Image acquisition test method and device, equipment and storage medium

Abstract

The invention relates to the technical field of image acquisition and testing, and discloses an image acquisition and testing method, device and equipment and a storage medium, and the method comprises the steps: collecting a standard test card image through an image acquisition card in a multi-stage standard illumination environment, and calculating sensitivity benchmark test data; executing time domain and space domain combined sampling processing to obtain super-sensitivity image data; performing brightness analysis and dynamic adjustment on the currently shot first scene image to obtain a first image acquisition parameter; acquiring a second scene image, and performing image grid division and contrast compensation on the second scene image to obtain an enhanced image after contrast compensation; according to the super-sensitivity image acquisition method and the super-sensitivity image acquisition device, the sensitivity limitation of the position depth of a traditional image acquisition card is broken through, and super-sensitivity image acquisition is realized.

Description

TECHNICAL FIELD

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

[0002] Traditional image acquisition methods mainly rely on fixed parameter settings, which are difficult to cope with complex and variable lighting environments in actual applications. Especially on industrial production lines, lighting conditions may fluctuate due to time, location or environmental changes, causing the acquisition system to fail to maintain stable image quality, resulting in misjudgment in subsequent detection and analysis.

[0003] Existing image acquisition techniques have obvious limitations, especially in terms of sensitivity. The sensitivity of image acquisition cards is essentially limited by bit depth, with a theoretical single-pixel sensitivity limit of 1 / 2 n (n is bit depth), but in actual applications, this theoretical limit is difficult to achieve due to noise interference. In low light conditions, traditional acquisition methods are difficult to effectively capture low contrast or subtle image features, which poses a serious challenge to application scenarios that require detection of small defects or fine structures. In addition, existing techniques lack effective parameter prediction mechanisms and cannot intelligently adjust acquisition parameters according to lighting changes, resulting in system response lag when lighting changes suddenly. SUMMARY

[0004] The present application provides an image acquisition testing method, device, equipment and storage medium, which breaks through the sensitivity limit of traditional image acquisition card bit depth, realizes ultra-sensitivity image acquisition, and especially in limited pixel area, high-quality images can be obtained through time delay sampling technology.

[0005] In a first aspect, the present application provides an image acquisition testing method, which comprises: acquiring standard test card images through an image acquisition card under multi-level standard lighting environments and calculating sensitivity benchmark test data; performing time domain and spatial domain joint sampling processing according to the sensitivity benchmark test data to obtain ultra-sensitivity image data; using the ultra-sensitivity image data and the sensitivity benchmark test data to perform brightness analysis and dynamic adjustment on a first scene image currently captured to obtain first image acquisition parameters; obtaining a second scene image based on the first image acquisition parameters and performing image grid division and contrast compensation on the second scene image to obtain an enhanced image after contrast compensation; Based on the contrast-compensated enhanced image and the first image acquisition parameter, scene image acquisition parameter prediction is performed to obtain a second image acquisition parameter, and the image acquisition card is controlled to reacquire a target scene image based on the second image acquisition parameter.

[0006] In a second aspect, the present application provides an image acquisition testing device, comprising: An acquisition module is configured to acquire standard test card images by using an image acquisition card under multi-level standard light environments and to calculate sensitivity reference test data. A sampling processing module is configured to perform joint sampling processing in a time domain and a space domain based on the sensitivity reference test data to obtain hypersensitivity image data. A dynamic adjustment module is configured to perform brightness analysis and dynamic adjustment on a first scene image currently photographed based on the hypersensitivity image data and the sensitivity reference test data to obtain first image acquisition parameters. A compensation module is configured to acquire a second scene image based on the first image acquisition parameters, to perform image grid division and contrast compensation on the second scene image, and to obtain a contrast-compensated enhanced image. A parameter prediction module is configured to perform scene image acquisition parameter prediction based on the contrast-compensated enhanced image and the first image acquisition parameter to obtain second image acquisition parameters, and to control the image acquisition card to reacquire a target scene image based on the second image acquisition parameters.

[0007] In a third aspect, the present application provides a computer device, comprising a memory and at least one processor, wherein the memory stores instructions, and the at least one processor invokes the instructions in the memory to enable the computer device to perform the image acquisition testing method described above.

[0008] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer is enabled to perform the image acquisition testing method described above.

[0009] The technical scheme provided by the application establishes complete sensitivity benchmark test data by collecting and analyzing the standard test card under multi-stage standard light conditions, provides accurate reference standards for subsequent sensitivity optimization, and solves the problem of lack of unified standards for sensitivity evaluation. By using joint sampling processing in the time domain and the spatial domain, combining the pseudo-phase space model and singular value decomposition, the sensitivity limitation of the traditional image collection card bit depth is broken through, and super-sensitivity image collection is realized. Especially in the limited pixel area, high-quality images can also be obtained through time delay sampling technology. Based on the sensitivity benchmark test data, the brightness of the current scene is analyzed and dynamically adjusted, and the collection parameters are optimized through iterative sampling and bilinear interpolation strategy, so that the system can automatically adjust the optimal parameter setting according to different light conditions, and effectively cope with the change of light. By using grid division and contrast clustering analysis technology, adaptive local histogram equalization processing is implemented for low-contrast areas, so that the image features under weak light conditions are aligned with the features under standard light conditions, and the image quality under weak light environment is significantly improved. By constructing the probability relationship network among the light conditions, scene characteristics and collection parameters, intelligent prediction of the parameters is realized, the system can automatically adapt to new scenes according to historical experience, the parameter search time is reduced, and the system response speed is improved. The sensitivity optimization and contrast enhancement are integrated into a parallel double-channel structure, the dynamic processing mode switching is realized through the feature relationship fusion device and the decision controller, and images with high sensitivity and high-quality contrast can be obtained under various light conditions, which greatly improves the adaptability of the system in complex environments. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical scheme of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0011] Figure 1 The step schematic diagram of the image collection test method in the embodiment of the application is shown in the figure. Figure 2 The structure schematic diagram of the image collection test device in the embodiment of the application is shown in the figure. Figure 3 The structure schematic block diagram of the computer device in the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide an image acquisition test method, device, equipment and storage medium. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not have to be used to describe a particular order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific flow of the embodiments of the present application is described below. Please refer to Figure 1 One embodiment of the image acquisition test method in the embodiments of the present application includes: Step S1, acquiring standard test card images under multi-level standard light environment through an image acquisition card and calculating sensitivity reference test data; It can be understood that the execution subject of the present application can be an image acquisition test device, and can also be a terminal or a server, and the specific place is not limited. The embodiments of the present application take the server as the execution subject for example.

[0014] Specifically, a sensitivity test benchmark environment is constructed, which is established based on three core components: a standard test card, a controllable light source system, and a precision fixing device. The standard test card has a linearly distributed gray scale gradient area, the gray scale value range of which covers the complete bit depth supported by the image acquisition system, for example, from 0 to 255 in an 8-bit system, or from 0 to 4095 in a 12-bit system, and each adjacent gray scale theoretically only differs by 1 unit. At the same time, the light source system supports multiple levels of brightness adjustment and has high stability and illumination uniformity control capability, and adopts five levels of brightness preset points; the fixing device ensures that the geometric relationship between the test card and the acquisition device is constant, thereby ensuring the spatial consistency of the image and the comparability of the data. After the environment is assembled, the initialization parameter set of the image acquisition card is set, including setting the exposure time, gain, black level correction value and other key parameters. Image acquisition operations are performed under each level of standard light conditions in turn, and image data is repeatedly collected under each light condition to average random interference, and the pixel value distribution of the gray scale gradient area in the standard test card image is extracted. The extraction and statistical feature analysis of the gray scale gradient data are performed, all pixel values are extracted from the gray scale area of the image, and a gray scale distribution curve is constructed, the average gray value, standard deviation and distribution range of each gray scale block are calculated, and gray scale gradient distribution data are formed. According to the gray scale gradient distribution data, the gray scale difference between adjacent gray scales is analyzed, and whether there is a gray scale indistinguishable area, that is, the gray scale fusion phenomenon appearing between different gray scale segments due to noise superposition or insensitive response is focused on. By calculating the minimum resolution unit between the theoretical gray scale difference and the actual collected gray scale difference, the minimum gray scale difference value that can be distinguished by the image acquisition card under the current light condition is quantified, which is defined as the sensitivity index under the light condition. At the same time, in order to improve the statistical stability, the image acquisition under each light level is repeated 10 times, and the sensitivity values obtained are averaged to reduce accidental factor interference. All light intensity levels and their corresponding sensitivity index values are mapped and matched to construct a light-sensitivity reference table containing five light levels and corresponding sensitivity indexes, and the noise level under each level is recorded synchronously to obtain sensitivity benchmark test data.

[0015] Step S2, performing joint sampling processing in the time domain and the spatial domain according to the sensitivity benchmark test data to obtain super-sensitivity image data; Specifically, a theoretical analysis model is established based on the sensitivity benchmark test data, which is used to quantify the coupling effect of the bit depth n of the image acquisition card and the noise level σ on the sensitivity performance. On this basis, by analyzing the formula S actual =max(1 / 2 n ,σ / √N), where 1 / 2 nThe minimum distinguishable gray unit brought by the representative bit depth, while σ / √N represents the theoretical sensitivity lower limit obtained by multi-pixel averaging under noise conditions, thus making it clear that the sensitivity limitation is not only due to the bit depth of the analog-to-digital converter, but also significantly restricted by the noise level of the photosensitive element itself and the number of sampling pixels N. Through this analysis process, a theoretical basis model of sensitivity is established, which is used to guide the optimal design of sampling parameter setting and data fusion mechanism. The target scene is planned for spatial domain sampling strategy. The scene area to be measured is divided into multiple MxM grid cells according to the importance and complexity of the image content, and a time-space joint sampling strategy is designed by combining the preset number of frames T. This strategy requires consistent sampling in the time dimension for each spatial grid area in the consecutive T frames of images, so that the pixels at the same position form a time sequence sampling sequence in multiple frames, and all grid cells form a two-dimensional spatial structure, forming a three-dimensional time-space sampling body. When the spatial domain is limited (such as edges, dark parts, and low-contrast areas), increase the value of T to compensate for the lack of information and improve the dimension redundancy of information sampling. Based on the above strategy, consecutive T frames of images are collected in the same scene, and each frame of image is divided into MxM spatial grid areas, and the pixel sampling sequence in the time dimension of each grid is constructed. These sampling sequences constitute the original time-space joint sampling data. The pixel values of the same grid area in the T frames of images are unfolded into a one-dimensional vector in frame order, and all area sampling vectors are stacked to form a pseudo-phase space data point set with a dimension of D=M²xT. A pseudo-phase space model is established, and a pseudo-phase space data matrix X is constructed, each column of which represents a time sequence vector of a spatial area in T frames of images, and each row represents the response value of a certain pixel dimension in all grid areas. Perform singular value decomposition operation on the constructed pseudo-phase space data matrix X to obtain the singular value sequence representing the main information structure of the image and the corresponding principal component feature vector set. According to the energy distribution of the singular values in Σ, the first k principal components are selected for retention, and the high-order components with extremely low energy proportion are removed to filter out random noise interference, thereby realizing the suppression of time-space redundant noise. The retained principal components are used to reconstruct the pseudo-phase space feature data to obtain the multi-dimensional feature representation after noise reduction processing. Perform joint weighted averaging processing on the reconstructed feature data in the spatial domain and the time domain. The pixels in each grid area in the T frames of images are weighted and averaged, and the weight factor is adaptively determined according to the energy distribution of the singular values, the inter-frame brightness fluctuation, the spatial consistency and other multi-dimensional indexes, thereby improving the signal fidelity and noise reduction efficiency; then the inter-regional fusion is used to eliminate the grid boundary effect, and the edge smoothing function is used to construct the continuity of the overall image, and finally the super-sensitivity image data is obtained.

[0016] The bit depth parameter of the image acquisition card is analyzed. The bit depth parameter of the image acquisition card represents its theoretical upper limit in gray scale resolution. The higher the bit depth, the more delicate the image gray scale layering, and theoretically, the stronger the sensitivity performance. It is clear that in the ideal state without any noise interference, what is the minimum gray scale difference that the acquisition system can distinguish, and this difference constitutes the basic relationship of the influence of bit depth on sensitivity. Through the understanding of the gray scale layer number under different bit depths, a reference model is established to define the limit value of the theoretical sensitivity. Based on the noise level record in the sensitivity benchmark test data, the noise influence factor of noise on actual sensitivity under different light conditions is analyzed, and the inhibitory effect of noise on the acquisition sensitivity under different light conditions is described. In a high-brightness environment, the image acquisition signal is strong, and the noise proportion is relatively low, so the acquisition system approaches the theoretical sensitivity upper limit; but in a low-light scene, noise signals dominate, making the gray scale regions that should be distinguished appear blurred and fused, resulting in a significant decrease in actual sensitivity. By statistically analyzing the noise performance under multiple light levels and analyzing the corresponding relationship between the noise performance and the sensitivity test results, a descriptive model is constructed to reveal the degree of constraint of noise on image gray scale resolution, and this constraint is quantified as a noise influence factor. At the same time, the participation degree of the image structure itself is introduced, that is, the potential effect of the number of pixels in different image regions on the improvement of sensitivity. Local area averaging, multi-pixel fusion, spatial filtering and other methods are used to improve the signal-to-noise ratio of the image, and more pixel samples are used to reduce the uncertainty of the image, so the pixel set in different regions of the benchmark test image is analyzed, the support effect of different pixel densities on sensitivity performance is evaluated, and the pixel number influence factor is extracted. The bit depth sensitivity basic relationship, noise influence factor and pixel number influence factor are subjected to multiple regression analysis, and the coupling relationship between the three factors is fitted using statistical modeling and regression analysis methods to construct a sensitivity theoretical basic model. The model can clearly depict the response capability boundary of the image acquisition card under different parameter and environmental configurations, and has high predictability and transferability.

[0017] Step S3, using the ultra-sensitivity image data and the sensitivity benchmark test data, performing brightness analysis and dynamic adjustment on the first scene image currently shot to obtain first image acquisition parameters; Specifically, based on the established illumination-sensitivity mapping relationship table in the sensitivity benchmark test data, a parameter adjustment basic function between the illumination condition and the exposure time, gain value, and black level correction parameter is constructed. The image brightness feature extraction and statistical analysis are performed on the first scene image currently captured. The overall gray scale distribution of the image is calculated without pre-processing. The parameters including the overall average brightness, brightness histogram peak position, and brightness standard deviation are extracted. The brightness feature value representing the current illumination condition is constructed. The feature value is used as the input quantity of the parameter adjustment function and is used to replace the standard illumination value for subsequent function operation. Due to the existence of local shielding, light spot, high reflection, or dark corner and other complex influences in the actual scene, the brightness feature value is smoothed by integral gray scale analysis and median filtering method, so as to ensure that the input value has high representativeness and anti-interference ability. The current scene brightness feature value is substituted into the parameter adjustment basic function. The first parameter set suitable for the scene, that is, the first set of acquisition initial value configuration, is calculated. The parameters in this set are close to the optimal solution in space. Since they are derived from the fitting function under the standard condition, they still cannot completely meet the image quality requirements under the complex illumination distribution. Therefore, they are further optimized by dynamic iteration. In this process, the acquisition system controls the image acquisition card to perform test sampling according to the first parameter set, acquires one or more images, and performs image quality scoring on the acquired images. The scoring comprehensively considers the signal-to-noise ratio, edge definition, local contrast, color uniformity, and other indexes of the image, and outputs an evaluation value reflecting the adaptability of the current acquisition parameter. According to the scoring result, the gradient relationship between the current parameter and the score is analyzed. The current parameter adjustment direction is judged. The local search is performed in the parameter space. The exposure, gain, or black level parameter is adjusted by small steps. The optimal point is gradually approached under the guidance of the scoring function. The iteration is stopped until the score converges or the set upper limit number is reached. The second parameter set is obtained. The second parameter set and the hypersensitivity image data are aligned and verified in multiple dimensions such as gray scale distribution, structure texture, and edge response. The difference analysis algorithm is used to correct the brightness deviation or detail loss caused by the acquisition setting. The related parameters in the second parameter set are fine-tuned and corrected. The first image acquisition parameter is obtained.

[0018] Step S4, a second scene image is acquired based on the first image acquisition parameter. Image grid division and contrast compensation are performed on the second scene image. An enhanced image after contrast compensation is obtained. Specifically, the first image acquisition parameter is input as a control instruction to the image acquisition card to perform formal acquisition on the target scene to obtain a second scene image. The second scene image is subjected to grid division processing to divide the entire image into grid units of fixed size, for example, 8x8 pixels or 16x16 pixels, and a local contrast value is calculated in each unit, which is obtained by analyzing the relative difference between the maximum and minimum gray scales in the region to form a complete grid contrast distribution data graph. The grid contrast distribution data is subjected to clustering processing, and a K-means or other unsupervised clustering algorithm is used to divide all grid units into several categories according to the similarity of the contrast values, for example, three categories of high contrast, medium contrast and low contrast, or five categories. A contrast distribution feature vector is obtained after clustering, wherein each dimension represents the proportion of grid units of a certain contrast level in the entire image. The feature vector quantitatively describes the overall contrast distribution feature of the image, and is compared with a reference contrast vector under standard lighting conditions to analyze the contrast deviation exhibited by the current image under the specific lighting conditions. Based on the difference analysis of the contrast distribution feature vector, a mapping model is constructed to express the overall change trend of the current image relative to the standard lighting image in the contrast level, and to describe the contrast compensation amplitude required to be performed by different gray scale regions in the target image. Through the mapping model, the system identifies the regions in the current image that have a contrast significantly lower than the standard level, that is, the grid units that have information loss, blurred details or unclear structure. These regions are marked as local enhancement image regions. For the low-contrast regions, local enhancement processing is performed, for example, using an adaptive histogram equalization algorithm, and a CLAHE algorithm with a limited contrast limiting mechanism is used for the low-illumination regions to ensure that the enhancement process does not excessively amplify noise or damage edge smoothness. The local regions after enhancement processing are fused with the original image, and the fusion process uses a weighted superposition method to smoothly transition the boundary to obtain an enhanced image after contrast compensation.

[0019] Step S5, based on the enhanced image after contrast compensation and the first image acquisition parameter, scene image acquisition parameter prediction is performed to obtain a second image acquisition parameter, and the image acquisition card is controlled to re-acquire the target scene image based on the second image acquisition parameter.

[0020] Specifically, the scene characteristic analysis of the structured enhanced image after the contrast compensation processing includes the extraction of multi-dimensional features such as texture complexity, detail density, edge sharpness, and local contrast distribution, thereby describing the lighting state, visual structure, and imaging requirements of the current scene. By constructing a scene description vector oriented to these features, the visual content characteristics of the current target image are quantitatively represented, and they are combined with known image acquisition parameters to drive the generation process of the parameter prediction model. In this stage, a parameter prediction network model is constructed based on historical acquisition data. The model uses a graph structure representation, including three categories of lighting condition nodes, scene characteristic nodes, and acquisition parameter nodes, and by analyzing the corresponding relationship between lighting, scene, and parameters in historical acquisition samples, the conditional dependency paths between nodes are constructed to form a network framework with learning and reasoning capabilities. After establishing the parameter prediction network model, the conditional probability distribution of each acquisition parameter under different lighting and scene characteristic combinations is calculated based on the current input data. According to the network structure, probability propagation and joint distribution derivation are performed to generate a conditional probability distribution table containing core parameters such as exposure time, gain value, and black level correction value. This table describes the value range of each parameter under the given input, and quantifies the probability intensity corresponding to each value, thereby providing a prior information basis for subsequent reasoning. Based on the conditional probability distribution table, variable elimination is performed to gradually eliminate irrelevant nodes to simplify the calculation complexity, and the posterior probability of each target parameter node is solved through joint distribution reduction. In this way, the parameter value combination that is most likely to improve image quality under the current scene characteristics and first image acquisition parameter background is derived, and the preliminary parameter prediction result is obtained. Since directly using the parameter corresponding to the maximum probability value may lead to excessive parameter jump or poor coordination between multiple parameters in some cases, parameter transformation is performed on the preliminary result. This process introduces multiple adjustment rules to ensure the coordination between parameters and the stability of the system, such as automatically reducing the gain appropriately when the exposure time increases significantly to balance the imaging signal-to-noise ratio, or fixing adjacent parameters when the black level correction value changes critically to avoid picture jumping. Through these strategies, the second image acquisition parameters that meet the actual image acquisition requirements are constructed. The image acquisition card is controlled using the second image acquisition parameters to perform image acquisition on the target scene, and the target scene image is obtained.

[0021] In the embodiment of the present application, by collecting and analyzing the standard test card under multi-level standard light conditions, a complete sensitivity benchmark test data is established, which provides an accurate reference standard for subsequent sensitivity optimization, and solves the problem of lack of unified standard for sensitivity evaluation. By using joint sampling processing in time domain and spatial domain, combining with pseudo-phase space model and singular value decomposition, the sensitivity limitation of traditional image acquisition card bit depth is broken through, and super-sensitivity image acquisition is realized, especially in the limited pixel area, high-quality images can also be obtained through time delay sampling technology. Based on the sensitivity benchmark test data, the brightness of the current scene is analyzed and dynamically adjusted, and the acquisition parameters are optimized through iterative sampling and bilinear interpolation strategy, so that the system can automatically adjust the optimal parameter setting according to different light conditions, and effectively cope with the change of light. By using grid division and contrast clustering analysis technology, adaptive local histogram equalization processing is implemented for low-contrast areas, so that the image features under weak light conditions are aligned with the features under standard light conditions, and the image quality under weak light environment is significantly improved. By constructing the probability relationship network among light conditions, scene characteristics and acquisition parameters, intelligent prediction of parameters is realized, and the system can automatically adapt to new scenes according to historical experience, reducing the parameter search time and improving the system response speed. The sensitivity optimization and contrast enhancement are integrated into a parallel double-channel structure, and the dynamic processing mode switching is realized through the feature relationship fusion device and the decision controller, so that images with high sensitivity and high-quality contrast can be obtained under various light conditions, and the adaptability of the system in complex environment is greatly improved.

[0022] In a specific embodiment, the process of performing step S1 can specifically include the following steps: The standard test card, the controllable light source system and the fixing device are configured and assembled to establish a sensitivity test benchmark environment; The initialization parameter set is set for the image acquisition card, and based on the sensitivity test benchmark environment and the initialization parameter set, the standard test card images are collected under multi-level standard light environments; According to the standard test card image, the pixel value distribution data of the gray scale gradient region is extracted and the pixel value statistical features are calculated to obtain the gray scale gradient distribution data; According to the gray scale gradient distribution data, the pixel value difference between adjacent gray scales is analyzed to obtain the sensitivity index value under different light conditions; The sensitivity index value under different light conditions is associated and matched with the corresponding light intensity data to establish a mapping relationship table between the light intensity and the sensitivity benchmark, and the sensitivity benchmark test data is obtained.

[0023] Specifically, a high-repeatability and highly standardized image acquisition experimental environment is constructed, which is composed of three parts, namely, a standard test card, a controllable light source system, and a precision mechanical fixing device. The standard test card, as a basic tool for evaluating image acquisition performance, provides an imaging reference area with clear gray scale gradient distribution, continuous gray scale, and good physical consistency. The linear gray scale strip from 0 to 255 is selected, and in higher bit depth systems, the gray scale distribution from 0 to 1023 or 0 to 4095 is selected to adapt to 10-bit or 12-bit acquisition systems. The test card has good surface diffuse reflection characteristics and material stability, avoiding spot, texture interference or gray scale reflection distortion during acquisition, ensuring that the subsequent extracted pixel values accurately reflect the true relationship between illumination and system response. At the same time, the controllable light source system has high stability, high adjustability, and low noise output characteristics to achieve fine control of the lighting conditions in the test environment. Multiple light levels are configured to cover typical imaging conditions from low to high illumination. The brightness control unit of the light source needs to support linear adjustment or step switching, and is equipped with a light stability feedback mechanism, such as a high-sensitivity photoelectric sensor to monitor the actual illumination changes, to ensure the uniformity and stability of the light source output at each level of set illumination. In order to avoid the interference of ambient light, the test environment needs to be placed in a closed dark box to ensure that only the specified light source participates in the illumination during the image acquisition process, preventing external light interference from causing measurement deviation. The fixing device needs to have micron-level displacement control capability to ensure that the geometric position relationship between the standard test card and the image acquisition system remains consistent at all times, preventing image content distortion caused by angle deviation or position drift, which affects the accuracy of subsequent pixel extraction. After completing the assembly of the test environment, set the initialization parameter set for the image acquisition card, which represents the basic working state and does not introduce any additional enhancement, compensation, or nonlinear adjustment processing. Under five levels of illumination, the standard test card images are collected step by step, and no less than ten images are collected at each level for statistical smoothing. At the same time, during the acquisition process, the test card position is fixed, the camera does not change the focal length, and the system temperature remains stable, thereby avoiding systematic errors caused by device changes. After image acquisition, the image data analysis phase begins, automatically identifying and extracting the gray scale gradient area in the standard test card from each image, and performing pixel value distribution statistics for each gray scale. For each gray scale strip area, all pixel values are extracted, and statistical features such as mean, maximum, minimum, and standard deviation are calculated, and whether the gray scale distribution has abnormal jumps, fusion, drift, etc. is recorded to form a gray scale gradient distribution data matrix. Based on the gray scale gradient distribution data, the pixel response difference between adjacent gray scales is analyzed, that is, by comparing the pixel mean difference of adjacent two gray scale blocks to determine the actual minimum gray scale change unit that can be distinguished.If the difference is close to the theoretical value, it indicates that the system has high sensitivity under the light condition; if the difference deviates significantly, especially in the low gray scale region, it indicates that the acquisition system has a significant decrease in sensitivity under the light level. By traversing the response characteristics of the entire gray scale range, an index value of sensitivity is extracted at each light level to measure the actual performance of the minimum gray scale resolution of the system under the light intensity. This index is measured by the "minimum effective gray scale difference under unit light", which is regarded as the quantitative expression of system sensitivity. The sensitivity index values calculated under different light conditions are matched with the corresponding light intensity data to establish a mapping relationship table between light intensity and sensitivity performance. This table records the image response capability corresponding to each level of light in the standard test environment, and also contains statistical information of the corresponding noise level, forming a sensitivity benchmark test data system.

[0024] In a specific embodiment, the process of performing step S2 can specifically include the following steps: Based on the sensitivity benchmark test data, analyze the influence relationship of bit depth and noise of the image acquisition card on sensitivity, and establish a sensitivity theoretical basis model; Spatial domain sampling planning is performed on the scene area to be tested to obtain a space-time domain joint sampling strategy; Based on the space-time domain joint sampling strategy, T frames of images of the same scene are continuously acquired and the same region is divided to obtain original space-time sampling data; Based on the original space-time sampling data, a pseudo phase space model is constructed, and the corresponding region of each grid unit in the T frames of images is regarded as a high-dimensional data point to construct a pseudo phase space data matrix; Perform singular value decomposition operation on the pseudo phase space data matrix to obtain feature data after noise reduction processing, and perform space-time domain joint average processing on the feature data after noise reduction processing to obtain hypersensitivity image data.

[0025] Specifically, based on the sensitivity benchmark test data, the response relationship between the bit depth performance and noise suppression ability of the image acquisition card under different illumination conditions is analyzed. Bit depth determines the minimum gray scale resolution of the image acquisition system. In the ideal state without any noise interference, the higher the bit depth, the more delicate the gray scale change that the system can perceive, and the higher the theoretical sensitivity. However, the image acquisition system is inevitably affected by noise from factors such as circuit, sensor, and environmental light, making the actual sensitivity lower than the theoretical upper limit. Therefore, by analyzing the gray scale response, pixel stability, and noise level contained in the sensitivity benchmark test data, the deviation between the bit depth and the actual gray scale response ability under different illumination conditions is extracted to construct a theoretical foundation model describing the limited relationship of acquisition sensitivity. This model not only covers the theoretical gray scale analysis ability corresponding to the bit depth, but also considers the superposition effect of noise in space and time distribution, and quantifies it as a constraint factor of the image system perception ability, so as to clearly define the performance boundary of image detail acquisition ability under different noise and bit depth combinations. Under the guidance of this theoretical model, in order to improve the actual sensitivity of image acquisition in real scenes, the spatial structure of the scene to be tested is divided into regions and the sampling strategy is designed, and a spatial domain sampling planning scheme is developed. According to the content complexity, structure uniformity, and texture density of the target image, the image is divided into multiple regular grid regions, and an M×M pixel grid is used as the basic sampling unit for subsequent high-consistency data superposition and statistical analysis. At the same time, combined with the scene dynamics and the acquisition frame rate setting, the number of consecutive sampling frames T is set in the time dimension to form a time sample set of T consecutive images, constituting a combined sampling strategy in space and time domains. This strategy enhances the sampling depth through time redundancy, and can still accumulate sufficient information when the spatial texture is insufficient or the illumination is uneven, thereby improving the sensitivity performance. Under this spatio-temporal joint sampling strategy, T consecutive image data of the same target scene are continuously acquired, and each image uses a consistent spatial division method to ensure that all images are aligned in terms of coordinate positioning, grid size, and pixel distribution structure, generating a structured original spatio-temporal sampling data set. Each image is regarded as a two-dimensional data surface, and the T consecutive images constitute a three-dimensional data body. Based on this three-dimensional body, the pixel region corresponding to each spatial grid element in the T images is extracted and unfolded into a high-dimensional feature vector in chronological order, so that each grid element is mapped to a high-dimensional data point with time series characteristics, thereby constructing a pseudo-phase space structure. The high-dimensional feature vectors of all grid elements are collected and arranged to form a pseudo-phase space data matrix, which takes the time series data of each sampling point as the column vector, integrates the observation data of the entire image in time and space dimensions, and contains the change rule and stability pattern of each region in the image in the time dimension.Based on the pseudo-phase space data matrix, singular value decomposition operation is performed to separate the principal component structure and noise component hidden in the matrix. Through the singular value decomposition method, high-dimensional observation data is decomposed into principal component signals and low-energy disturbances, thereby effectively identifying the structural features that exist stably in continuous images, and eliminating noise data that is randomly distributed, changes dramatically and has no consistency, to obtain feature data after noise reduction processing. All high-dimensional feature data after noise reduction is processed by joint time and space domain average, that is, weighted average is performed on the multi-frame response values of each pixel position in the time dimension, local response integration is performed on each sampling area in the spatial dimension, and an edge preservation mechanism is introduced to prevent blurred edges or texture loss during the average process. The weighting factor is dynamically allocated according to the principal component distribution, signal-to-noise ratio estimation result or image local variation degree, so as to ensure that strong structure areas obtain more reservation weight and weak texture areas receive greater degree of smoothing processing. Finally, through the joint average mechanism, the optimal response in each frame image is fused into a data set that is superior to the original image in gray resolution, detail performance and structural continuity, to obtain hypersensitivity image data.

[0026] In a specific embodiment, the performing step analyzes the relationship between the bit depth of the image acquisition card and the sensitivity of the noise based on the sensitivity benchmark test data, and the process of establishing the sensitivity theoretical basis model can specifically include the following steps: Analyzing the bit depth parameter of the image acquisition card to obtain a bit depth sensitivity basis relationship; Based on the noise level record in the sensitivity benchmark test data, analyzing the noise influence factor of the noise on the actual sensitivity under different light conditions; Analyzing different regional pixel sets in the sensitivity benchmark test data to obtain a pixel quantity influence factor; Performing multiple regression analysis on the bit depth sensitivity basis relationship, the noise influence factor and the pixel quantity influence factor to obtain the sensitivity theoretical basis model.

[0027] Specifically, the bit depth parameter of the image acquisition card is analyzed. In the ideal noise-free theoretical state, the bit depth of the image acquisition card determines the total number of distinguishable gray levels and determines the minimum resolution capability of the system to perceive brightness differences. The higher the bit depth, the finer the theoretical gray scale accuracy, and the more sensitive the image acquisition system to small changes in light, the higher the sensitivity. There is a positive correlation between bit depth and sensitivity. The relationship is derived by analyzing the minimum gray scale difference that can be recognized by different bit depth systems under the same test card conditions, and a bit depth sensitivity basic relationship model is obtained to characterize the theoretical response lower limit of the acquisition card. Extract the noise-related raw records from the sensitivity benchmark test data, analyze the image noise level under different light conditions, and quantify the inhibitory effect of noise on gray scale resolution. At each standard light level, the standard deviation and offset of the gray scale region in each frame of image are calculated. These statistics comprehensively reflect the stability and response consistency of the image under that light condition. By comparing with the theoretical minimum gray scale difference, the degree of interference caused by noise on the actual sensitivity is calculated, that is, the noise influence factor. The larger the value, the more significant the noise influence, the lower the sensitivity. Analyze the pixel distribution in the sensitivity benchmark test image, model the number, distribution density and region texture complexity of the sampled pixels in different gray scale regions. By comparing the influence of different pixel number sets on the image response stability and sensitivity index under the same gray scale level, an pixel number influence factor reflecting the sampling redundancy is extracted. This factor describes the auxiliary role of pixel sampling size in improving image sensitivity, especially in dark, edge or sparse detail areas. Increasing the sampling area or time accumulation can significantly improve the distinguishable gray scale difference, so this factor provides theoretical support for the noise reduction algorithm, time series superposition mechanism and region weighted enhancement strategy of the image processing module. Perform multiple regression analysis on the bit depth sensitivity basic relationship, noise influence factor and pixel number influence factor, and use the actual sensitivity index value in the sensitivity benchmark test data as the explained variable. In the modeling process, the original data is standardized for pretreatment to eliminate the interference of variable scale on regression weight, and then linear regression, polynomial fitting or principal component regression methods are used for modeling training. Select the model form with the smallest regression residual and the best fitting degree as the final output to obtain the sensitivity theoretical basic model. The regression model can structure and quantify the influence path of multiple factors on sensitivity, clearly the contribution of each variable in the formation process of sensitivity, and reveal the change trend and response interval of image system sensitivity under different bit depths, different noise levels and different pixel sets.

[0028] In a specific embodiment, the process of performing step S3 can specifically include the following steps: Based on the light-sensitivity mapping relationship table in the sensitivity benchmark test data, a parameter adjustment basic function between the light condition and the exposure time, the gain value and the black level correction parameter is constructed; Statistical analysis is performed on the first scene image currently photographed to obtain a current scene brightness characteristic value; The first parameter set suitable for the current scene light condition is calculated by substituting the current scene brightness characteristic value into the parameter adjustment basic function; The first parameter set is used to control the image acquisition card to perform test sampling, an image quality score is obtained, and the optimal parameter direction is found according to the image quality score, and iterative search is performed in the parameter space to obtain a second parameter set; The second parameter set is matched and calibrated with the hypersensitivity image data to obtain the first image acquisition parameter.

[0029] Specifically, based on the illumination-sensitivity mapping relationship table in the sensitivity benchmark test data, a parameter adjustment basic function between the illumination condition and the exposure time, gain value, and black level correction parameter is constructed. The first scene image currently captured is statistically analyzed to estimate the current real illumination condition. By calculating the average brightness, local brightness distribution density, gray histogram peak position, and brightness standard deviation of the entire image, a brightness feature value representing the overall lighting level of the current image is comprehensively formed. In order to prevent a few high-contrast regions in the image from interfering with the overall brightness evaluation, this process combines median filtering and boundary exclusion strategies to preferentially select the central region or typical texture region of the image for analysis, obtaining a stable current scene brightness feature. The current scene brightness feature value is substituted into the parameter adjustment basic function to calculate the estimated optimal exposure time, gain value, and black level correction value under the current illumination condition, obtaining a first parameter set suitable for the current scene. The first parameter set is used to control the image acquisition card to test and sample the target scene, and an image quality score operation is performed on the acquired image, using a weighted fusion method to output a unified score value. After obtaining the preliminary score, it is compared with the set threshold value, and the parameter adjustment direction is determined according to the score curve trend, i.e., identifying which parameter has a dominant influence on the current image quality limitation. Based on this judgment, a local iterative search process in the parameter space is entered, one or more parameters are adjusted by a small amount, the image is reacquired and the quality score is executed, and the "parameter adjustment-image sampling-score feedback" process is repeatedly executed, and the parameter changes and score results of each iteration are recorded in real time. When the score converges or reaches the set maximum number of iterations, a second parameter set with the optimal comprehensive score is locked. The second parameter set is matched and calibrated with the hypersensitivity image data, and the gray scale details, boundary texture, and dark response in the hypersensitivity image data are used to align the features of the image acquired by the current parameter set. When local deviations are found, the micro values of some parameters are adjusted, such as slightly adjusting the gain when the bright and dark dynamic responses are inconsistent, appropriately lengthening the exposure time when the detail resolution is insufficient, and fine-tuning the black level correction value when the black level response is compressed, so as to generate the first image acquisition parameter which is highly adapted to the current scene content and compatible with the upper limit of the hardware performance of the image acquisition card through matching and calibration operations.

[0030] In a specific embodiment, the process of performing step S4 can specifically include the following steps: using the first image acquisition parameter to control the image acquisition card to acquire the target scene, obtaining a second scene image; performing grid division processing on the second scene image to obtain grid cell contrast distribution data; performing clustering on the grid cell contrast distribution data to obtain a contrast distribution feature vector; obtaining a contrast mapping model based on the contrast distribution feature vector analyzing the difference of the contrast under the standard lighting condition and the current lighting condition; identifying a low-contrast region in the second scene image according to the contrast mapping model, and obtaining a local enhanced image region; fusing the local enhanced image region with the second scene image, and obtaining an enhanced image after contrast compensation.

[0031] Specifically, the image acquisition card is controlled to capture the target scene using the first image acquisition parameters to obtain a second scene image. To achieve accurate modeling of the local contrast structure of the image, the image is divided into a plurality of grid cells of equal size through structured image region division processing, for example, the image is spatially divided into 8x8 pixels, 16x16 pixels or 32x32 pixels window distribution, each grid cell is taken as a local image sub-region, and the contrast index of each grid cell is calculated independently. The contrast in each grid is normalized by the difference between the maximum pixel value and the minimum pixel value, or the scale is adjusted using a normalization coefficient combined with the local brightness average of the image to form a contrast distribution map. The map is a thermal expression of the contrast in the spatial dimension of the image, reflecting which regions have obvious structural edges or texture clarity, and which regions are relatively flat or lack clear levels due to insufficient lighting causing gray scale compression. After obtaining the local contrast data of each grid, clustering processing is performed to identify the distribution pattern of the overall contrast structure of the image. The clustering method selects the K-means algorithm or the Gaussian mixture model, and all grids are divided into several contrast levels, such as low contrast, medium contrast, and high contrast, according to the density distribution of the contrast values, and the proportion of each category in the entire image is counted to construct a contrast distribution feature vector. Each dimension of the vector represents the proportion of the number of grid cells of a certain contrast level to the total image area, so it can describe the statistical characteristics of the contrast distribution in the spatial structure of the image. In this way, the contrast state of the image is transformed into a structured, measurable and comparable vector representation. The contrast distribution feature vector described above is compared with the feature vector corresponding to the standard test image under standard lighting conditions, the deviation of the two in each level distribution is calculated, and a contrast mapping model is established using the deviation relationship. The model aims to express the visual information loss path caused by contrast compression, gray aggregation or brightness drift in the current lighting environment, and its essence is a function mapping structure to describe how to transform from the current contrast state to the enhanced target state close to the standard state. The mapping model not only considers the balance of the overall distribution, but also includes the matching degree of the spatial distribution characteristics, so a hierarchical weighting mechanism is set in the model to give higher adjustment priority to low-contrast areas, thereby guiding the system to focus on and respond to the image parts that are most likely to cause visual degradation. The contrast mapping model is applied to the contrast distribution map of the second scene image to identify the areas in the current image that have lower contrast than the target reference standard, i.e., the potential information missing area or visual blur area. This step compares the contrast value of each grid cell with the corresponding compensation target in the model, and when the actual contrast value is significantly lower than the expected value, the area is marked as an image area that needs to be enhanced. The marking result forms a local enhanced image region mask map indicating which areas of the image need to enter the local contrast enhancement process and be selected for processing in subsequent processing.For the marked region, a local adaptive enhancement method is used for processing, which is improved based on the histogram equalization framework, for example, using the CLAHE algorithm with limited contrast limit, which sets the local enhancement threshold and window range for each enhanced region, and performs dynamic histogram reconstruction according to the brightness distribution in the region, to ensure that the contrast is improved while not introducing artifacts or over-enhancement problems. To avoid obvious boundaries between the local processing result and the original image, a region fusion operation is performed after the enhancement is completed, and the weighted average or Gaussian edge transition technology is used to embed the processed region into the original image naturally, to obtain the contrast-compensated enhanced image.

[0032] In a specific embodiment, the process of performing step S5 can specifically include the following steps: Performing scene characteristic analysis on the contrast-compensated enhanced image to obtain current scene characteristic data, and constructing a parameter prediction network model based on the current scene characteristic data and the first image acquisition parameters; Calculating the probability distribution of each parameter value under different combinations of lighting conditions and scene characteristics through the parameter prediction network model to obtain a conditional probability distribution table; Performing variable elimination and posterior probability calculation based on the conditional probability distribution table to obtain a preliminary parameter prediction result, and performing parameter transformation based on the preliminary parameter prediction result to obtain second image acquisition parameters; Using the second image acquisition parameters to control the image acquisition card to perform image acquisition operation to obtain the target scene image.

[0033] Specifically, scene characteristic analysis is performed on the contrast-compensated enhanced image, including extraction of indicators such as local contrast distribution, texture density, edge sharpness, regional detail complexity, and tone distribution balance, and a set of structured scene characteristic data is formed based on the same. The data is quantifiable, comparable, and embeddable, facilitating its use as input features for the modeling process of the prediction network. At the same time, the first image acquisition parameters determined through brightness evaluation, score optimization, and hypersensitivity calibration are input synchronously as the description of the initial configuration state of the current system, so that the input dimension of the model includes not only the environmental characteristics reflected by the current image but also the acquisition mechanism parameters relied on by the current image. Based on the current scene characteristic data and the first image acquisition parameters, a parameter prediction network model is constructed, which is modeled in a probabilistic graph structure and adopts a directed acyclic graph form to construct a node relationship system. Three types of nodes are set in the graph, namely, illumination condition nodes, scene characteristic nodes, and acquisition parameter nodes. The illumination nodes are divided into several levels of states such as low illumination, medium-low illumination, medium illumination, medium-high illumination, and high illumination. The scene characteristic nodes include image contrast level, texture complexity level, edge distribution level, etc. The acquisition parameter nodes include exposure time, gain value, black level correction value, and other controllable variables. The edges between nodes in the graph represent conditional dependency relationships, for example, the acquisition parameter nodes depend on the combined state of the illumination nodes and the scene characteristic nodes. The scene characteristic nodes are directly obtained through image analysis, and the illumination node state can be determined by brightness feature estimation, thereby forming a prediction network driven by image content and known parameters. After the network structure is built, a complete conditional probability distribution table is constructed using the pre-test data, standard image samples, and real acquisition records. The table records the probability distribution of different parameter values under various combinations of illumination conditions and scene characteristics, thereby forming a parameter probability mapping system. This system covers the entire parameter space and can provide corresponding parameter prediction results for any input feature combination. The system eliminates the probability gaps caused by sparse states through Laplace smoothing or Bayesian adjustment techniques. In real-time operation, the system calls the corresponding distribution structure from the conditional probability distribution table based on the scene characteristic state extracted from the current image, the probability dependency relationship between the nodes in the prediction network, and the current illumination judgment result, to obtain the parameter probability matrix for the current scene. Based on the above conditional probability table, the variable elimination and posterior probability calculation process is performed to gradually eliminate the intermediate variable nodes that have no direct relationship with the target parameter, reduce the computational complexity, and calculate the posterior probability distribution of the target parameter node while retaining the necessary dependency paths. In this process, the maximum probability of exposure time, gain, and black level is solved to obtain the preliminary parameter combination that is most likely to improve image quality. Parameter transformation is performed on the preliminary parameter combination.The parameter transformation stage adjusts the preliminary prediction result by setting multiple coordination rules and optimization functions, for example, appropriately reducing the gain to prevent overexposure when the exposure time is increased, or compensating for brightness jump phenomenon by adapting gain delay when the black level correction value is critically changed, to ensure that the entire parameter set reaches a balanced state in terms of imaging quality, signal-to-noise ratio, gray scale response, detail retention, and system running stability. The second image acquisition parameters output by the parameter prediction network and optimized through adjustment are sent to the image acquisition card for controlling it to perform formal image acquisition operations. In this stage, the system reacquires the target scene image according to the new parameter configuration, and this acquisition is based on the dynamic decision result generated by the three-dimensional fusion of content understanding, condition prediction, and acquisition mechanism, and has adaptive and predictive intelligent acquisition behavior. After the acquisition is completed, the target scene image is obtained.

[0034] The image acquisition test method in the embodiments of the present application is described above, and the image acquisition test device in the embodiments of the present application is described below. Please refer to Figure 2 An embodiment of the image acquisition test device in the embodiments of the present application includes: The acquisition module is configured to acquire the standard test card image through the image acquisition card and calculate the sensitivity reference test data under the multi-level standard light environment. The sampling processing module is configured to perform joint sampling processing in the time domain and the spatial domain based on the sensitivity reference test data to obtain the hypersensitivity image data. The dynamic adjustment module is configured to perform brightness analysis and dynamic adjustment on the first scene image based on the hypersensitivity image data and the sensitivity reference test data to obtain the first image acquisition parameter. The compensation module is configured to acquire the second scene image based on the first image acquisition parameter, perform image grid division and contrast compensation on the second scene image, and obtain the enhanced image after contrast compensation. The parameter prediction module is configured to predict the scene image acquisition parameter based on the enhanced image after contrast compensation and the first image acquisition parameter, obtain the second image acquisition parameter, and control the image acquisition card to reacquire the target scene image based on the second image acquisition parameter.

[0035] Through the synergistic cooperation of the above-mentioned various components, through the collection and analysis of standard test cards under multi-level standard light conditions, a complete sensitivity benchmark test data is established, which provides an accurate reference standard for subsequent sensitivity optimization, and solves the problem of lack of unified standard for sensitivity evaluation. By using the joint sampling processing of time domain and spatial domain, combining the pseudo-phase space model and singular value decomposition, the sensitivity limitation of the traditional image acquisition card bit depth is broken through, and super-sensitivity image acquisition is realized, especially in the limited pixel area, high-quality images can also be obtained through time delay sampling technology. Based on the sensitivity benchmark test data, the brightness of the current scene is analyzed and dynamically adjusted, and the acquisition parameters are optimized through iterative sampling and bilinear interpolation strategy, so that the system can automatically adjust the optimal parameter setting according to different light conditions, and effectively cope with the change of light. By using grid division and contrast clustering analysis technology, adaptive local histogram equalization processing is implemented for low-contrast areas, so that the image features under weak light conditions are aligned with the features under standard light conditions, and the image quality under weak light environment is significantly improved. By constructing the probability relationship network among the light conditions, scene characteristics and acquisition parameters, intelligent prediction of the parameters is realized, and the system can automatically adapt to new scenes according to historical experience, reducing the parameter search time and improving the system response speed. The sensitivity optimization and contrast enhancement are integrated into a parallel double-channel structure, and the dynamic processing mode switching is realized through the feature relationship fusioner and the decision controller, so that images with high sensitivity and high-quality contrast can be obtained under various light conditions, which greatly improves the adaptability of the system in complex environments.

[0036] Reference Figure 3 In the embodiment of the present application, a computer device is also provided, which can be a server, and the internal structure thereof can be as shown in Figure 3 The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in the embodiment. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement the above method.

[0037] Those skilled in the art can understand Figure 3 that the structure shown in the embodiment is only a block diagram of part of the structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied.

[0038] An embodiment of the present application further provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the method described above. It can be understood that the computer readable storage medium in the embodiment can be a volatile or non-volatile computer readable storage medium.

[0039] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium provided by the present application and used in the embodiments can include non-volatile and / or volatile memory. The non-volatile memory can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. The volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and memory bus dynamic RAM (MDRAM), etc.

[0040] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-mentioned system, system and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein.

[0041] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0042] The above-described embodiments are merely used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some of the technical features; and these modifications or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An image acquisition test method, characterized in that: include: Capture standard test card images using an image acquisition card in a multi-level standard lighting environment and calculate sensitivity benchmark test data; Performing time domain and space domain joint sampling processing according to the sensitivity benchmark test data to obtain ultra-sensitive image data; Using the ultra-sensitive image data and the sensitivity benchmark test data, performing brightness analysis and dynamic adjustment on a first scene image currently being captured to obtain first image acquisition parameters; Acquire a second scene image based on the first image acquisition parameters, and perform image grid division and contrast compensation on the second scene image to obtain a contrast-compensated enhanced image; A scene image acquisition parameter prediction is performed based on the contrast-compensated enhanced image and the first image acquisition parameter to obtain a second image acquisition parameter, and the image acquisition card is controlled to re-acquire the target scene image based on the second image acquisition parameter.

2. The image acquisition test method according to claim 1, characterized in that: The method of acquiring a standard test card image by an image acquisition card under a multi-level standard illumination environment and calculating sensitivity benchmark test data includes: Configure and assemble standard test cards, controllable light source systems, and fixtures to establish a sensitivity test benchmark environment; Setting an initialization parameter set for an image acquisition card, and respectively acquiring standard test card images under multiple levels of standard lighting environments based on the sensitivity test benchmark environment and the initialization parameter set; Extracting pixel value distribution data of the grayscale gradient area according to the standard test card image and calculating pixel value statistical features to obtain grayscale gradient distribution data; Analyzing the pixel value differences between adjacent grayscales according to the grayscale gradient distribution data to obtain sensitivity index values ​​under different lighting conditions; The sensitivity index values ​​of the different lighting conditions are associated and matched with the corresponding light intensity data, a mapping relationship table between light intensity and sensitivity benchmark is established, and sensitivity benchmark test data is obtained.

3. The image acquisition test method according to claim 2, characterized in that: The performing time domain and space domain joint sampling processing according to the sensitivity benchmark test data to obtain ultra-sensitive image data includes: Analyze the relationship between the image acquisition card bit depth and noise on sensitivity based on the sensitivity benchmark test data, and establish a sensitivity theory basic model; Perform spatial domain sampling planning on the scene area to be tested and obtain a joint sampling strategy in the spatiotemporal domain; Based on the spatiotemporal domain joint sampling strategy, T frames of images are continuously collected for the same scene and the same area is divided to obtain the original spatiotemporal sampling data; constructing a pseudo phase space model based on the original spatiotemporal sampling data, and treating the corresponding area of ​​each grid unit in the T-frame image as a high-dimensional data point to construct a pseudo phase space data matrix; A singular value decomposition operation is performed on the pseudo-phase space data matrix to obtain feature data after noise reduction processing, and a spatiotemporal domain joint averaging process is performed on the feature data after noise reduction processing to obtain ultra-sensitive image data.

4. The image acquisition test method according to claim 1, characterized in that: The analysis of the relationship between the influence of image acquisition card bit depth and noise on sensitivity based on the sensitivity benchmark test data and the establishment of a sensitivity theory basic model include: Analyze the bit depth parameters of the image acquisition card to obtain the basic relationship between bit depth and sensitivity; Analyzing the noise impact factor of noise on actual sensitivity under different lighting conditions based on the noise level records in the sensitivity benchmark test data; Analyzing pixel sets in different regions of the sensitivity benchmark test data to obtain pixel quantity influencing factors; A multiple regression analysis is performed on the bit depth sensitivity basic relationship, the noise influencing factor, and the pixel number influencing factor to obtain a sensitivity theory basic model.

5. The image acquisition test method according to claim 4, characterized in that: The method of using the ultra-sensitive image data and the sensitivity benchmark test data to perform brightness analysis and dynamic adjustment on the currently captured first scene image to obtain first image acquisition parameters includes: Based on the illumination-sensitivity mapping relationship table in the sensitivity benchmark test data, constructing a parameter adjustment basic function between illumination conditions and exposure time, gain value, and black level correction parameters; Performing statistical analysis on the first scene image currently captured to obtain a brightness characteristic value of the current scene; Substituting the current scene brightness characteristic value into the parameter adjustment basic function to calculate a first parameter set suitable for the current scene lighting conditions; Using the first parameter set to control the image acquisition card to perform test sampling to obtain an image quality score, and finding an optimal parameter direction based on the image quality score, performing an iterative search in the parameter space to obtain a second parameter set; The second parameter set is matched and calibrated with the ultra-sensitive image data to obtain first image acquisition parameters.

6. The image acquisition test method according to claim 1, characterized in that: The acquiring of a second scene image based on the first image acquisition parameter, and performing image grid division and contrast compensation on the second scene image to obtain a contrast-compensated enhanced image, includes: Using the first image acquisition parameter to control the image acquisition card to acquire a target scene to obtain a second scene image; Performing grid division processing on the second scene image to obtain grid unit contrast distribution data; Clustering the grid unit contrast distribution data to obtain a contrast distribution feature vector; Analyzing the contrast difference between the standard lighting condition and the current lighting condition based on the contrast distribution feature vector to obtain a contrast mapping model; identifying a low-contrast region in the second scene image according to the contrast mapping model to obtain a local enhanced image region; The local enhanced image region is fused with the second scene image to obtain a contrast-compensated enhanced image.

7. The image acquisition test method according to claim 1, characterized in that: The method of predicting scene image acquisition parameters based on the contrast-compensated enhanced image and the first image acquisition parameters to obtain second image acquisition parameters, and controlling the image acquisition card to re-acquire the target scene image based on the second image acquisition parameters, includes: Performing scene characteristic analysis on the contrast-compensated enhanced image to obtain current scene characteristic data, and constructing a parameter prediction network model based on the current scene characteristic data and the first image acquisition parameters; Calculate the probability distribution of each parameter value under different combinations of lighting conditions and scene characteristics through the parameter prediction network model to obtain a conditional probability distribution table; performing variable elimination and posterior probability calculation based on the conditional probability distribution table to obtain preliminary parameter prediction results, and performing parameter transformation based on the preliminary parameter prediction results to obtain second image acquisition parameters; The image acquisition card is controlled using the second image acquisition parameter to perform an image acquisition operation to obtain a target scene image.

8. An image acquisition test device, characterized in that: Used to perform the image acquisition test method according to any one of claims 1 to 7, the image acquisition test device comprising: An acquisition module is used to acquire standard test card images through an image acquisition card under multi-level standard lighting environments and calculate sensitivity benchmark test data; a sampling and processing module, configured to perform time-domain and space-domain joint sampling processing according to the sensitivity benchmark test data to obtain ultra-sensitive image data; a dynamic adjustment module, configured to perform brightness analysis and dynamic adjustment on a currently captured first scene image using the ultra-sensitive image data and the sensitivity benchmark test data to obtain first image acquisition parameters; a compensation module, configured to acquire a second scene image based on the first image acquisition parameters, and perform image grid division and contrast compensation on the second scene image to obtain an enhanced image after contrast compensation; A parameter prediction module is used to predict scene image acquisition parameters based on the contrast-compensated enhanced image and the first image acquisition parameters to obtain second image acquisition parameters, and control the image acquisition card to re-acquire the target scene image based on the second image acquisition parameters.

9. A computer device, characterized in that: The apparatus comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the image acquisition test method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the processor is enabled to perform the image acquisition test method according to any one of claims 1 to 7.

Images