Community video optimization acquisition method based on artificial intelligence

Abstract

The invention relates to the technical field of image data processing, in particular to a community video optimized acquisition method based on artificial intelligence, which comprises the following steps: acquiring each video frame of a community video, and dividing any video frame into a plurality of target areas; performing noise removal processing on each video frame by using an improved Gaussian filtering algorithm so as to obtain each video frame after noise removal; and reconstructing a community video based on each video frame after noise removal. Wherein the improved Gaussian filtering algorithm comprises a weight value, and the weight value is related to a noise parameter of the target area and a mean value of gradient amplitudes of all pixel points in the target area. The problem that edge information is lost or noise remains when an existing Gaussian filtering algorithm is used for noise removal processing is solved.

Description

Technical Field

[0001] The present invention relates to the field of image data processing technology. More specifically, the present invention relates to an artificial intelligence-based community video optimization acquisition method. Background Art

[0002] With the rapid development of artificial intelligence technology, video surveillance is increasingly used in community security, environmental monitoring, traffic management and other fields. However, as the complexity and dynamism of the community environment increase, traditional video acquisition methods face many challenges, such as unstable video quality, poor adaptability to scene changes, and inaccurate information extraction.

[0003] At present, intelligent video analysis technology based on artificial intelligence, especially the application of deep learning and computer vision, has provided a new development direction for community video collection. Through intelligent algorithms, cameras can not only dynamically adjust collection parameters according to changes in scenes and events, but also identify and classify important events in real time, improving the quality of video data and the accuracy of information extraction.

[0004] In practical applications, community video collection still faces technical bottlenecks such as multi-camera collaboration, real-time data processing, and transmission bandwidth limitations. How to optimize video collection methods through effective algorithms, improve video quality, enhance intelligent recognition capabilities, and solve the problems of large-scale data storage and real-time analysis has become a key technical issue that needs to be solved urgently.

[0005] The traditional Gaussian filtering algorithm smoothes the image by using a weight matrix generated based on the Gaussian distribution function. This means that no matter what the content of any area in the image is, the filter will process it in the same way. This will cause unnecessary blurring in areas with rich details and fail to effectively distinguish important structural information. The weight value of each pixel is determined by its distance from the current pixel and the standard deviation of the Gaussian function. The farther the point is from the center pixel, the smaller its weight. Due to its fixed weight matrix, it does not distinguish between edge areas or detailed areas of the image. In places with rich edges or textures, over-smoothing may cause image distortion and loss of details.

[0006] Therefore, it is difficult for the traditional existing Gaussian filtering algorithm to achieve both high efficiency and high precision, which leads to the problem of edge information loss or noise residue when removing noise. Summary of the invention

[0007] In order to solve the problem of edge information loss or noise residue caused by noise removal proposed in the above background technology, the present invention provides the following solution.

[0008] The present invention provides a community video optimization acquisition method based on artificial intelligence, comprising: obtaining each video frame of the community video, dividing any video frame into a plurality of target areas; performing noise removal processing on each video frame by using an improved Gaussian filtering algorithm to obtain each video frame after noise removal; and reconstructing the community video based on each video frame after noise removal; wherein the improved Gaussian filtering algorithm includes a weight value, the first In the video frame The weight of the target area for, , where For the In the video frame The noise parameters of the target area, For the In the video frame The noise parameters of the target area, For the The total number of target areas in a video frame, The natural constant The exponential function with base , For the In the video frame The mean value of the gradient amplitude of all pixels in the target area; the noise parameter is In the video frame The noise characteristic value of the target area and the standard deviation of the grayscale values ​​of all pixels are positively correlated with the The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic value in the target area and the standard deviation of the grayscale values ​​of all pixels are inversely correlated.

[0009] The above technical solution introduces an improved Gaussian filtering algorithm, combines the noise characteristics of the target area and the pixel gradient information, effectively removes the noise in the community video frame, improves the video quality, and solves the problem of edge information loss or noise residue when the existing Gaussian filtering algorithm is used to remove noise.

[0010] Furthermore, the noise parameter is: , where For the In the video frame The standard deviation of the grayscale values ​​of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The standard deviation of the grayscale values ​​of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The noise characteristic value of the target area, For the first The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic values ​​of the target area, is a normalized function, and the noise characteristic value represents the credibility of the target area.

[0011] The above technical solution constructs a noise parameter to measure the credibility of the target area by calculating the standard deviation difference of the grayscale value of the pixel in the target area and the degree of deviation of the noise characteristic value. The normalization function is used to enhance the difference contrast, making the features of the low credibility area more prominent. At the same time, through the comprehensive calculation of multi-frame information in the neighborhood, the robustness to noise interference is improved, which can effectively suppress the influence of random noise on single-frame data.

[0012] Furthermore, the noise parameter is: , where For the In the video frame The mean gray value of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean gray value of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The noise characteristic value of the target area, For the first The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic values ​​of the target area, is a normalized function, and the noise characteristic value represents the credibility of the target area.

[0013] The above technical solution constructs a noise parameter to measure the credibility of the target area by calculating the difference in the mean grayscale value of the pixels in the target area and combining it with the degree of deviation of the noise characteristic value. The introduction of the normalization function enhances the sensitivity to grayscale changes, making the characteristics of the abnormal area more obvious. At the same time, the comprehensive calculation using the neighborhood multi-frame information improves the adaptability to local illumination changes and random noise.

[0014] Furthermore, the neighborhood radius is set to 5.

[0015] Furthermore, the noise characteristic value is: , where For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean value of the gradient amplitude of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The mean value of the gradient direction of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean value of the gradient directions of all pixels in the target area.

[0016] The above technical solution constructs parameters to measure the noise characteristics of the target area by calculating the mean of the pixel gradient amplitude in the target area and combining the degree of change in the gradient direction of multiple frames in the neighborhood. By using the amplitude and direction differences of the gradient information, the noise characteristics can more accurately reflect the changes in the local image structure, and at the same time, the influence of single-frame random noise is reduced through multi-frame calculation. The sensitivity to edge features is improved, and the real target and noise interference can be more effectively distinguished in complex backgrounds, thereby improving the accuracy of the determination of the credibility of the target area.

[0017] Furthermore, each video frame of the community video is acquired using a high frame rate camera.

[0018] Furthermore, the method also includes grayscale processing of each video frame.

[0019] Furthermore, the multiple target areas include: dividing any video frame into multiple target areas of the same size.

[0020] Furthermore, the multiple target areas include: dividing any video frame into multiple target areas of any size.

[0021] Furthermore, the community video is reconstructed based on each video frame after noise removal, specifically: The video frames and timestamps after the noise removal process are obtained, the video frames after the noise removal process are sorted in chronological order based on the timestamps, and the community video is reconstructed based on the sorting results.

[0022] The above technical solution ensures the temporal consistency of the video content by sorting the timestamps of the denoised video frames, and reconstructs the video based on the sorting results, thereby effectively eliminating noise interference while retaining the original time logic of the video.

[0023] The beneficial effects of the present invention are: The present invention optimizes the collection of community videos by introducing an improved filtering algorithm and noise assessment mechanism, thereby achieving efficient denoising and video reconstruction. First, the video frame is divided into target areas and a weight-controlled filtering method is used to effectively suppress noise while maintaining edge and texture information, thereby improving video clarity. Secondly, the noise characteristic value is calculated by combining the pixel grayscale characteristics, gradient amplitude and direction changes, so that noise suppression is more accurate, and stability is enhanced by neighborhood multi-frame calculation to reduce the impact of illumination changes and random interference. In addition, the quality and availability of video data are further optimized through high frame rate camera technology and grayscale processing. In the process of video reconstruction after denoising, the time logic of the video is restored by timestamp sorting to ensure timing consistency and picture smoothness, making the optimized video more coherent and reliable. The present invention improves the quality of community videos and solves the problem of edge information loss or noise residue when the existing Gaussian filtering algorithm processes noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flowchart schematically illustrating an artificial intelligence-based community video optimization acquisition method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] An implementation example of a community video optimization collection method based on artificial intelligence.

[0026] like Figure 1 As shown, the flowchart of the community video optimization acquisition method based on artificial intelligence in an embodiment of the present invention includes the following steps: S1: Obtain each video frame of the community video and divide any video frame into multiple target areas.

[0027] In one embodiment, a high frame rate camera can be used to capture community videos, and the acquired video frames can be grayed. Graying can effectively reduce the computational complexity, while enhancing the feature contrast of the video, making subsequent analysis, detection or enhancement processing more efficient and stable. In addition, graying can reduce the interference caused by color changes and improve the extraction accuracy of the target area; The multiple target areas include: dividing any video frame into multiple target areas of the same size. This uniform division method can ensure that each area has a similar resolution while ensuring computational efficiency, which is helpful for subsequent noise reduction, motion detection or feature extraction. For example, when performing noise removal or target recognition, the data distribution of each area is more uniform, thereby improving the adaptability and stability of the algorithm. In addition, the division of target areas of the same size facilitates parallel computing, accelerates the video processing process, and improves overall processing efficiency.

[0028] In another embodiment, the multiple target areas include: dividing any video frame into multiple target areas of any size, that is, dividing the video frame into multiple target areas of any size according to the complexity of the video frame content or specific application requirements. For example, in the presence of moving targets or mutation areas, smaller division units can be used to capture key changes more finely; while in static areas or low-texture areas, larger division units can be used to reduce computational redundancy and improve processing efficiency. This adaptive division strategy can effectively optimize the allocation of computing resources, improve the accuracy of video processing, and reduce the computational burden on irrelevant areas, thereby improving the real-time and robustness of the overall system.

[0029] S2: using an improved Gaussian filtering algorithm to perform noise removal processing on each of the video frames to obtain each video frame after noise removal.

[0030] In one embodiment, the improved Gaussian filtering algorithm includes weight values, In the video frame The weight of the target area for, , where For the In the video frame The noise parameters of the target area, For the In the video frame The noise parameters of the target area, For the The total number of target areas in a video frame, The natural constant The exponential function with base , For the In the video frame The mean value of the gradient amplitude of all pixels in the target area; By constructing weight values ​​based on noise parameters and gradient amplitude, the improved filtering algorithm can adaptively adjust the smoothing degree of different target areas. The normalized noise parameter is used to assign weights to improve the suppression effect on high noise areas. At the same time, combined with the exponential decay of the gradient amplitude, the edge and texture information can be retained to avoid the loss of details caused by over-smoothing.

[0031] The noise parameters are: , where For the In the video frame The standard deviation of the grayscale values ​​of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The standard deviation of the grayscale values ​​of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The noise characteristic value of the target area, For the first The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic values ​​of the target area, is the normalization function.

[0032] By calculating the variation of the standard deviation of the grayscale of the target area pixels and combining it with the deviation of the noise characteristic value, a noise parameter is constructed to measure the credibility of the target area. The introduction of the normalization function enhances the ability to distinguish the noise levels of different areas, making the low credibility area more obvious. At the same time, the comprehensive calculation using the neighborhood multi-frame information improves the robustness to random noise and local illumination changes.

[0033] The noise characteristic value is: , where For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean value of the gradient amplitude of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The mean value of the gradient direction of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean value of the gradient directions of all pixels in the target area.

[0034] By calculating the mean of the pixel gradient amplitude in the target area and combining the gradient direction changes of multiple frames in the neighborhood, a feature value to measure the impact of noise is constructed. By using the amplitude and direction differences of the gradient information, the noise feature can more accurately reflect the changes in the local image structure, while reducing the interference of abnormal fluctuations in a single frame on the results. The sensitivity to texture and edge features is enhanced, so that real targets and noise interference can be more effectively distinguished in complex backgrounds, thereby improving the stability of the video frame and the credibility of the target area.

[0035] In another embodiment, the noise parameter is: , where For the In the video frame The mean gray value of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean gray value of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The noise characteristic value of the target area, For the first The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic values ​​of the target area, is the normalization function.

[0036] By calculating the change degree of the grayscale mean of the target area and combining it with the deviation degree of the noise characteristic value, a noise parameter is constructed to measure the credibility of the target area. The introduction of the normalization function enhances the adaptive adjustment ability of the noise level, making the low credibility area more prominent. At the same time, the comprehensive calculation using the neighborhood multi-frame information improves the suppression effect of local illumination changes and random noise.

[0037] S3: Reconstruct the community video based on the video frames after noise removal.

[0038] In one embodiment, the reconstructing the community video based on each video frame after noise removal is specifically: Obtain each video frame and timestamp after noise removal, sort the video frames after noise removal in chronological order based on the timestamp, and reconstruct the community video based on the sorting result. By sorting and reconstructing the denoised video frames according to the timestamp, the temporal order and continuity of the video are ensured, and the interference of noise on the video quality is eliminated, making the video clearer and more stable. This method not only improves the realism and visualization of the video content, but also enhances the accuracy of subsequent analysis, especially in complex environments, and can more effectively identify target areas and abnormal situations.

[0039] The solution of the present invention significantly improves the denoising effect of video frames and enhances the clarity and stability of video content by introducing an improved Gaussian filtering algorithm and noise parameter calculation. During the processing, by accurately calculating the noise characteristic value and credibility of the target area, background noise and irrelevant information can be effectively filtered to ensure that important details in the video are more prominent. In addition, based on the timestamp sorting of the denoised video frames, the time order of the video can be restored, the smoothness and coherence of the video reconstruction can be improved, and the collection and processing process of community videos can be further optimized, solving the problem of edge information loss or noise residue when the existing Gaussian filtering algorithm processes noise.

[0040] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0041] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. The community video optimization acquisition method based on artificial intelligence is characterized by: include: Obtain each video frame of the community video, and divide any video frame into multiple target areas; Using an improved Gaussian filtering algorithm to remove noise from each video frame to obtain each video frame after noise removal; And reconstruct the community video based on each video frame after noise removal; Among them, the improved Gaussian filtering algorithm includes weight values, In the video frame The weight of the target area for, , where For the In the video frame The noise parameters of the target area, For the In the video frame The noise parameters of the target area, For the The total number of target areas in a video frame, The natural constant The exponential function with base , For the In the video frame The mean value of the gradient amplitude of all pixels in the target area; The noise parameter is related to In the video frame The noise characteristic value of the target area and the standard deviation of the grayscale values ​​of all pixels are positively correlated with the The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic value in the target area and the standard deviation of the grayscale values ​​of all pixels are inversely correlated.

2. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The noise parameters are: , where For the In the video frame The standard deviation of the grayscale values ​​of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The standard deviation of the grayscale values ​​of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The noise characteristic value of the target area, For the first The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic values ​​of the target area, is a normalized function, and the noise characteristic value represents the credibility of the target area.

3. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The noise parameters are: , where For the In the video frame The mean gray value of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean gray value of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The noise characteristic value of the target area, For the first The first video frame is used as the center to set the neighborhood radius of all video frames. The mean of the noise characteristic values ​​of the target area, is a normalized function, and the noise characteristic value represents the credibility of the target area.

4. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The neighborhood radius is set to 5.

5. The community video optimization acquisition method based on artificial intelligence according to claim 2 or 3 is characterized in that: The noise characteristic value is: , where For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean value of the gradient amplitude of all pixels in the target area, is the total number of video frames within the set neighborhood radius, For the In the video frame The mean value of the gradient direction of all pixels in the target area, For the first The video frame is taken as the center to set the neighborhood radius The first video frame The mean value of the gradient directions of all pixels in the target area.

6. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: Use a high frame rate camera to obtain each video frame of the community video.

7. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The method also includes performing grayscale processing on each of the video frames.

8. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The multiple target areas include: dividing any video frame into multiple target areas of the same size.

9. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The multiple target areas include: dividing any video frame into multiple target areas of any size.

10. The community video optimization acquisition method based on artificial intelligence according to claim 1 is characterized in that: The community video is reconstructed based on each video frame after noise removal, specifically: The video frames and timestamps after the noise removal process are obtained, the video frames after the noise removal process are sorted in chronological order based on the timestamps, and the community video is reconstructed based on the sorting results.

Images