Computer image processing system based on deep learning

Abstract

The present application relates to the technical field of image processing, and especially relates to a computer image processing system based on deep learning, which determines the first contour boundary of a target object according to a deep learning model through a preprocessing module, draws the brightness performance curves of the pixel points of each sub-region in different direction dimensions through a feature extraction module, marks the non-confidence sub-region in the aggregation analysis region through a feature aggregation module, obtains the brightness performance curves of the pixel points after adjusting the exposure in different direction dimensions through a local processing module to determine the contour boundary fitting points, fits the contour boundary section of the non-confidence sub-region, updates the first contour boundary through an output end and outputs the second contour boundary of the target object. The present application performs local region optimization for the local feature deficiency in the contour boundary extraction process of the target object under complex lighting conditions, and realizes high-precision extraction and dynamic updating of the target object contour in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more particularly to a computer image processing system based on deep learning. Background Technology

[0002] In the field of computer image processing, accurate extraction of object contour boundaries is fundamental to tasks such as image analysis, object recognition, and scene understanding. Traditional image processing methods often rely on fixed thresholds and edge detection operators to achieve contour extraction. However, such methods are prone to problems such as contour breakage, false detection, or loss of details in scenarios with low contrast between the object and the background, blurred texture features, or noise interference. In complex situations such as areas with abrupt changes in brightness under non-uniform illumination or where the edge of the object is similar to the background features, contour extraction models based on a single global feature still struggle to accurately capture boundary details. In existing technologies, deep learning models lack fine-grained analysis of local brightness features, especially when dealing with the problem of blurred contour boundaries caused by exposure differences. They lack targeted optimization mechanisms for unconfidence regions. Therefore, there is an urgent need for an image processing system that combines deep learning models with local brightness feature analysis to solve the technical problem of insufficient accuracy in extracting object contour boundaries under complex lighting conditions.

[0003] For example, Chinese Patent Publication No. CN119338745A discloses a computer image processing system based on intelligent recognition technology, comprising: a real-time monitoring module that uses real-time image quality assessment technology to monitor and learn the acquired image data information; an image recognition module equipped with an alarm system that triggers the alarm system based on changes in the acquired image data information; an image optimization module that receives the alarm signal, locks the damaged position of the image, extracts image information before and after the damage, and uses a neural network-based degraded image restoration algorithm to deeply learn the image information; using the neural network-based degraded image restoration algorithm to deeply learn the image information, extracts clear original information, and performs blurring and noise reduction processing to effectively fill in the damaged position of the image.

[0004] The following problems still exist in the existing technology:

[0005] Existing technologies cannot optimize local regions to address the problem of insufficient local features in the extraction process of target object contour boundaries under complex lighting conditions, thus affecting the accuracy of contour extraction. Summary of the Invention

[0006] To address this issue, the present invention provides a computer image processing system based on deep learning, which overcomes the problem that existing technologies cannot optimize local regions due to insufficient local features in the process of extracting the contour boundaries of targets under complex lighting conditions.

[0007] To achieve the above objectives, the present invention provides a deep learning-based computer image processing system, comprising:

[0008] The preprocessing module is used to divide the reference image into several rectangular sub-regions, determine the first contour boundary of the target object according to the deep learning model, and mark the location information of the sub-regions where contour boundary segments exist.

[0009] The feature extraction module, which is connected to the preprocessing module, is used to obtain the brightness of the pixels in each sub-region in order to draw the brightness performance curves of the pixels in each sub-region in different directional dimensions.

[0010] The feature aggregation module is connected to the preprocessing module and the feature extraction module respectively, and is used to determine the aggregation analysis area based on the location information matching conditions, and to mark the non-confidence sub-regions according to the comparison of the brightness performance curves of each sub-region within the aggregation analysis area.

[0011] A local processing module, connected to the feature aggregation module, is used to obtain the brightness performance curves of the pixels in the non-confidence sub-region after adjusting the exposure in different directional dimensions, and to determine the contour boundary fitting point based on the comparison of the brightness performance curves after adjusting the exposure, so as to fit the contour boundary segment of the non-confidence sub-region.

[0012] The output end is connected to the local processing module and the preprocessing module respectively, and is used to update the first contour boundary using the contour boundary segment of the non-confidence sub-region, so as to output the second contour boundary of the target object in the reference image.

[0013] Furthermore, the feature extraction module is used to calculate the average brightness of pixels in each sub-region across different directional dimensions, wherein,

[0014] The feature extraction module is used to use the width data of each sub-region as the horizontal axis of the coordinate system to calculate the average brightness of the pixels in the sub-region based on each width value point.

[0015] And the length data of each sub-region is used as the horizontal axis of the coordinate system to calculate the average brightness of the pixels within the sub-region based on each length value point.

[0016] Furthermore, the feature extraction module is used to draw brightness performance curves in different directional dimensions, wherein,

[0017] The feature extraction module is used to draw a first brightness performance curve in a first direction dimension based on the average brightness of the pixels determined by each width value point, and to draw a second brightness performance curve in a second direction dimension based on the average brightness of the pixels determined by each length value point.

[0018] The horizontal axis of the first brightness performance curve represents the width of the sub-region, and the vertical axis represents the average brightness of the pixels. The horizontal axis of the second brightness performance curve represents the length of the sub-region, and the vertical axis represents the average brightness of the pixels.

[0019] Furthermore, the feature aggregation module is used to determine the aggregation analysis region based on location information matching conditions, wherein,

[0020] The feature aggregation module is used to obtain the location information of each sub-region with contour boundary segments, including a row sequence determined along the length direction of the reference image and a column sequence determined along the width direction of the reference image based on the location information.

[0021] The feature aggregation module is used to determine the region composed of several sub-regions that meet the location information matching conditions as the aggregation analysis region;

[0022] The aggregation analysis area includes at least two sub-regions, and the location information matching condition is that the row sequence or column sequence of each sub-region is consistent.

[0023] Furthermore, the feature aggregation module is used to determine the comparison of brightness performance curves for each sub-region, wherein,

[0024] The feature aggregation module is used to determine the maximum difference in brightness based on the brightness performance curves of each sub-region within the aggregation analysis area, and to calculate the standard deviation of the maximum difference in brightness among the sub-regions within the aggregation analysis area.

[0025] Furthermore, the feature aggregation module is used to mark non-confidence sub-regions, wherein,

[0026] The feature aggregation module is used to filter the aggregation analysis region corresponding to the maximum value of the standard deviation, and to mark the sub-regions corresponding to the maximum and minimum values ​​of the maximum difference in brightness within the corresponding aggregation analysis region as non-confidence sub-regions.

[0027] Furthermore, the local processing module is used to obtain the brightness performance curves of the pixels in the non-confidence sub-region after exposure adjustment in different directional dimensions, wherein,

[0028] The local processing module is used to draw a third brightness performance curve in the first direction dimension based on the average brightness of the pixels at each width value point of the non-confidence sub-region after exposure adjustment, and to draw a fourth brightness performance curve in the second direction dimension based on the average brightness of the pixels at each length value point of the non-confidence sub-region after exposure adjustment.

[0029] The exposure of the non-confidence sub-region is higher after adjustment than the exposure before adjustment.

[0030] Furthermore, the local processing module is used to determine the contour boundary fitting points, wherein,

[0031] The local processing module is used to compare the average brightness of the third brightness performance curve with the average brightness of the first brightness performance curve on the vertical axis to determine the width of the maximum difference in the average brightness on the horizontal axis, and to compare the average brightness of the fourth brightness performance curve with the average brightness of the second brightness performance curve on the vertical axis to determine the length of the maximum difference in the average brightness on the horizontal axis.

[0032] The local processing module is used to determine the contour boundary fitting point in the non-confidence sub-region based on the width value and the length value.

[0033] Furthermore, the local processing module is used to fit the contour boundary segment of the non-confidence sub-region, wherein,

[0034] The local processing module is used to obtain the initial contour boundary segment of the non-confidence sub-region, and to fit the initial contour boundary segment based on the contour boundary fitting point, so that the fitted contour boundary segment passes through the contour boundary fitting point.

[0035] Furthermore, the output terminal is used to update the first contour boundary using the fitted contour boundary segment, wherein,

[0036] The output terminal is used to replace the initial contour boundary segment with the fitted contour boundary segment, so as to update the first contour boundary to the second contour boundary.

[0037] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention uses a preprocessing module to determine the first contour boundary of the target object based on a deep learning model and marks the position information of sub-regions containing contour boundary segments. A feature extraction module draws the brightness performance curves of pixels in each sub-region in different directional dimensions. A feature aggregation module marks unconfident sub-regions within the aggregation analysis area. A local processing module obtains the brightness performance curves of pixels in unconfident sub-regions after exposure adjustment in different directional dimensions. The contour boundary segments of unconfident sub-regions are fitted based on contour boundary fitting points. Finally, the first contour boundary is updated and the second contour boundary of the target object is output through the output terminal. This invention optimizes local regions to address the lack of local features in the target object contour boundary extraction process under complex lighting conditions, achieving high-precision extraction and dynamic updating of the target object contour under complex lighting conditions.

[0038] Furthermore, this invention divides the reference image into several rectangular sub-regions, and performs brightness statistics along both the width and length directions for each sub-region. In the width direction, the average brightness of each column of pixels is calculated; in the length direction, the average brightness of each row of pixels is calculated. This two-dimensional analysis can capture the brightness variation patterns in different directions within the sub-regions. A first brightness performance curve reflects the brightness distribution characteristics in the width direction, and a second brightness performance curve reflects the brightness distribution characteristics in the length direction.

[0039] Furthermore, based on the phenomenon that the contour boundary of a target object is usually composed of adjacent or continuously arranged sub-regions, the present invention defines the aggregation analysis region by means of a feature aggregation module according to the matching conditions of row consistency or column consistency: if multiple sub-regions have the same row sequence or the same column sequence, they are divided into an aggregation analysis region. This strategy of aggregating and analyzing sub-regions in the same row or column can effectively capture the linear extension features of the contour boundary and improve the continuity of the contour boundary analysis.

[0040] Furthermore, this invention calculates the maximum difference in brightness along the corresponding dimension for the brightness performance curve of each sub-region within the aggregation analysis area using a feature aggregation module. This difference reflects the severity of brightness changes within the sub-region. By calculating the standard deviation of the maximum difference in brightness among all sub-regions within the aggregation area, the larger the standard deviation, the more significant the difference in brightness changes among the sub-regions within the aggregation analysis area. This indicates that the brightness characteristics of some sub-regions at the contour boundary of the same row or column do not conform to the overall trend. The aggregation analysis area with the largest standard deviation among all aggregation areas is selected. The sub-region brightness feature consistency within this aggregation analysis area is the worst, which may include contour blurring caused by complex lighting, resulting in abnormal contour boundary segments. By quantifying feature consistency through standard deviation and combining it with extreme value localization of abnormal sub-regions, reliable boundary segments and unreliable boundary segments can be effectively distinguished in complex lighting scenarios, improving the robustness of overall contour extraction.

[0041] Furthermore, this invention enhances the exposure of marked, unconfident sub-regions, amplifying the pixel brightness differences that were previously masked by underexposure. In particular, the brightness gradient between the target and the background becomes more pronounced. After the exposure is enhanced, the module recalculates the average pixel brightness at each coordinate point in the width and length directions of the sub-region, and plots the third and fourth brightness performance curves respectively. Since the increased exposure causes the pixel values ​​in the dark areas to shift towards the high brightness range, the brightness abrupt changes at the originally blurred boundaries become more significant, thus highlighting boundary features through local exposure enhancement.

[0042] Furthermore, this invention generates a first brightness performance curve and a second brightness performance curve for the non-confidence sub-region before exposure adjustment, and a third and fourth brightness performance curve after exposure adjustment. The local processing module compares the curves before and after adjustment on the vertical axis to find the position with the largest difference between them: in the width direction, the horizontal axis width value corresponding to the maximum difference between the average brightness of the third curve and the first curve reflects the position where the brightness change is most significant after exposure enhancement in that direction, which usually corresponds to the boundary between the target object and the background; similarly, the maximum difference between the fourth curve and the second curve in the length direction corresponds to the length value. The point determined by the intersection of the determined width and length values ​​in the sub-region coordinate system is the contour boundary fitting point; those skilled in the art should understand that exposure adjustment amplifies the brightness difference between the target object and the background, and this difference change at the boundary is often more drastic than in non-boundary areas. By capturing the extreme difference position in the curve comparison, the boundary point that was originally blurred due to lighting problems can be accurately located. This invention utilizes the complementarity of information before and after exposure adjustment, avoids the limitations of contour feature extraction, and realizes local area optimization for insufficient local features. Attached Figure Description

[0043] Figure 1 This is a block diagram of a deep learning-based computer image processing system according to an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the sub-region distribution in an embodiment of the present invention;

[0045] Figure 3 The flowchart below shows the logic of the feature aggregation module determining the aggregation analysis region in an embodiment of the present invention. Detailed Implementation

[0046] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0047] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0048] It should be noted that in the description of this invention, the terms "upper," "lower," "inner," "outer," etc., which indicate the direction or positional relationship, are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0049] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation" and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0050] Please see Figure 1 The diagram shown is a block diagram of a deep learning-based computer image processing system according to an embodiment of the present invention. The deep learning-based computer image processing system of the present invention includes:

[0051] The preprocessing module is used to divide the reference image into several rectangular sub-regions, determine the first contour boundary of the target object according to the deep learning model, and mark the location information of the sub-regions where contour boundary segments exist.

[0052] The reference image used in this invention is an image that has been cropped and denoised. Cropping and denoising are common operations in image processing, and will not be described in detail here.

[0053] Specifically, the preprocessing module in this invention can be a processor that integrates image segmentation and contour extraction functions, used to divide the reference image into several rectangular sub-regions of a preset size using a grid division method. Preferably, the length of the rectangular sub-region is 1 / 20 of the length of the reference image, and the width of the rectangular sub-region is 1 / 20 of the width of the reference image.

[0054] Specifically, the preprocessing module determines the row sequence of each sub-region along the length direction of the reference image, determines the column sequence of each sub-region along the width direction of the reference image, and determines the position information of the sub-regions with contour boundary segments based on the row sequence and column sequence.

[0055] Specifically, the deep learning model in the preprocessing module is pre-trained and used to identify the first contour boundary of the target object. Those skilled in the art will understand that the training process of the deep learning model can be as follows: Multiple training samples are pre-acquired, each training sample including information about an image and a label corresponding to that training sample. The label corresponding to the training sample refers to the first contour boundary of the target object in the image. The model processor trains the initialized learning model based on the multiple training samples and the labels corresponding to each training sample to obtain the deep learning model. The deep learning model may include convolutional layers, non-linear activation layers, normalization layers, and deconvolutional layers, etc. There are no restrictions on the construction layers of the deep learning model. The construction layers of the deep learning model in this invention are all constructed based on the above method, and the construction layers of the deep learning model will not be described here.

[0056] Preferably, each training sample includes individual pixels of a reference image under different lighting conditions. The reference image information includes an RGB image or a grayscale image. If the image is an RGB image, the pixel values ​​can be the pixel values ​​of each pixel in the R, G, and B channels of the RGB image. If the image is a grayscale image, the pixel values ​​are the grayscale values ​​of each pixel in the grayscale image. The training process of the deep learning model involves training an initialized learning model based on the individual pixels and pixel values ​​of the acquired reference image and the result of recognizing the first contour of the target object in the image, thus obtaining the deep learning model.

[0057] The feature extraction module, which is connected to the preprocessing module, is used to obtain the brightness of the pixels in each sub-region in order to draw the brightness performance curves of the pixels in each sub-region in different directional dimensions.

[0058] Specifically, the feature extraction module in this invention can be a processor that integrates pixel brightness analysis and curve construction functions. In this invention, the brightness of a pixel can be achieved by converting RGB to grayscale values, which is existing technology and will not be elaborated here.

[0059] The feature aggregation module is connected to the preprocessing module and the feature extraction module respectively, and is used to determine the aggregation analysis area based on the location information matching conditions, and to mark the non-confidence sub-regions according to the comparison of the brightness performance curves of each sub-region within the aggregation analysis area.

[0060] Specifically, the feature aggregation module in this invention can be an image processor used to perform region marking processing on the image, which is existing technology and will not be described in detail here.

[0061] A local processing module, connected to the feature aggregation module, is used to obtain the brightness performance curves of the pixels in the non-confidence sub-region after adjusting the exposure in different directional dimensions, and to determine the contour boundary fitting point based on the comparison of the brightness performance curves after adjusting the exposure, so as to fit the contour boundary segment of the non-confidence sub-region.

[0062] Specifically, the local processing module in this invention is a processor with curve fitting function. It determines discrete points on the contour based on the known contour and refits the contour boundary with the newly determined contour boundary fitting points. This can be achieved by existing image processing algorithms, and will not be elaborated here.

[0063] The output end is connected to the local processing module and the preprocessing module respectively, and is used to update the first contour boundary using the contour boundary segment of the non-confidence sub-region, so as to output the second contour boundary of the target object in the reference image.

[0064] Specifically, the output terminal in this invention can be a display, which receives the signal after image contour processing and displays the processed image contour. This will not be elaborated further.

[0065] Specifically, the feature extraction module is used to calculate the average brightness of pixels in each sub-region across different directional dimensions, wherein,

[0066] The feature extraction module is used to use the width data of each sub-region as the horizontal axis of the coordinate system to calculate the average brightness of the pixels in the sub-region based on each width value point.

[0067] And the length data of each sub-region is used as the horizontal axis of the coordinate system to calculate the average brightness of the pixels within the sub-region based on each length value point.

[0068] Specifically, the feature extraction module is used to draw brightness performance curves in different directional dimensions, wherein,

[0069] The feature extraction module is used to draw a first brightness performance curve in a first direction dimension based on the average brightness of the pixels determined by each width value point, and to draw a second brightness performance curve in a second direction dimension based on the average brightness of the pixels determined by each length value point.

[0070] The horizontal axis of the first brightness performance curve represents the width of the sub-region, and the vertical axis represents the average brightness of the pixels. The horizontal axis of the second brightness performance curve represents the length of the sub-region, and the vertical axis represents the average brightness of the pixels.

[0071] Please see Figure 2As shown, this is a schematic diagram of the sub-region distribution in an embodiment of the present invention. For example, the reference image can be divided into 20 columns along the width direction and 20 rows along the length direction. For the sub-region of the 6th column along the width direction, the average brightness of the pixels in the 20 sub-regions of the 6th column is calculated, and the calculated average brightness is used as the average brightness of the point with the width value of the 6th column. Similarly, for the sub-region of the 13th row along the length direction, the average brightness of the pixels in the 20 sub-regions of the 13th row is calculated, and the calculated average brightness is used as the average brightness of the point with the length value of the 13th row.

[0072] Understandably, this invention divides the reference image into several rectangular sub-regions, and performs brightness statistics on each sub-region along both the width and length directions. In the width direction, the average brightness of each column of pixels is calculated; in the length direction, the average brightness of each row of pixels is calculated. This two-dimensional analysis can capture the brightness variation patterns in different directions within the sub-regions. A first brightness performance curve reflects the brightness distribution characteristics in the width direction, and a second brightness performance curve reflects the brightness distribution characteristics in the length direction.

[0073] Specifically, please refer to Figure 3 The diagram shown is a logical flowchart of the feature aggregation module determining the aggregation analysis region according to an embodiment of the present invention. The feature aggregation module is used to determine the aggregation analysis region based on location information matching conditions.

[0074] The feature aggregation module is used to obtain the location information of each sub-region with contour boundary segments, including a row sequence determined along the length direction of the reference image and a column sequence determined along the width direction of the reference image based on the location information.

[0075] The feature aggregation module is used to determine the region composed of several sub-regions that meet the location information matching conditions as the aggregation analysis region, and to not analyze the sub-regions that do not meet the location information matching conditions.

[0076] The aggregation analysis area includes at least two sub-regions, and the location information matching condition is that the row sequence or column sequence of each sub-region is consistent.

[0077] It is understandable that the outline boundary of a target object is usually composed of adjacent or consecutively arranged sub-regions. The feature aggregation module defines the aggregation analysis region based on the matching conditions of row consistency or column consistency: if multiple sub-regions have the same row sequence or column sequence, they are divided into an aggregation analysis region. This strategy of aggregating and analyzing sub-regions in the same row or column can effectively capture the linear extension features of the outline boundary and improve the continuity of the outline boundary analysis.

[0078] Specifically, the feature aggregation module is used to determine the comparison of brightness performance curves for each sub-region, wherein,

[0079] The feature aggregation module is used to determine the maximum difference in brightness based on the brightness performance curves of each sub-region within the aggregation analysis area, and to calculate the standard deviation of the maximum difference in brightness among the sub-regions within the aggregation analysis area.

[0080] Specifically, the feature aggregation module is used to mark non-confidence sub-regions, wherein,

[0081] The feature aggregation module is used to filter the aggregation analysis region corresponding to the maximum value of the standard deviation, and to mark the sub-regions corresponding to the maximum and minimum values ​​of the maximum difference in brightness within the corresponding aggregation analysis region as non-confidence sub-regions.

[0082] Understandably, this invention uses a feature aggregation module to calculate the maximum difference in brightness across the corresponding dimension for the brightness performance curve of each sub-region within the aggregation analysis area. This difference reflects the severity of brightness changes within the sub-region. By calculating the standard deviation of the maximum difference in brightness across all sub-regions within the aggregation area, the larger the standard deviation, the more significant the difference in brightness changes among the sub-regions within the aggregation analysis area. This means that the brightness characteristics of some sub-regions at the contour boundary of the same row or column do not match the overall trend. The aggregation analysis area with the largest standard deviation among all aggregation areas is selected. The sub-regions within this aggregation analysis area have the worst consistency in brightness characteristics and may contain contour blurring caused by complex lighting, resulting in abnormal contour boundary segments. By quantifying feature consistency through standard deviation and combining it with extreme value localization of abnormal sub-regions, reliable and unreliable boundary segments can be effectively distinguished in complex lighting scenarios, improving the robustness of overall contour extraction.

[0083] Specifically, the local processing module is used to obtain the brightness performance curves of the pixels in the non-confidence sub-region after exposure adjustment in different directional dimensions, wherein,

[0084] The local processing module is used to draw a third brightness performance curve in the first direction dimension based on the average brightness of the pixels at each width value point of the non-confidence sub-region after exposure adjustment, and to draw a fourth brightness performance curve in the second direction dimension based on the average brightness of the pixels at each length value point of the non-confidence sub-region after exposure adjustment.

[0085] The exposure of the non-confidence sub-region is higher after adjustment than the exposure before adjustment.

[0086] In this invention, those skilled in the art can adjust the exposure of the non-confidence sub-region of the reference image according to processing requirements. It is understood that adjusting the exposure will amplify the difference between the target object boundary and the image background, but overexposure will also cause the target object outline details to be lost. Based on this, in this invention, the exposure of the non-confidence sub-region can be adjusted to [103%, 110%] of the original exposure. Preferably, the exposure of the non-confidence sub-region can be adjusted to 105% of the original exposure.

[0087] Understandably, by increasing the exposure of the marked unconfidence sub-regions, the pixel brightness differences that were originally masked by underexposure in the region are amplified. In particular, the brightness gradient change between the target object and the background becomes more obvious. After the exposure is enhanced, the module recalculates the average pixel brightness of each coordinate point in the width and length directions of the sub-region and draws the third and fourth brightness performance curves respectively. Since the increase in exposure will cause the pixel values ​​in the dark area to shift to the high brightness range, the brightness change features at the originally blurred boundary will be more significant, thus achieving the goal of highlighting boundary features through local exposure enhancement.

[0088] Specifically, the local processing module is used to determine the contour boundary fitting points, wherein,

[0089] The local processing module is used to compare the average brightness of the third brightness performance curve with the average brightness of the first brightness performance curve on the vertical axis to determine the width of the maximum difference in the average brightness on the horizontal axis, and to compare the average brightness of the fourth brightness performance curve with the average brightness of the second brightness performance curve on the vertical axis to determine the length of the maximum difference in the average brightness on the horizontal axis.

[0090] The local processing module is used to determine the contour boundary fitting point in the non-confidence sub-region based on the width value and the length value.

[0091] Understandably, this invention generates a first and second brightness performance curve for non-confidence sub-regions before exposure adjustment, and a third and fourth brightness performance curve after exposure adjustment. The local processing module compares the curves before and after adjustment on the vertical axis to find the position with the largest difference: in the width direction, the horizontal axis width value corresponding to the maximum difference between the average brightness of the third curve and the first curve reflects the position where the brightness change is most significant after exposure enhancement in that direction, which usually corresponds to the boundary between the target object and the background; similarly, the maximum difference between the fourth curve and the second curve in the length direction corresponds to the length value. The point determined by the intersection of the determined width and length values ​​in the sub-region coordinate system is the contour boundary fitting point; those skilled in the art should understand that exposure adjustment amplifies the brightness difference between the target object and the background, and this difference change at the boundary is often more drastic than in non-boundary areas. By capturing the extreme difference positions in the curve comparison, the boundary points that were originally blurred due to lighting issues can be accurately located. This invention utilizes the complementarity of information before and after exposure adjustment, avoids the limitations of contour feature extraction, and realizes local area optimization for insufficient local features.

[0092] Specifically, the local processing module is used to fit the contour boundary segment of the non-confidence sub-region, wherein,

[0093] The local processing module is used to obtain the initial contour boundary segment of the non-confidence sub-region, and to fit the initial contour boundary segment based on the contour boundary fitting point, so that the fitted contour boundary segment passes through the contour boundary fitting point.

[0094] In this invention, the initial contour boundary segment of the non-confidence sub-region obtained by the preprocessing module can be uniformly sampled in advance to generate a discrete point set. The determined contour boundary fitting points are added to the discrete point set. Then, Gaussian filtering is performed to smooth the fitting based on the points in the discrete point set to obtain the fitted contour boundary segment. The fitting curve based on the discrete points is an existing technology and will not be described in detail here.

[0095] Specifically, the output terminal is used to update the first contour boundary using the fitted contour boundary segment, wherein,

[0096] The output terminal is used to replace the initial contour boundary segment with the fitted contour boundary segment, so as to update the first contour boundary to the second contour boundary.

[0097] Understandably, compared to the first contour boundary, the contour boundary segments within the non-confidence sub-region of the second contour boundary have been refitted and replaced, and the second contour boundary is the final contour boundary of the target object determined from the reference image.

[0098] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A deep learning based computer image processing system, characterized by, The method comprises the following steps: a preprocessing module is used to divide a reference image into a plurality of rectangular sub-regions, and determine a first contour boundary of a target object according to a deep learning model, and mark position information of a sub-region where a contour boundary segment exists; a feature extraction module connected with the preprocessing module is used to obtain the brightness of the pixel points in each sub-region to draw the brightness performance curve of the pixel points in each sub-region in different direction dimensions; the feature extraction module is used to calculate the brightness average value of the pixel points in each sub-region in different direction dimensions, wherein the feature extraction module is used to take the width data of each sub-region as the horizontal axis of the coordinate system to calculate the brightness average value of the pixel points determined based on each width numerical point in the sub-region, and take the length data of each sub-region as the horizontal axis of the coordinate system to calculate the brightness average value of the pixel points determined based on each length numerical point in the sub-region; the feature extraction module is used to draw the brightness performance curve in different direction dimensions, wherein the feature extraction module is used to draw a first brightness performance curve in a first direction dimension based on the brightness average value of the pixel points determined based on each width numerical point, and draw a second brightness performance curve in a second direction dimension based on the brightness average value of the pixel points determined based on each length numerical point; the horizontal axis of the first brightness performance curve is the width numerical value of the sub-region, and the vertical axis is the brightness average value of the pixel points, and the horizontal axis of the second brightness performance curve is the length numerical value of the sub-region, and the vertical axis is the brightness average value of the pixel points; a feature aggregation module connected with the preprocessing module and the feature extraction module respectively is used to determine an aggregated analysis region based on a position information matching condition, and mark non-confident sub-regions according to the comparison of the brightness performance curves of each sub-region in the aggregated analysis region; a local processing module connected with the feature aggregation module is used to obtain the brightness performance curve of the pixel points in the non-confident sub-region after adjusting the exposure, determine contour boundary fitting points according to the comparison of the brightness performance curve after adjusting the exposure, and fit the contour boundary segment of the non-confident sub-region; an output end connected with the local processing module and the preprocessing module respectively is used to update the first contour boundary by using the contour boundary segment of the non-confident sub-region, and output the second contour boundary of the target object of the reference image.

2. The deep learning based computer image processing system of claim 1, wherein, the feature aggregation module is used to determine the aggregated analysis region based on the position information matching condition, wherein the feature aggregation module is used to obtain the position information of each sub-region where the contour boundary segment exists, including a row sequence determined along the length direction of the reference image and a column sequence determined along the width direction of the reference image; the feature aggregation module is used to determine the region composed of a plurality of sub-regions meeting the position information matching condition as the aggregated analysis region; the aggregated analysis region comprises at least two sub-regions, and the position information matching condition is that the row sequences of the sub-regions are consistent or the column sequences of the sub-regions are consistent.

3. The deep learning based computer image processing system of claim 2, wherein, the feature aggregation module is used to determine the comparison of the brightness performance curves of each sub-region, wherein The feature aggregation module is used to determine the maximum difference in brightness based on the brightness performance curves of each sub-region within the aggregation analysis area, and to calculate the standard deviation of the maximum difference in brightness among the sub-regions within the aggregation analysis area.

4. The deep learning based computer image processing system of claim 3, wherein, The feature aggregation module is used to mark non-confidence sub-regions, wherein... The feature aggregation module is used to filter the aggregation analysis region corresponding to the maximum value of the standard deviation, and to mark the sub-regions corresponding to the maximum and minimum values ​​of the maximum difference in brightness within the corresponding aggregation analysis region as non-confidence sub-regions.

5. The deep learning based computer image processing system of claim 4, wherein, The local processing module is used to obtain the brightness performance curves of the pixels in the non-confidence sub-region after exposure adjustment in different directional dimensions, wherein... The local processing module is used to draw a third brightness performance curve in the first direction dimension based on the average brightness of the pixels at each width value point of the non-confidence sub-region after exposure adjustment, and to draw a fourth brightness performance curve in the second direction dimension based on the average brightness of the pixels at each length value point of the non-confidence sub-region after exposure adjustment. The exposure of the non-confidence sub-region is higher after adjustment than the exposure before adjustment.

6. The deep learning based computer image processing system of claim 5, wherein, The local processing module is used to determine the contour boundary fitting points, wherein... The local processing module is used to compare the average brightness of the third brightness performance curve with the average brightness of the first brightness performance curve on the vertical axis to determine the width of the maximum difference in the average brightness on the horizontal axis, and to compare the average brightness of the fourth brightness performance curve with the average brightness of the second brightness performance curve on the vertical axis to determine the length of the maximum difference in the average brightness on the horizontal axis. The local processing module is used to determine the contour boundary fitting point in the non-confidence sub-region based on the width value and the length value.

7. The deep learning based computer image processing system of claim 6, wherein, The local processing module is used to fit the contour boundary segment of the non-confidence sub-region, wherein... The local processing module is used to obtain the initial contour boundary segment of the non-confidence sub-region, and to fit the initial contour boundary segment based on the contour boundary fitting point, so that the fitted contour boundary segment passes through the contour boundary fitting point.

8. The deep learning based computer image processing system of claim 7, wherein, The output terminal is used to update the first contour boundary using the fitted contour boundary segment, wherein... The output terminal is used to replace the initial contour boundary segment with the fitted contour boundary segment, so as to update the first contour boundary to the second contour boundary.

Images