An image acquisition method, an image acquisition device and a computer readable storage medium

Abstract

The application discloses an image acquisition method, an image acquisition device and a computer readable storage medium, the method is applied to a camera, and the method comprises the following steps: acquiring a current monitoring image frame and a historical monitoring image frame of a monitoring scene, wherein the current monitoring image frame and the historical monitoring image frame comprise a monitoring target; using the current monitoring image frame and the historical monitoring image frame to estimate a region where the monitoring target is located in a next frame monitoring image frame, and obtaining a predicted screenshot region; and shooting the monitoring scene in a preset processing period to obtain a plurality of shot image frames, wherein a first image in the plurality of shot image frames is a global monitoring image, and the remaining frames in the plurality of shot image frames are local monitoring images obtained by shooting the predicted screenshot region. Through the above method, the application can reduce the output data amount, shorten the image acquisition time, and help improve the image noise reduction effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an image acquisition method, an image acquisition device and a computer readable storage medium. BACKGROUND

[0002] Due to the performance of components in the camera, the quality of the image captured by the camera in a low-illumination scene (such as a night scene or an indoor scene without illumination, etc.) will be greatly reduced, and the low-illumination scene is a key scene for security monitoring. In order to improve the effect of the low-illumination image, it is necessary to improve the light signal and reduce the noise, and the ways to improve the light signal mainly include: increasing the light source energy (using a light supplement lamp), collecting more light (increasing the aperture or increasing the exposure time), reducing the loss (increasing the light transmittance of the optical path such as the lens or the optical filter), etc.; the ways to reduce the noise mainly include: hardware methods (such as a hardware noise reduction circuit) and software methods (such as a noise reduction algorithm). However, increasing the light source energy has the problem of increasing light pollution, and increasing the aperture will increase the hardware cost and also face the problem of shallow depth of field. Considering the problems of light pollution and cost, in order to improve the quality of the moving target, only the method of reducing the noise can be used; but in a complex motion state, the time-domain noise reduction algorithm will be less effective, although the problem can be solved by increasing the frame rate, but increasing the frame rate means increasing the amount of data transmission, and the data transmission speed (clock frequency) is limited, and the highest frame rate commonly used at present is only 60Hz, and a small number of sensors can support 90Hz / 120Hz, which cannot realize high frame rate, resulting in weak noise reduction performance for moving objects. SUMMARY

[0003] The present application provides an image acquisition method, an image acquisition device and a computer readable storage medium, which can reduce the output data amount, shorten the image acquisition time, and help improve the image noise reduction effect.

[0004] To solve the above technical problems, the technical scheme adopted by the present application is: providing an image acquisition method, the method is applied to a camera, and the method comprises: acquiring a current monitoring image frame and a historical monitoring image frame of a monitoring scene, wherein the current monitoring image frame and the historical monitoring image frame comprise a monitoring target; using the current monitoring image frame and the historical monitoring image frame to estimate a region where the monitoring target is located in a next frame of monitoring image frame, to obtain a predicted screenshot region; and capturing the monitoring scene to obtain a plurality of captured image frames within a preset processing period, wherein a first image in the plurality of captured image frames is a global monitoring image, and the remaining frames in the plurality of captured image frames are local monitoring images obtained by capturing the predicted screenshot region.

[0005] To solve the above technical problems, another technical solution adopted by the present application is to provide an image acquisition device, comprising a memory and a processor connected to each other, wherein the memory is used to store a computer program, and the computer program is used to implement the image acquisition method in the above technical solution when executed by the processor.

[0006] To solve the above technical problems, another technical solution adopted by the present application is to provide a computer readable storage medium, which is used to store a computer program, and the computer program is used to implement the image acquisition method in the above technical solution when executed by the processor.

[0007] Through the above scheme, the beneficial effects of the present application are: first, the current monitoring image frame and the historical monitoring image frame containing the monitoring target in the monitoring scene are acquired; then, the region where the monitoring target is located in the next frame of monitoring image frame is estimated using the current monitoring image frame and the historical monitoring image frame, and a predicted screenshot region is obtained; then, the monitoring scene is photographed to obtain a plurality of frames of photographed image frames within a preset processing period, the first frame of image in the plurality of frames of photographed image frames is a global monitoring image, and the remaining frames in the plurality of frames of photographed image frames are local monitoring images obtained by photographing the predicted screenshot region; since the traditional output large image mode is changed to output local images that meet the optimal screenshot requirements, the data amount of the output image is reduced, the instantaneous frame rate can be improved, which is equivalent to reducing the motion speed of the monitoring target, and thus the denoising effect of the multi-frame denoising algorithm can be improved; and since the image size to be photographed is reduced, a shorter exposure time can be set for photographing, so that the overall photographing time is reduced and the image acquisition time is shortened. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0009] Figure 1 is a flowchart of an embodiment of the image acquisition method provided by the present application;

[0010] Figure 2 is a flowchart of another embodiment of the image acquisition method provided by the present application;

[0011] Figure 3 is Figure 2 is a flowchart of step 25 in the embodiment shown in the figure;

[0012] Figure 4 is a schematic diagram of N frames of photographed image frames provided by the present application;

[0013] Figure 5 is a schematic diagram of N+1 image frames taken by the present application;

[0014] Figure 6 is Figure 2 is a flowchart of step 28 in the embodiment shown in FIG. 8;

[0015] Figure 7 is Figure 2 is another flowchart of step 28 in the embodiment shown in FIG. 9;

[0016] Figure 8 is still another flowchart of step 28 in the embodiment shown in FIG. 10; Figure 2

[0017] Figure 9 is a structural schematic diagram of an embodiment of the image acquisition device provided by the present application;

[0018] Figure 10 is a structural schematic diagram of an embodiment of the computer readable storage medium provided by the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0020] Since the multi-frame noise reduction algorithm searches for similar pixels near the pixels that need to be reduced, and uses the found similar blocks for noise reduction, when the multi-frame noise reduction algorithm is used, the motion speed of the monitoring target will reduce the effect of the algorithm. In the worst case, if the motion speed of the monitoring target is too fast, the similar block search will not find similar pixels that can be used for noise reduction, and the multi-frame noise reduction algorithm will have no noise reduction effect on this part of the area, thereby forming a very ugly noise band around the moving monitoring target.

[0021] ​In view of the problem of slow multi-frame noise reduction processing speed, the existing scheme speeds up by changing the algorithm itself, and the application provides a prediction preferred image capturing mode, which aims to speed up the image acquisition speed of the monitoring target during multi-frame noise reduction. The prediction preferred image capturing mode finds the area where the monitoring target will appear that meets the preferred image capturing requirements through historical information, then performs local area exposure and output on these areas, and then transmits the local images obtained by using the prediction preferred image capturing mode to the multi-frame noise reduction algorithm for noise reduction. Since the output of the local images that meet the preferred image capturing requirements to the multi-frame noise reduction algorithm reduces the data volume, improves the instantaneous frame rate, helps to improve the multi-frame noise reduction effect, and speeds up the acquisition of images. The specific implementation scheme of the application will be described below.

[0022] Please refer to Figure 1 , Figure 1 is a flowchart of an embodiment of an image acquisition method provided by the application, which is applied to a camera, and the method comprises the following steps:

[0023] Step 11: acquiring a current monitoring image frame and a historical monitoring image frame of a monitoring scene.

[0024] The camera can capture the monitored scene to generate a current monitoring image frame and store it for subsequent use. The historical monitoring image frame is an image frame located before the current monitoring image frame in time, for example, the previous image frame of the current monitoring image frame. The current monitoring image frame and the historical monitoring image frame include a monitoring target, which is a target of interest in the monitoring scene, for example, personnel entering and leaving a supermarket.

[0025] Step 12: estimating the area where the monitoring target is located in the next monitoring image frame by using the current monitoring image frame and the historical monitoring image frame to obtain a predicted image capturing area.

[0026] After the current monitoring image frame and the historical monitoring image frame are acquired, they can be analyzed and processed to determine the predicted image capturing area, which is an area that meets the preferred image capturing requirements and is obtained by estimating the area where the monitoring target is located in the next monitoring image frame. By setting the preferred image capturing requirements, the traditional image capturing and subsequent optimization process are moved forward to improve the effectiveness of image capturing.

[0027] Step 13: capturing the monitoring scene to obtain a plurality of captured image frames within a preset processing period.

[0028] After the predicted image capturing area is determined, capturing can be performed to generate a plurality of captured image frames. The first image in the plurality of captured image frames is a global monitoring image, and the remaining images in the plurality of captured image frames are local monitoring images obtained by capturing the predicted image capturing area. The length of the preset period can be set according to the specific application scenario.

[0029] It can be understood that after the end of a preset processing period, the next preset processing period will come, and there is no other process between the two.

[0030] The embodiment provides a prediction preferred snapshot mode, which modifies a conventional mode of outputting a large image into a mode of outputting only a small image meeting a preferred snapshot requirement. Due to reduction of the output data amount, the instantaneous frame rate can be improved, which is equivalent to reducing the motion speed of a monitoring target, and thus the effect of a multi-frame noise reduction algorithm can be improved. In addition, due to the fact that only a partial image of the monitoring scene is acquired, the image acquisition speed can be accelerated, the time spent in image acquisition can be reduced, and the image processing efficiency can be improved.

[0031] Please refer to Figure 2 , Figure 2 is a flowchart of another embodiment of an image acquisition method provided by the application, the method being applied to a camera, and the method comprises the following steps:

[0032] Step 21: acquiring a current monitoring image frame and a historical monitoring image frame of a monitoring scene.

[0033] The camera comprises a memory, in order to acquire the historical monitoring image frame, the historical monitoring image frame can be read from the memory; the current monitoring image frame is obtained by shooting the monitoring scene and is stored.

[0034] Step 22: acquiring detection and recognition information of the historical monitoring image frame, denoted as historical detection and recognition information.

[0035] After the historical monitoring image frame is acquired, if the historical monitoring image frame has been subjected to target detection and rotation angle recognition processing, the detection and recognition information corresponding to the historical monitoring image frame can be directly read from the memory; if the historical monitoring image frame has not been subjected to target detection and rotation angle recognition processing, target detection and rotation angle recognition processing is performed to generate the detection and recognition information of the historical monitoring image frame.

[0036] Step 23: performing target detection and rotation angle recognition on the current monitoring image frame to obtain current detection and recognition information and storing the current detection and recognition information.

[0037] After the current monitoring image frame is acquired, target detection and rotation angle recognition can be performed on the current monitoring image frame to obtain detection and recognition information, the detection and recognition information comprising the position of a detected target, installation preset information of the camera, the focal length of the camera, the rotation angle of the monitoring target or image light and dark distribution information, and the installation preset information can be the installation height or angle of the camera.

[0038] In one specific embodiment, the camera includes a first image sensor, which can be used to expose the image for a first preset exposure time to obtain the current monitoring image frame. Specifically, the first preset exposure time is relatively long, that is, the visible light signal of all photosensitive units of the first image sensor is output with a longer exposure (e.g., 10ms) to form the current monitoring image frame. Then, target detection and three-dimensional rotation angle recognition are performed on the current monitoring image frame, and the predicted capture area is determined using the current monitoring image frame.

[0039] In another specific embodiment, if the camera has two image sensors, one sensor can be used to obtain an infrared image under infrared illumination, while the other image sensor is exposed for a longer time (e.g., 10ms) to output a visible light image of all photosensitive units of that image sensor, thereby enabling the output of a video stream composed of visible light images. Since the illumination brightness of the visible light path may be relatively dim or absent to reduce light pollution, the imaging quality of the infrared light path is better. Target detection, three-dimensional rotation angle recognition, and determination of the predicted capture area are performed on the infrared image. That is, the camera includes a first image sensor and a second image sensor, and the current monitoring image frame includes a first monitoring image frame and a second monitoring image frame. The first image sensor is used to expose the image for a first preset exposure time to obtain the first monitoring image frame. The second image sensor is used to acquire an infrared image, which is then used as the second monitoring image frame. Target detection and rotation angle recognition are performed on the infrared image to obtain the current detection and recognition information.

[0040] Understandably, the infrared and visible light images output at all times can be separately composed of video streams for output.

[0041] Step 24: Using historical detection and recognition information and current detection and recognition information, estimate the area where the monitored target is located in the next frame of the monitoring image, and record it as the candidate image capture area.

[0042] After obtaining the detection and recognition information of the current monitoring image frame and the detection and recognition information of the historical monitoring image frames, the current detection and recognition information and the historical detection and recognition information are processed to predict the area where the monitoring target may appear in the next frame image, and the area is recorded as a candidate capture area. The candidate capture area can be a rectangular area, and the coordinates of the upper left corner, length and width of the rectangular area can be stored.

[0043] Step 25: Score the candidate capture regions to obtain the first score value, and determine the predicted capture region based on the first score value.

[0044] After identifying the candidate capture areas, they can be scored to select the best candidate capture areas. The specific method is as follows: Figure 3 As shown, it includes the following steps:

[0045] Step 251: Obtain evaluation dimensions of the candidate screenshot region.

[0046] According to the current detection recognition information and the historical detection recognition information, each evaluation dimension of the candidate screenshot region is calculated, the evaluation dimension includes multiple evaluation parameters, and each evaluation parameter corresponds to a score value; specifically, the evaluation dimension includes at least one of a pixel size of the monitoring target, a rotation angle of the monitoring target, a speed and acceleration of the monitoring target, a definition of the monitoring target, a brightness of the monitoring target, a predicted position of the monitoring target, and whether the monitoring target is about to leave a monitoring region corresponding to the monitoring scene.

[0047] Further, the pixel size of the monitoring target is the number of pixels occupied by the monitoring target in the whole image, for example, taking a person as the monitoring target, a face width pixel greater than a set value can be set, if the face width is less than or equal to the set value, the score value is 0, and if the face width is greater than the set value, the score value is 1.

[0048] The rotation angle of the monitoring target is a three-dimensional rotation angle of the monitoring target relative to the camera, and the three-dimensional rotation angle includes a pitch angle, a yaw angle, and a roll angle, for example: the posture of the face, the face can be preferably directed to the camera or away from the camera, and when away from the camera, the features of the back clothes or bag of the monitoring target can be captured.

[0049] For the speed and acceleration of the monitoring target, the faster the monitoring target, the lower the corresponding score value, that is, the motion speed of the monitoring target is slow.

[0050] The definition of the monitoring target is the definition of the monitoring target predicted according to the depth of field range calculated after the camera is installed, the higher the definition, the higher the definition corresponding to the score value.

[0051] The brightness of the monitoring target is the brightness of the area where the monitoring target is about to enter, which is predicted according to the brightness of the picture and the walking direction of the monitoring target, the brighter the brightness, the higher the brightness corresponding to the score value.

[0052] Whether the monitoring target is about to leave the monitoring region is whether the monitoring target leaves the monitoring region within a set time, if the monitoring target is about to leave the monitoring region and has not been predicted and selected for capturing a screenshot, the prediction and selection for capturing a screenshot is immediately performed.

[0053] It can be understood that if the prediction and selection for capturing a screenshot has been performed on the monitoring target, and the score value of the preferred processing in the intelligent processing during post-processing is greater than a certain value, the prediction and selection for capturing a screenshot is not performed again, and the specific process of the preferred processing will be described below.

[0054] Step 252: performing weighted summation on the score values corresponding to the at least part of the evaluation parameters to obtain a first score value.

[0055] The prediction score is generated by using the at least part of the evaluation dimensions, and a score value of the candidate image capture region is generated, which is denoted as a first score value.

[0056] Step 253: determining whether the first score value is greater than a first preset score value.

[0057] It is determined whether the first score value of each candidate image capture region is greater than a first preset score value. If the first score value is less than or equal to the first preset score value, no processing is performed.

[0058] Step 254: if the first score value is greater than the first preset score value, the candidate image capture region is recorded as a predicted image capture region.

[0059] If the first score value is greater than the first preset score value, it indicates that the candidate image capture region meets the preferred image capture requirement, and the candidate image capture region is recorded as a predicted image capture region. Understandably, the number of the predicted image capture regions is not limited to one, and can be two or more.

[0060] Step 26: capturing a plurality of image frames of the monitoring scene in a preset processing period.

[0061] The embodiment provides two modes: a linear mode and a wide dynamic mode. In the linear mode, one image of all the light sensing units of an image sensor is captured, and N-1 (N≥2) images of regions meeting the preferred image capture requirement are captured, a total of N images, and a preset processing period is a total time of exposure and output of the N images. In the wide dynamic mode, one image of all the light sensing units of an image sensor is captured, N-1 images of regions meeting the preferred image capture requirement are captured, and one image of a target region of a motor vehicle is captured for super-short exposure, a total of N+1 images, and a preset processing period is a total time of exposure and output of the N+1 images.

[0062] Further, for the linear mode, in the preset processing period, the monitoring scene is exposed for a first preset exposure time to obtain a global monitoring image, the first preset exposure time can be 0-10 ms; and the predicted image capture region is exposed for a second preset exposure time to obtain a local monitoring image, the second preset exposure time can be 0-4 ms. For example, in the linear mode, in the preset processing period, the monitoring scene is exposed for a first preset exposure time to obtain a global monitoring image, and the predicted image capture region is exposed for a second preset exposure time to obtain a local monitoring image. Figure 4 For example, there are four monitoring targets: O1-O4, the targets O1 and O3 meet the preferred image capture requirement, a full image is exposed in the first frame, and the targets O1 and O3 are exposed multiple times in the second frame to the Nth frame, and then an image is output.

[0063] For the wide dynamic mode, the monitoring targets include motor vehicles, and when the motor vehicles exist in the monitoring scene, the area where the motor vehicles are located is detected, which is recorded as the motor vehicle target area; the motor vehicle target area is exposed with a third preset exposure time to obtain the last local monitoring image, and the motor vehicle is a high-speed moving target. For the high-speed moving target, in order to reduce the trailing, a short exposure can be used, that is, the motor vehicle target area is exposed for a very short time, and the third preset exposure time is about 0.25 ms. For example, taking Figure 5 for example, there are four monitoring targets: O1-O4, and the targets O1 and O3 meet the preferred snapshot requirement. In the 2nd frame to the Nth frame, only the targets O1 and O3 are exposed and the images are output, and in the N+1th frame, only the motor vehicle target O5 is exposed for a very short time and the image is output.

[0064] It can be understood that when there are two sensors, the local monitoring image includes the first local monitoring image and the second local monitoring image. The first sensor can be used to expose the predicted snapshot area with a second preset exposure time to obtain the first local monitoring image; and the second sensor can be used to expose the predicted snapshot area with the second preset exposure time to obtain the second local monitoring image.

[0065] In a specific embodiment, in order to verify the feasibility of the scheme, tests can be carried out:

[0066] 1) Precondition

[0067] A. The preset processing period is 40 ms.

[0068] B. The output resolution of the image sensor is 4 million, that is, 2688x1520; and the main control chip in the camera can support a processing speed of 90 fps.

[0069] C. The exposure time of the first frame full image is 0-10 ms, the exposure time of the local monitoring image is 0-4 ms, and the exposure time of the shortest frame is about 0.25 ms.

[0070] D. The pixel size of the monitoring target is 120x120, and the pixel size of a motor vehicle is 900x450. It is assumed that there are 10 motor vehicles in the monitoring scene.

[0071] 2) Performance calculation

[0072] A. Except for the large image output in the 1st frame in a preset processing period, the main control chip can process the data of about 2 large images, and the calculation formula is as follows:

[0073] (90 / (1000 / 40)-1)≈2

[0074] B. The number of frames of small images that can be processed in a preset processing period is about 7, and the calculation formula is as follows:

[0075] (40-10-0.25) / 4≈7

[0076] C. In linear mode, the number of monitoring targets that can be processed in the 2nd frame to the 8th frame in a preset processing period is:

[0077] (2x2688x1520) / 7 / (120x120)=81

[0078] D. In wide dynamic mode, the number of monitoring targets that can be processed in the 2nd frame to the 8th frame in a preset processing period is:

[0079] (2x2688x1520-900x450x10) / 7 / (120x120)=40

[0080] Currently, it is required to process more than 30 monitoring targets per second, and in the present scheme, 40-81 targets can be processed per 40 ms, which fully meets the requirement.

[0081] After the shooting is completed, image signal processing (ISP) can be performed, such as image noise reduction, image enhancement, linearization, bad pixel removal, color space conversion, rotation, sharpening or scaling, etc. The present embodiment is described by taking image noise reduction as an example.

[0082] Step 27: Perform multi-frame noise reduction processing on the multi-frame local monitoring images, and output one frame of noise reduction image.

[0083] The global monitoring image obtained by shooting is output, and the multi-frame noise reduction or other image processing can be performed on the monitoring targets for the prediction preferred image capture result (i.e. the multi-frame local monitoring image output in step 26); after the multi-frame noise reduction processing, the multi-frame images obtained by the prediction preferred image capture can be synthesized and output as one frame, and different image parameters (such as gain parameter or sharpening parameter) can be selected for different monitoring targets with different brightness or clarity to improve the processing effect; for example, assuming that each preset processing period includes M frames of images, for the 2nd to Mth frames, one frame of noise reduction image is output after multi-frame noise reduction processing; then the M+1th frame of global monitoring image is output, and another frame of noise reduction image is output after multi-frame noise reduction processing for the M+2th to 2Mth frames, and so on, to form a video stream output. It can be understood that the multi-frame noise reduction can be performed in RAW domain or YUV domain.

[0084] In another embodiment, the noise reduction image can also be scored to obtain a second score value; then it is judged whether the second score value is greater than a second preset score value; if yes, the noise reduction image is output.

[0085] In other embodiments, steps 21-27 can be repeatedly performed to obtain multiple denoised images, and each denoised image can be scored to obtain a second score value; after each denoised image is obtained, it is determined whether the monitoring target is located in the monitoring scene; if the monitoring target is not located in the monitoring scene, the denoised image corresponding to the maximum value of all second score values is output.

[0086] In other embodiments, after the denoised image is obtained, intelligent processing can be performed, which includes one or more of the processes of optimization, attribute extraction or recognition, etc. The optimization process is to optimize the results when the same monitoring target has multiple predicted optimal image capture results, and obtain an optimal image capture region. The attribute extraction or recognition can be located before or after the optimization process in the intelligent processing, which will be described in detail below.

[0087] In a specific embodiment, the intelligent processing includes the optimization process, as shown in FIG. 8, the processing scheme is as follows: Figure 6

[0088] Step 281a: scoring the multiple denoised images to obtain second score values.

[0089] Steps 21-27 can be repeatedly performed to obtain multiple denoised images, and the denoised images can be scored to obtain second score values corresponding to each denoised image. Specifically, traditional scoring methods can be used, such as evaluation dimensions of clarity, brightness, three-dimensional rotation angle of the monitoring target, whether it is a positive or negative face, or whether it is blocked, etc. Deep learning methods can also be used for scoring, or the image quality of the denoised image can be used as the second score value.

[0090] Step 282a: determining whether the maximum value of all second score values is greater than a second preset score value.

[0091] After obtaining the second score value of each denoised image, the maximum value of all second score values can be calculated, and then the size relationship between the maximum value and the second preset score value is determined, so as to select the optimal denoised image.

[0092] Step 283a: if the maximum value of all second score values is greater than the second preset score value, the denoised image with the maximum second score value is output.

[0093] If the maximum value of all second score values is less than or equal to the second preset score value, it indicates that the image quality of the denoised image is poor and does not meet the user's set requirements. At this time, the step of obtaining the current monitoring image frame and the historical monitoring image frame of the monitoring scene can be returned to be performed, i.e., step 21 is performed, until the maximum value of all second score values is greater than the second preset score value.

[0094] ​It can be understood that a preset number of times can also be set. If the maximum value in all second scores is still less than or equal to the second preset score after the preset number of times is exceeded, it indicates that the image quality of the current denoised image is not good, at which time the prediction and preferred image capturing mode can be exited.

[0095] In another specific embodiment, the intelligent processing includes preferred processing, attribute extraction and identification, and the preferred processing is arranged before the attribute extraction and identification, as shown in Figure 7 The processing scheme is as follows:

[0096] Step 281b: scoring the multiple denoised images to obtain second scores.

[0097] Step 281b is the same as step 281a and will not be described again here.

[0098] Step 282b: taking the denoised image with the maximum second score as a candidate denoised image.

[0099] The second score of each denoised image can be counted, and the denoised image with the highest second score is recorded as a candidate denoised image.

[0100] Step 283b: processing the candidate denoised image to obtain attribute identification information of the candidate denoised image.

[0101] The attribute extraction and identification processing can be performed on the candidate denoised image to generate attribute identification information. The attribute identification information includes attribute features and identification results of the denoised image. The attribute features include the category, gender, age, clothing color or hair color of the monitored target. The identification result can be searched in a specific registration library to determine whether the identified monitored target belongs to the registration library. If it belongs to the registration library, the corresponding target is determined.

[0102] It can be understood that in other embodiments, only attribute extraction or identification processing can be performed.

[0103] Step 284b: scoring the candidate denoised image based on the attribute identification information to obtain a third score.

[0104] After the attribute identification information is obtained, the candidate denoised image can be scored according to the attribute identification information to obtain a third score corresponding to the candidate denoised image.

[0105] Step 285b: determining whether the third score is greater than a third preset score.

[0106] After the candidate denoised image is scored, the size relationship between the third score and the third preset score can be determined.

[0107] Step 286b: if the third score value is greater than the third preset score value, output the candidate denoising image.

[0108] If the third score value is greater than the third preset score value, it indicates that the image quality of the current candidate denoising image meets the requirements, and the candidate denoising image can be directly output. If the third score value is less than or equal to the third preset score value, return to execute the step of obtaining the current monitoring image frame and the historical monitoring image frame of the monitoring scene, that is, execute step 21, until the third score value is greater than the third preset score value.

[0109] In other specific embodiments, the intelligent processing includes preferred processing, attribute extraction, and identification, and the preferred processing is arranged after the attribute extraction and the identification, as shown in Figure 8 The processing scheme is as follows:

[0110] Step 281c: processing the multiple frames of denoising images to obtain attribute identification information of the denoising images.

[0111] After repeatedly executing steps 21-27, multiple frames of denoising images can be obtained, and attribute extraction and identification processing are performed on these denoising images to generate attribute identification information corresponding to the denoising images.

[0112] Step 282c: scoring the denoising image frame based on the attribute identification information to obtain a second score value.

[0113] After obtaining the attribute identification information of the denoising images, the attribute identification information can be used to score the denoising images to obtain a second score value.

[0114] Step 283c: taking the denoising image with the maximum second score value as a candidate denoising image.

[0115] Step 283c is the same as step 282b and will not be described here.

[0116] Step 284c: scoring the candidate denoising image to obtain a third score value.

[0117] Step 285c: determining whether the third score value is greater than a third preset score value.

[0118] Step 286c: if the third score value is greater than the third preset score value, output the candidate denoising image.

[0119] If the third score value is less than or equal to the third preset score value, return to execute the step of obtaining the current monitoring image frame and the historical monitoring image frame of the monitoring scene, until the third score value is greater than the third preset score value; steps 284c-286c are the same as steps 284b-286b and will not be described here.

[0120] The scheme provided in the application can complete the block matching based time domain noise reduction processing of the moving target in a very short time, avoid the noise reduction effect decline caused by target turning change, the higher the frame rate of the image capture, the slower the equivalent speed, the closer to the effect of the static object, the image capture time can be greatly shortened, taking a preset processing period of 40ms, an image capture frame rate of 25fps, and 7 images required by a multi-frame noise reduction algorithm as an example, in the existing mode, 7 images are captured for the full image, and 280ms is required for capturing, and the data capture can be completed in 40ms by using the scheme of the application, and the time consumption is reduced to one seventh of the original. In addition, the results of the prediction and optimization of the image capture can be respectively subjected to multi-frame noise reduction and other traditional ISP processing in the ISP processing process. Different image parameters can be selected for different monitoring targets such as brightness and clarity, for example: a larger gain is required for a darker target, and a smaller gain is required for a brighter target. If the same gain is used, the noise of the brighter target may be too large or the image of the darker target may be too dark, and different image parameters can achieve better results.

[0121] Referring to Figure 9 , Figure 9 is a structural schematic diagram of an embodiment of an image acquisition device provided by the application. The image acquisition device 90 comprises a memory 91 and a processor 92 connected to each other. The memory 91 is used to store a computer program. The computer program is used to realize the image acquisition method in the above embodiment when executed by the processor 92.

[0122] Referring to Figure 10 , Figure 10 is a structural schematic diagram of an embodiment of a computer readable storage medium provided by the application. The computer readable storage medium 100 is used to store a computer program 101. The computer program 101 is used to realize the image acquisition method in the above embodiment when executed by a processor.

[0123] The computer readable storage medium 100 can be a server, a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0124] In several embodiments provided in the application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules or units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0125] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e. may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0126] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0127] The above is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. An image acquisition method, characterized in that, The application is applied to a camera, comprising: acquiring a current monitoring image frame and a historical monitoring image frame of a monitoring scene, wherein the current monitoring image frame and the historical monitoring image frame comprise a monitoring target; estimating a region where the monitoring target is located in a next monitoring image frame by using the current monitoring image frame and the historical monitoring image frame, to obtain a predicted screenshot region; acquiring a plurality of photographed image frames of the monitoring scene within a preset processing period, wherein a first image frame in the plurality of photographed image frames is a global monitoring image, and the remaining image frames in the plurality of photographed image frames are local monitoring images obtained by photographing the predicted screenshot region; wherein the step of estimating the region where the monitoring target is located in the next monitoring image frame by using the current monitoring image frame and the historical monitoring image frame to obtain the predicted screenshot region comprises: acquiring detection and recognition information of the historical monitoring image frame, denoted as historical detection and recognition information; performing target detection and rotation angle recognition on the current monitoring image frame to obtain current detection and recognition information, and storing the current detection and recognition information; estimating a region where the monitoring target is located in the next monitoring image frame by using the historical detection and recognition information and the current detection and recognition information, denoted as a candidate screenshot region; scoring the candidate screenshot region to obtain a first score value, and determining the predicted screenshot region based on the first score value.

2. The image acquisition method of claim 1, wherein, The detection and recognition information comprises a position of the monitoring target, installation preset information of the camera, a focal length of the camera, a rotation angle of the monitoring target, or image light and dark distribution information, and the step of scoring the candidate screenshot region to obtain the first score value comprises: acquiring an evaluation dimension of the candidate screenshot region, wherein the evaluation dimension comprises a plurality of evaluation parameters, and each evaluation parameter corresponds to a score value; performing weighted summation on score values corresponding to at least part of the evaluation parameters to obtain the first score value.

3. The image acquisition method according to claim 2, wherein the evaluation dimension comprises at least one of a pixel size of the monitoring target, a rotation angle of the monitoring target, a speed and acceleration of the monitoring target, a definition of the monitoring target, a brightness of the monitoring target, a predicted position of the monitoring target, and whether the monitoring target is about to leave the monitoring scene.

4. The image acquisition method of claim 1, wherein, The step of determining the predicted screenshot region based on the first score value comprises: determining whether the first score value is greater than a first preset score value; if yes, the candidate screenshot region is recorded as the predicted screenshot region.

5. The image acquisition method of claim 1, wherein, The step of acquiring a plurality of photographed image frames of the monitoring scene within a preset processing period comprises: exposing the monitoring scene to light with a first preset exposure time within the preset processing period to acquire the global monitoring image; exposing the predicted screenshot region to light with a second preset exposure time to obtain the local monitoring image.

6. The image acquisition method of claim 5, wherein, The monitoring target comprises a motor vehicle, and the step of acquiring a plurality of photographed image frames of the monitoring scene within a preset processing period further comprises: When the motor vehicle exists in the monitoring scene, an area where the motor vehicle exists is detected, denoted as a motor vehicle target area; Exposure is performed on the motor vehicle target area with a third preset exposure time, to obtain a last local monitoring image.

7. The image acquisition method of claim 1, wherein, The method further comprises: Multi-frame noise reduction processing is performed on the multiple local monitoring images, to output a noise reduction image.

8. The image acquisition method of claim 7, wherein, The method further comprises: Scoring is performed on the noise reduction image, to obtain a second score value; It is determined whether the second score value is greater than a second preset score value; If yes, the noise reduction image is output.

9. The image acquisition method of claim 7, wherein, The method further comprises: Multiple noise reduction images are obtained, and scoring is performed on the noise reduction images, to obtain a second score value; It is determined whether the monitoring target exists in the monitoring scene; If no, a noise reduction image corresponding to a maximum value among all the second score values is output.

10. The image acquisition method of claim 1, wherein, The camera comprises a memory, and the step of obtaining a current monitoring image frame and a historical monitoring image frame of the monitoring scene comprises: The historical monitoring image frame is read from the memory; The monitoring scene is photographed, to obtain the current monitoring image frame, which is stored.

11. The image acquisition method of claim 10, wherein, The camera comprises a first image sensor, and the step of photographing the monitoring scene to obtain the current monitoring image frame comprises: Exposure is performed on the monitoring scene with the first image sensor and a first preset exposure time, to obtain the current monitoring image frame.

12. The image acquisition method of claim 10, wherein, The camera comprises a first image sensor and a second image sensor, the current monitoring image frame comprises a first monitoring image frame and a second monitoring image frame, and the step of photographing the monitoring scene to obtain the current monitoring image frame comprises: Exposure is performed on the monitoring scene with the first image sensor and a first preset exposure time, to obtain the first monitoring image frame; An infrared image is obtained with the second image sensor, the infrared image is taken as the second monitoring image frame, target detection and rotation angle recognition are performed on the infrared image, and current detection recognition information is obtained.

13. The image acquisition method of claim 12, wherein, The local monitoring image comprises a first local monitoring image and a second local monitoring image, and the method further comprises: Exposure is performed on the predicted image capturing area with a second preset exposure time by using a first sensor, to obtain the first local monitoring image; Exposure is performed on the predicted image capturing area with the second preset exposure time by using a second sensor, to obtain the second local monitoring image.

14. An image acquisition device, characterized in that The computer program, when executed by the processor, is configured to implement the image acquisition method in any one of claims 1-13.

15. A computer readable storage medium for storing a computer program, characterized in that, The computer program, when executed by the processor, is configured to implement the image acquisition method in any one of claims 1-14.

Images