A chest-mounted device with night shooting video function

Abstract

The present application relates to the technical field of image processing, in particular to a chest-worn device with night shooting and recording function; a scale adjustment parameter is obtained according to the gray level feature and the gradient difference feature of the pixel point; the adaptive scale and the illumination component are obtained by adjusting the filter kernel scale according to the scale adjustment parameter; the local similarity is obtained according to the gray level difference feature of the local area between the pixel point and other pixel points in the historical video frame; the corresponding pixel point and the maximum credibility are obtained according to the local similarity between the pixel point and other pixel points in the local area; the smoothing weight is obtained according to the video frame time interval feature between the pixel point and the corresponding pixel point and the maximum credibility; the illumination component smoothing value is obtained according to the smoothing weight, the illumination component smoothing value of the corresponding pixel point and the illumination component of the pixel point, and the current video frame is enhanced through the Retinex enhancement algorithm, thereby improving the video imaging quality and visual stability.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a chest-worn device with nighttime video recording capabilities. Background Technology

[0002] Chest-worn video recording devices have gradually become the main portable video recording devices due to their advantages of being easy to wear and small in size. However, in nighttime or low-light environments, these devices are easily affected by factors such as external lighting conditions, the movement of the wearer, and environmental changes, which can lead to uneven image brightness and noise interference, thus affecting the image quality at night. Therefore, it is necessary to improve the shortcomings of these devices in nighttime shooting and enhance the image quality in low-light environments.

[0003] While existing image enhancement algorithms based on the Retinex theory of the retina and cerebral cortex offer some solutions for low-light imaging, their application in chest-mounted shooting scenarios remains limited. When separating illuminance and reflectance components, these algorithms typically rely on fixed parameters or single-frame brightness distribution for estimation, lacking constraints on the temporal continuity between video frames. This leads to unstable calculation of the illumination component, and the enhancement results are prone to brightness jumps, amplification of noise in dark areas, and local overexposure, significantly reducing the visual stability and realism of the video. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a chest-worn device with nighttime video recording capabilities. The specific technical solution adopted is as follows:

[0005] The video acquisition module is used to acquire video frames from the captured content;

[0006] The first video frame analysis module is used to obtain scale adjustment parameters based on the grayscale features of pixels in the current video frame and the gradient difference features within the neighborhood range; and to adjust the filter kernel scale in the Retinex enhancement algorithm based on the pixel scale adjustment parameters to obtain the adaptive scale and illumination component of the pixel.

[0007] The second video frame analysis module is used to obtain the corresponding local region and scanned region based on the adaptive scale of the pixel; obtain local similarity based on the grayscale difference features of the local region between the pixel and other pixels in the scanned region of historical video frames; obtain matching confidence based on the local similarity between the pixel and other pixels in the local region; obtain the corresponding pixel and maximum confidence of the pixel in historical video frames based on the matching confidence; and obtain the smoothing weight of the corresponding pixel based on the video frame time interval features between the pixel and the corresponding pixel and the maximum confidence.

[0008] The video frame enhancement module is used to obtain the illumination component smoothing value of the pixel based on the smoothing weight and the illumination component smoothing value of the corresponding pixel, and the illumination component of the pixel; and to perform image enhancement on the current video frame using the Retinex enhancement algorithm based on the illumination component smoothing value of the pixel in the current video frame.

[0009] Furthermore, the step of obtaining the scale adjustment parameters based on the grayscale features of pixels in the current video frame and the gradient difference features within the neighborhood includes:

[0010] Calculate the ratio of the grayscale value of the pixel to the upper limit of grayscale to obtain a first value; calculate the gradient range of the pixel within the range corresponding to the preset minimum filter kernel scale and perform negative correlation mapping to obtain a second value; calculate the information entropy of the difference in gradient directions between any two adjacent pixels within the range corresponding to the preset minimum filter kernel scale to obtain a third value; calculate the product of the second value and the third value and normalize it to obtain a fourth value; calculate the sum of the first value and the fourth value and normalize it to obtain the scale adjustment parameter of the pixel.

[0011] Further, the step of adjusting the filter kernel scale in the Retinex enhancement algorithm according to the pixel scale adjustment parameters to obtain the adaptive scale and illumination component of the pixel includes:

[0012] Calculate the difference between the preset maximum filter kernel scale and the preset minimum filter kernel scale to obtain the scale difference; calculate the product of the scale difference and the scale adjustment parameter of the pixel to obtain the scale adjustment amount; calculate the sum of the scale adjustment amount and the preset minimum filter kernel scale to obtain the adaptive scale of the pixel; obtain the illumination component of the pixel using the Retinex enhancement algorithm based on the adaptive scale.

[0013] Furthermore, the step of obtaining the corresponding local region and scanned region based on the adaptive scale of the pixel includes:

[0014] The range of the filter kernel corresponding to the adaptive scale of the pixel is taken as the local region of the pixel; the scanning region of the pixel covers and is larger than the local region.

[0015] Further, the step of obtaining local similarity based on the grayscale difference features of the local region between the pixel and other pixels in the scanned area of ​​the historical video frame includes:

[0016] Starting from the right-side adjacent pixel of the pixel, traverse all pixels in the local area from the inside out in a counter-clockwise direction, and construct a grayscale value normalization sequence according to the traversal order; calculate the dynamic time-normalized distance of the grayscale value normalization sequence between the pixel and any other pixel in the scanning area of ​​the historical video frame and perform negative correlation mapping to obtain the local similarity between the pixel and any other pixel.

[0017] Further, the step of obtaining the matching confidence level based on the local similarity between the pixel and other pixels within the local region includes:

[0018] When calculating the local similarity between the pixel and any other pixel within the scanning range of the historical video frame, the average local similarity between all other pixels within the local region of the pixel and their relative positions in the same historical video frame is calculated to obtain the neighborhood similarity; the product of the local similarity and the neighborhood similarity is calculated to obtain the matching confidence between the pixel and any other pixel within the scanning range of the historical video frame.

[0019] Further, the step of obtaining the corresponding pixel in the historical video frame and the maximum confidence level based on the matching confidence level includes:

[0020] The pixel point corresponding to the maximum matching confidence value in each historical video frame is taken as the corresponding pixel point, and the matching confidence value of the corresponding pixel point is taken as the maximum confidence value.

[0021] Further, the step of obtaining the smoothing weight of the corresponding pixel based on the video frame time interval feature between the pixel and the corresponding pixel, and the maximum confidence level, includes:

[0022] In the formula, R represents the smoothing weight of the corresponding pixel, and N represents the number of corresponding pixels. This represents the time interval between the nth corresponding pixel and the video frame of that pixel. This represents the maximum confidence level of the nth corresponding pixel.

[0023] Further, the step of obtaining the smoothed value of the illumination component of the pixel based on the smoothing weight and the smoothed value of the illumination component of the corresponding pixel, and the illumination component of the pixel, includes:

[0024] Set the smoothing weight of the pixel to a constant of 1; calculate the proportion of the smoothing weight of the pixel and each corresponding pixel to obtain different final weights; calculate the product of the illumination component of the pixel and the corresponding final weight to obtain a fifth value; calculate the sum of the products of the illumination component smoothing value of the corresponding pixel and the corresponding final weight to obtain a sixth value; calculate the sum of the fifth value and the sixth value to obtain the illumination component smoothing value of the pixel.

[0025] The present invention has the following beneficial effects:

[0026] In this invention, obtaining the scale adjustment parameter adaptively adjusts the filter kernel scale in the Retinex enhancement algorithm based on the brightness and local texture features of the pixel, thereby obtaining a more accurate illumination component for the pixel and initially improving the enhancement effect of the pixel. Obtaining local similarity determines matching pixels based on the gray-level similarity features at the same physical location between video frames; obtaining matching confidence further improves the matching accuracy of pixels. Obtaining corresponding pixels and maximum confidence accurately smooths the illumination components of pixels in the current video frame; obtaining smoothing weights determines the contribution of different corresponding pixels to the illumination component smoothing, further improving smoothing accuracy. Obtaining the illumination component smoothing value of the pixel ensures that the illumination components of pixels in the video frame are stable in the time domain; finally, based on the illumination component smoothing value of the pixels in the current video frame, the Retinex enhancement algorithm is used to enhance the image of the current video frame, which not only improves the imaging brightness and detail performance under night shooting conditions, but also significantly reduces the brightness fluctuation and flicker problems between consecutive video frames, thereby improving the imaging clarity and visual stability of the chest-worn device in night recording. Attached Figure Description

[0027] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a block diagram of a chest-worn device module with nighttime video recording function, provided as an embodiment of the present invention. Detailed Implementation

[0029] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a chest-worn device with nighttime video recording function proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] The following description, in conjunction with the accompanying drawings, details a specific solution for a chest-worn device with nighttime video recording capabilities provided by the present invention.

[0032] Please see Figure 1 The diagram illustrates a block diagram of a chest-worn device module with nighttime video recording function according to an embodiment of the present invention. The device includes the following modules.

[0033] The video acquisition module S1 is used to acquire video frames from the captured content.

[0034] The system acquires video frames captured by a chest-worn video recording device in a nighttime environment, and then optimizes these frames to obtain video content with more stable brightness and better video continuity, thereby improving the device's imaging quality.

[0035] The first video frame analysis module S2 is used to obtain scale adjustment parameters based on the grayscale features of pixels in the current video frame and the gradient difference features within the neighborhood; and to adjust the filter kernel scale in the Retinex enhancement algorithm based on the pixel scale adjustment parameters to obtain the adaptive scale and illumination components of the pixels.

[0036] In nighttime or low-light shooting scenarios, ambient lighting often exhibits characteristics such as alternating intensity, interference from point light sources, and dynamic changes. The wearer's movement also causes frequent changes in perspective, resulting in significant fluctuations in brightness across video frames. Traditional image enhancement algorithms based on the Retinex theory typically employ a fixed smoothing scale during illumination estimation. This scale, representing the standard deviation of the filter kernel, applies the same filtering parameters to all pixels in the image, making it difficult to simultaneously preserve details in dark areas and suppress overexposure in bright areas. This can easily lead to problems such as edge blurring, halo effects, and uneven brightness. For areas with high brightness and gentle texture changes, a larger smoothing scale can be used due to the continuous and stable illumination distribution, effectively suppressing local bright spots and maintaining the naturalness of illumination transitions. Conversely, in areas with low brightness or large texture gradients, a smaller smoothing scale can better preserve local details and avoid edge blurring. This adaptive smoothing strategy achieves dynamic smoothing of illumination and detail preservation globally, making the calculation of illumination components for any pixel in any video frame more accurate, stable, and visually consistent. Therefore, the scale adjustment parameters are obtained based on the grayscale characteristics of pixels in the current video frame and the gradient difference characteristics within their neighborhood.

[0037] Preferably, in this embodiment of the invention, the step of obtaining the scale adjustment parameters includes: calculating the ratio of the gray value of a pixel to the upper limit of gray value to obtain a first value; the upper limit of gray value is 255. The larger the first value, the greater the brightness of the pixel, and the larger the smoothing scale of the pixel. Calculating the gradient range of the pixel within the range corresponding to the preset minimum filter kernel scale and negatively correlated mapping to obtain a second value; the larger the second value, the smaller the gradient range within that range, and the larger the smoothing scale required. Calculating the information entropy of the difference in gradient directions between any adjacent pixels within the range corresponding to the preset minimum filter kernel scale to obtain a third value; it should be noted that when calculating the difference in gradient directions, pixels with gradient values ​​not equal to 0 need to be selected. The larger the information entropy, the lower the probability of texture existing in the neighborhood of the pixel, and the larger the smoothing scale required. Calculating the product of the second and third values ​​and normalizing it to obtain a fourth value; this normalization is linear normalization. The larger the fourth value, the smoother the texture change, and the larger the smoothing scale required. The sum of the first and fourth values ​​is calculated and normalized to obtain the pixel's scale adjustment parameter. This normalization is performed using the sigmoid function. A larger scale adjustment parameter indicates higher brightness and smoother texture changes at the pixel, resulting in a larger smoothing scale; conversely, a smaller parameter indicates lower brightness or a larger texture gradient, resulting in a smaller smoothing scale. The formulas for obtaining the scale adjustment parameter include:

[0038]

[0039] In the formula, W represents the pixel scale adjustment parameter. This indicates normalization using the sigmoid function, where H represents the grayscale value of the pixel. Indicates the first value. denoted by linear normalization, Q represents the third value, e represents the natural constant, and Y represents the gradient range of the pixel within the range corresponding to the preset minimum filter kernel scale. Indicates the second value. This represents the fourth numerical value.

[0040] Furthermore, after obtaining the scale adjustment parameters of any pixel in the current video frame, the filter kernel scale in the Retinex enhancement algorithm can be adjusted according to the pixel scale adjustment parameters to obtain the adaptive scale and illumination component of the pixel. Preferably, in this embodiment of the invention, the steps of obtaining the adaptive scale and illumination component include: calculating the difference between a preset maximum filter kernel scale and a preset minimum filter kernel scale to obtain the scale difference; in this embodiment of the invention, the preset maximum filter kernel scale parameter is 10, and the preset minimum filter kernel scale parameter is 1. This range can preserve texture in dark areas and suppress overexposure in bright areas, achieving illumination balance and visual stability; the implementer can determine this according to the implementation scenario. Calculate the product of the scale difference and the scale adjustment parameter of the pixel to obtain the scale adjustment amount; the larger the scale adjustment parameter, the larger the scale adjustment amount. Calculate the sum of the scale adjustment amount and the preset minimum filter kernel scale to obtain the adaptive scale of the pixel. Obtain the illumination component of the pixel through the Retinex enhancement algorithm according to the adaptive scale. It should be noted that this image enhancement algorithm belongs to the prior art, and the specific steps will not be described in detail.

[0041] The second video frame analysis module S3 is used to obtain the corresponding local region and scanning region based on the adaptive scale of the pixel; to obtain the local similarity based on the grayscale difference features of the local region between the pixel and other pixels in the scanning region in the historical video frame; to obtain the matching confidence based on the local similarity between the pixel and other pixels in the local region; to obtain the corresponding pixel and maximum confidence in the historical video frame based on the matching confidence; and to obtain the smoothing weight of the corresponding pixel based on the video frame time interval features between the pixel and the corresponding pixel and the maximum confidence.

[0042] In continuously captured video frames, due to changes in ambient lighting, interference from point light sources, and the movement of the wearer and the subject, the pixel brightness corresponding to the same scene or object may fluctuate significantly in different video frames, resulting in flickering in dark areas, local overexposure, or uneven overall brightness. Therefore, it is necessary to smooth the illumination components that may correspond to the same object or pixel in the temporal domain. Smoothing the illumination components of each pixel in consecutive video frames suppresses abrupt fluctuations and maintains the continuity and stability of the enhanced video frame brightness. To identify objects that may correspond to the same pixel in consecutive video frames, the corresponding local region and scanning region are first obtained based on the pixel's adaptive scale. Preferably, in this embodiment, the steps of obtaining the local region and scanning region include: using the filter kernel range corresponding to the pixel's adaptive scale as the local region of the pixel; the scanning region of the pixel covers and is larger than the local region. In this embodiment, the area of ​​the scanning region is twice that of the local region. The purpose of the scanning region is to determine the corresponding pixel in historical video frames. The smaller the scanning region, the higher the computational efficiency; the implementer can determine this based on the implementation scenario. The adaptive scale reflects the brightness characteristics and texture complexity of the pixel. When the scale is large, it means that the texture changes gently and the details are weak, so a larger matching range is needed to improve the probability of correct matching. When the scale is small, the texture details are rich, so accurate matching can be achieved within a smaller range.

[0043] Furthermore, after obtaining the local region and scanning region of a pixel, pixel matching can be performed in different video frames. Local similarity is obtained based on the grayscale difference characteristics of the local region between the pixel and other pixels within the scanning region of the historical video frame. Preferably, in this embodiment, the step of obtaining local similarity includes: starting from the right-side adjacent pixel of the pixel, traversing all pixels within the local region counterclockwise from the inside out, and constructing a grayscale value normalization sequence according to the traversal order. If any pixel in a video frame corresponds to the same physical location as other pixels in a historical video frame, then the grayscale value change patterns of their neighborhoods have a certain similarity. The dynamic time-warped distance of the grayscale value normalization sequence between the pixel and any other pixel within the scanning region of the historical video frame is calculated and negatively correlated to obtain the local similarity between the pixel and any other pixel. In this embodiment, the range of the historical video frame is the previous second of the video frame where the pixel is located, which can be determined by the implementer according to the implementation scenario. The dynamic time-warped distance is obtained through existing dynamic time-warped algorithms; the more similar the changes of the two sequences, the smaller the dynamic time-warped distance. Therefore, the more similar the grayscale features of a pixel are to those of any other pixel in the scanning area of ​​a historical video frame, the greater the local similarity, and the more likely the two pixels represent the same physical location. It should be noted that the range of the grayscale value normalization sequence of any other pixel in the scanning area of ​​a historical video frame is consistent with the size of the local region of the pixel, and the scanning area in the historical video frame is consistent with the position of the scanning area of ​​the video frame in which the pixel is located.

[0044] Considering the stability of pixels on the surface of an object in its local structure, the normalized sequences of grayscale values ​​of other pixels at the same position within the local area of ​​the matched pixel also exhibit high consistency. Therefore, when the grayscale matching of the pixel itself and other pixels within the local area shows high similarity, it is possible to accurately determine that the two pixels represent the same physical location in different video frames. Furthermore, the matching confidence level is obtained based on the local similarity between the pixel and other pixels within the local area. Preferably, in this embodiment, the step of obtaining the matching confidence level includes: when calculating the local similarity between the pixel and any other pixel within the scanning range of a historical video frame, calculating the average of the local similarities between all other pixels within the local area of ​​the pixel and their relative positions in the same historical video frame to obtain the neighborhood similarity; the relative position is the position where the two matched pixels have the same position in the normalized sequence of grayscale values. The greater the neighborhood similarity, the greater the similarity between the grayscale features of other pixels within the local area of ​​the pixel and their relative positions in the historical video frame, and the more likely the two matched pixels represent the same physical location. The product of local similarity and neighborhood similarity is calculated to obtain the matching confidence of the pixel with any other pixel within the scanning range of the historical video frame. The higher the matching confidence, the more likely the pixel is to represent the same physical location as any other pixel within the scanning range of the historical video frame.

[0045] Furthermore, the corresponding pixel and maximum confidence of a pixel in historical video frames can be obtained based on the matching confidence. Specifically, this includes: taking the pixel corresponding to the maximum matching confidence in each historical video frame as the corresponding pixel, and taking the matching confidence of the corresponding pixel as the maximum confidence. After obtaining the corresponding pixel and maximum confidence of the pixel in each historical video frame, the weight of the illumination component smoothing value can be calculated; therefore, the smoothing weight of the corresponding pixel is obtained based on the video frame time interval characteristics between the pixel and the corresponding pixel, and the maximum confidence. Preferably, in this embodiment of the invention, the step of obtaining the smoothing weight includes:

[0046]

[0047] In the formula, R represents the smoothing weight of the corresponding pixel, and N represents the number of corresponding pixels. This represents the time interval between the nth corresponding pixel and the video frame containing that pixel. This represents the maximum confidence level of the nth corresponding pixel. The closer the time interval between the historical video frame containing the corresponding pixel and the current video frame, the greater the weight of the illumination component of that pixel in the smoothing process. A higher maximum confidence level indicates a higher confidence level for the illumination component of that pixel, and thus a greater weight for that illumination component in the smoothing process.

[0048] The video frame enhancement module S4 is used to obtain the illumination component smoothing value of a pixel based on the smoothing weight and the illumination component smoothing value of the corresponding pixel; and to perform image enhancement on the current video frame using the Retinex enhancement algorithm based on the illumination component smoothing value of the pixel in the current video frame.

[0049] After obtaining the smoothing weights of corresponding pixels in all historical video frames of the pixel, the smoothing value of the illumination component of the pixel can be obtained based on the smoothing weights, the smoothing value of the illumination component of the corresponding pixel, and the illumination component of the pixel. Preferably, in this embodiment of the invention, the step of obtaining the smoothing value of the illumination component includes: setting the smoothing weight of the pixel to a constant of 1, so that the illumination component weight of the pixel itself is the largest during the smoothing process; calculating the proportion of the smoothing weights of the pixel and each corresponding pixel to obtain different final weights; calculating the product of the illumination component of the pixel and the corresponding final weight to obtain a fifth value; calculating the sum of the products of the illumination component smoothing value of the corresponding pixel and the corresponding final weight to obtain a sixth value; the calculation steps of the smoothing value of the illumination component of the pixel in different video frames are the same and belong to a progressive process; calculating the sum of the fifth value and the sixth value to obtain the smoothing value of the illumination component of the pixel. The formula for obtaining the smoothing value of the illumination component of the pixel includes:

[0050]

[0051] In the formula, E represents the smoothing value of the illumination component of that pixel. This represents the final weight of the pixel, where D represents the illumination component of the pixel. This represents the fifth value, where N represents the number of corresponding pixels. This represents the final weight of the nth corresponding pixel. This represents the smoothing value of the illumination component of the nth corresponding pixel. This represents the sixth numerical value.

[0052] After obtaining the smoothed illumination component values ​​of each pixel in the current video frame, the illumination components of the video frame are stabilized in the time domain, effectively suppressing brightness jumps caused by changes in ambient light or wearer movement. Then, based on the smoothed illumination component values ​​of the pixels in the current video frame, the Retinex enhancement algorithm is used to enhance the image of the current video frame. The smoothed illumination component values ​​and the reflection component of the video frame image are separated; the reflection component of each pixel is calculated using a reflection retrieval model based on Retinex theory, thereby highlighting the inherent brightness and texture details of the scene. On this basis, adaptive dynamic range compression and color correction are applied to the reflection component to further improve visibility in dark areas and suppress overexposure in bright areas, achieving a natural balance of overall brightness and color in the video frame. Finally, the enhanced reflection component and the smoothed illumination component values ​​after time-domain smoothing are re-fused to generate a stable and visually consistent enhanced video frame. It should be noted that the Retinex enhancement algorithm is existing technology, and the specific steps are not described in detail here. This optimization method not only improves the imaging brightness and detail performance under nighttime shooting conditions but also significantly reduces brightness fluctuations and flickering in consecutive video frames, improving the imaging quality and visual stability of the device.

[0053] In summary, this invention provides a chest-worn device with nighttime video recording capabilities; it obtains scale adjustment parameters based on the grayscale features of pixels and gradient difference features within their neighborhood; it adjusts the filter kernel scale based on the pixel scale adjustment parameters to obtain adaptive scale and illumination components; it obtains local similarity based on the grayscale difference features of local regions between a pixel and other pixels in historical video frames; it obtains the corresponding pixel and maximum confidence based on the local similarity between a pixel and other pixels in the local region; it obtains smoothing weights based on the video frame time interval features between a pixel and its corresponding pixel and the maximum confidence; and it obtains illumination component smoothing values ​​based on the smoothing weights, the illumination component smoothing values ​​of the corresponding pixels, and the illumination components of the pixels, and enhances the current video frame using the Retinex enhancement algorithm, thereby improving video imaging quality and visual stability.

[0054] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0055] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A chest-worn device with nighttime video recording function, characterized in that, The device comprises the following modules: a video acquisition module, configured to acquire a video frame in a shooting content; a first video frame analysis module, configured to obtain a scale adjustment parameter according to a gray scale feature of a pixel point in a current video frame and a gradient difference feature in a neighborhood range; adjust a filter kernel scale in a Retinex enhancement algorithm according to the scale adjustment parameter of the pixel point, and obtain an adaptive scale and an illumination component of the pixel point; a second video frame analysis module, configured to obtain a corresponding local region and a scanning region according to the adaptive scale of the pixel point; obtain a local similarity according to a gray scale difference feature of the local region between the pixel point and other pixel points in a scanning region in a historical video frame; obtain a matching confidence according to the local similarity between the pixel point and other pixel points in the local region; obtain a corresponding pixel point and a maximum confidence of the pixel point in the historical video frame according to the matching confidence; obtain a smoothing weight of the corresponding pixel point according to a video frame time interval feature between the pixel point and the corresponding pixel point and the maximum confidence; a video frame enhancement module, configured to obtain an illumination component smoothing value of the pixel point according to the smoothing weight, the illumination component smoothing value of the corresponding pixel point and the illumination component of the pixel point; and perform image enhancement on the current video frame by the Retinex enhancement algorithm according to the illumination component smoothing value of the pixel point in the current video frame; the step of obtaining the scale adjustment parameter according to the gray scale feature of the pixel point in the current video frame and the gradient difference feature in the neighborhood range comprises: calculating a ratio of the gray scale value of the pixel point to an upper limit value of the gray scale to obtain a first value; calculating a gradient range value of the pixel point in a range corresponding to a preset minimum filter kernel scale and performing negative correlation mapping to obtain a second value; calculating an information entropy of a difference value of gradient directions of any adjacent pixel points of the pixel point in the range corresponding to the preset minimum filter kernel scale to obtain a third value; calculating a product of the second value and the third value and normalizing to obtain a fourth value; and calculating a sum value of the first value and the fourth value and normalizing to obtain the scale adjustment parameter of the pixel point; the step of obtaining the corresponding local region and the scanning region according to the adaptive scale of the pixel point comprises: taking a filter kernel range corresponding to the adaptive scale of the pixel point as a local region of the pixel point; and the scanning region of the pixel point covers and is larger than the local region; the step of obtaining the local similarity according to the gray scale difference feature of the local region between the pixel point and other pixel points in the scanning region in the historical video frame comprises: starting from a right adjacent pixel point of the pixel point, traversing all pixel points in the local region from inside to outside in a counterclockwise direction, and constructing a gray scale value normalization sequence according to the traversal order; calculating a dynamic time warping distance of the gray scale value normalization sequence between the pixel point and any other pixel point in the scanning region in the historical video frame and performing negative correlation mapping to obtain a local similarity of the pixel point and any other pixel point; and The step of obtaining the smoothing weight of the corresponding pixel point according to the video frame time interval feature between the pixel point and the corresponding pixel point and the maximum confidence comprises: wherein R represents a smoothing weight of the corresponding pixel point, N represents a number of the corresponding pixel points, represents a time interval of the nth corresponding pixel point and a video frame of the pixel point, represents a maximum credibility of the nth corresponding pixel point; The step of obtaining the illumination component smoothing value of the pixel point according to the smoothing weight and the illumination component smoothing value of the corresponding pixel point and the illumination component of the pixel point comprises: The smoothing weight of the pixel point is set as a constant 1; the proportion of the smoothing weight of the pixel point and each corresponding pixel point is calculated to obtain different maximum weights; the product of the illumination component of the pixel point and the corresponding maximum weight is calculated to obtain a fifth value; the sum of the product of the illumination component smoothing value of the corresponding pixel point and the corresponding maximum weight is calculated to obtain a sixth value; the sum of the fifth value and the sixth value is calculated to obtain the illumination component smoothing value of the pixel point.

2. The chest-worn device having a night shooting video function according to claim 1, characterized in that, The step of adjusting the filter kernel scale in the Retinex enhancement algorithm according to the scale adjustment parameter of the pixel point to obtain the adaptive scale and the illumination component of the pixel point comprises: The difference between the preset maximum filter kernel scale and the preset minimum filter kernel scale is calculated to obtain a scale difference; the product of the scale difference and the scale adjustment parameter of the pixel point is calculated to obtain a scale adjustment amount; the sum of the scale adjustment amount and the preset minimum filter kernel scale is calculated to obtain the adaptive scale of the pixel point; the illumination component of the pixel point is obtained by the Retinex enhancement algorithm according to the adaptive scale.

3. The chest-worn device having a night shooting video function according to claim 1, characterized in that, The step of obtaining the matching confidence according to the local similarity of the pixel point and other pixel points in the local region comprises: When calculating the local similarity of the pixel point and any other pixel point in the scanning range in the historical video frame, the average value of the local similarity of all other pixel points in the local region of the pixel point and the relative position in the same historical video frame is calculated to obtain a neighborhood similarity; the product of the local similarity and the neighborhood similarity is calculated to obtain the matching confidence of the pixel point and any other pixel point in the scanning range in the historical video frame.

4. The chest-worn device having a night shooting video function according to claim 1, characterized in that, The step of obtaining the corresponding pixel point and the maximum confidence of the pixel point in the historical video frame according to the matching confidence comprises: The pixel point corresponding to the maximum value of the matching confidence in each historical video frame is taken as the corresponding pixel point, and the matching confidence of the corresponding pixel point is taken as the maximum confidence.

Images