association relationship stored in the storage module 1.

[0111] In detail, the storage module 1 of the image processing device also stores a second association relationship between the trigger feature information and each of the scene conditions and a third association relationship between the automatic trigger information and each of the scene conditions.

[0112] The information acquisition module 2 includes an image feature capture module and an image information analysis module, which stores a facial feature capture algorithm, a license plate feature capture algorithm, etc., and is used to obtain trigger feature information and automatic trigger information.

[0113] The resolution adjustment module 3 is configured as an existing algorithm for enhancing or reducing the image resolution, which will not be described in detail in this application, and is used to adjust the image to an adaptive resolution.

[0114] The storage module 1 is configured as a memory, which can be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device, or an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the terminal device, etc., and the memory can also be a combination of an internal storage unit and an external storage device of the terminal device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0115] An NVR-based image processing system includes an NVR device with a set resolution and an image output device, wherein the NVR device includes or is connected to the image processing device as described above, wherein the image output end of the NVR device is data-connected to the data input end of an information acquisition module 2, and the image input end of the image output device is data-connected to the image output end of a resolution adjustment module 3, and images of different resolutions can be output according to scene conditions to make them more adaptable to the scene conditions.

[0116] The above is only the preferred implementation of the present application. The protection scope of the present application is not limited to the above embodiments. All technical solutions under the concept of the present application belong to the protection scope of the present application. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present application should also be regarded as the protection scope of the present application.

Claims

1. An image processing method, characterized in that: include: Establishing a first association relationship between scene conditions and resolution information; Get scene conditions and resolution information of the current image; Based on the scene condition and the resolution information, and in combination with the first association relationship, adjusting the resolution corresponding to the current image; The acquiring scene conditions includes: Setting and storing the trigger feature information obtainable via the image and the second association relationship between the trigger feature information and each of the scene conditions; Identify trigger feature information therein based on current image information; Based on the recognition result and the second association relationship, obtaining the scene condition; The trigger feature information includes: Facial feature information, license plate feature information, and dangerous object feature information in the image information; and The moving speed information of the target object obtained by processing the image information through a set algorithm; The identifying of trigger feature information based on the current image information includes: Acquire facial feature information, license plate feature information, and dangerous object feature information based on single-frame image recognition; Based on the adjacent frame images combined with the set algorithm, the moving speed information of the target object is obtained; The acquisition scene conditions include: Setting and storing automatic trigger information obtainable through image analysis or device input and a third association relationship between the information and each of the scene conditions; Automatically trigger information based on the current image recognition; Based on the recognition result and the third association relationship, obtaining the scene condition; The automatic trigger information also includes time information, geographic location information, environmental information and memory size information of the image storage location associated with the current image; the automatic trigger information based on the current image recognition includes obtaining the memory size information of the image storage location based on the memory detection algorithm, and the network bandwidth corresponding to the image transmission.

2. The image processing method according to claim 1, characterized in that: A target reference object is selected in the image, and the application scene is divided into a static scene and a dynamic scene according to whether the target reference object moves; Establishing basic image resolution information, first resolution information corresponding to static scenes, and second resolution information corresponding to dynamic scenes; Obtaining memory size information of the image storage location, if the memory size satisfies the storage of the set duration image based on the basic image resolution information, storing the acquired image at the basic image resolution; If the application scenario is a dynamic scene, the resolution corresponding to the current image is adjusted to the second resolution, and the resolution of the static scene image in the subsequent duration is adjusted to the first resolution, so that the memory size meets the requirement of storing images of the set duration.

3. The image processing method according to claim 2, characterized in that: The adjusting the resolution corresponding to the current image information to the second resolution includes: Divide the dynamic scene image into key frames and feature frames based on scene conditions; If the current image is a key frame, storing the acquired image information at the second resolution; Based on the memory size information of the image storage location and the required memory for storing the current image of the set duration, the existing feature frame is stored at the basic resolution or the first resolution.

4. An image processing device, used to implement the image processing method according to any one of claims 1 to 3, characterized in that: include: A storage module (1) for storing a first association relationship between scene conditions and resolution information; An information acquisition module (2) for acquiring scene conditions and current image resolution information; as well as A resolution adjustment module (3) for adjusting the resolution corresponding to the current image according to the scene condition and the resolution information combined with the first association relationship; wherein: The information acquisition module (2) acquires scene conditions and resolution information of the current image, and the resolution adjustment module (3) adjusts the resolution of the current image based on the first association relationship stored in the storage module (1).

5. The image processing device according to claim 4, characterized in that: The information acquisition module (2) comprises an image feature capture module for acquiring trigger feature information and an image information analysis module connected to external device data, and acquires trigger feature information and automatic trigger information in real time; The storage module (1) also stores a second association relationship between the trigger feature information and the scene condition and a third association relationship between the automatic trigger information and the scene condition.

6. An NVR-based image processing system, comprising an NVR device with a set resolution and an image output device; characterized in that: The NVR device includes or is connected to the image processing device as claimed in any one of claims 4 to 5; The image output end of the NVR device is data-connected to the data input end of the information acquisition module (2), and the image input end of the image output device is data-connected to the image output end of the resolution adjustment module (3).

Images