A method and system for target recognition and detection of a turnstile passage

Abstract

The present invention relates to the technical field of image recognition of targets in a turnstile passage, and specifically discloses a method and system for target recognition and detection in a turnstile passage. By collecting image data and through image preprocessing steps such as light compensation and noise removal, the brightness features and texture features in the image are extracted, and the visibility coefficient of the target area is calculated, so that the target maintains a high recognizability under various lighting conditions. By comprehensively analyzing the target visibility coefficient and the recognition confidence, the final recognition accuracy of the target in different environments is evaluated, and thus the image preprocessing and target recognition algorithms are dynamically adjusted to ensure that the target can be recognized quickly and accurately, improving the target recognition efficiency in the turnstile passage and ensuring the high efficiency and robustness of the system under complex lighting conditions, having practical value and application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition of targets in turnstile channels, and particularly relates to a method and system for recognizing and detecting targets in turnstile channels. Background Art

[0002] With the development of intelligent transportation systems, turnstile channels, as key access management areas, are widely used in places such as subways, airports, and scenic spots to achieve personnel identity authentication and access management. To ensure access safety and efficiency, target recognition technology has been introduced into turnstile systems to accurately identify passing personnel. However, target recognition in turnstile channels faces a series of challenges such as light changes and environmental interference. Especially under strong light, backlight, or low-light conditions, traditional target recognition methods are difficult to ensure efficient and accurate recognition effects. This makes the performance of existing image recognition systems significantly affected under complex lighting conditions and unable to meet the actual application requirements.

[0003] The existing technologies have the following deficiencies:

[0004] Although existing turnstile target recognition systems adopt various image processing methods such as light compensation, noise removal, and dynamic range adjustment, these technologies often cannot comprehensively address the challenges under different lighting environments. Especially in the case of strong light differences or rapid changes, the recognition effect of the system is unstable. In addition, most existing target recognition algorithms rely on static feature extraction, are difficult to effectively evaluate the recognizability of images, and do not fully consider the visibility changes of targets in images and the influence under different lighting conditions. Existing methods lack systematic evaluation means for the accuracy and robustness of target recognition, resulting in easy misrecognition or missed recognition problems in complex scenarios. Therefore, there is an urgent need for a new technical solution that can dynamically adjust image preprocessing, feature extraction, and target recognition algorithms to ensure stable and accurate target recognition tasks under various lighting changes. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for recognizing and detecting targets in turnstile channels to solve the problems in the above background.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A method for recognizing and detecting targets in turnstile channels includes the following steps:

[0008] S1: Collect image data of the turnstile channel through a camera, and perform light compensation, noise removal, and high-dynamic range imaging processing on the image to improve the image quality and adapt to target recognition under different lighting conditions;

[0009] S2: Extract the brightness features and texture features of the target region from the preprocessed image. The brightness features are obtained through the analysis of the brightness contrast of the image, and the texture features are obtained through the texture analysis algorithm to describe the brightness difference and texture clarity of the target region;

[0010] S3: Based on the extracted brightness features and texture features, calculate the target visibility coefficient of the image. The target visibility coefficient is used to evaluate the recognizable degree of the target region under different lighting conditions to ensure the visibility of the target region under changing lighting conditions;

[0011] S4: Use the target recognition algorithm to detect the targets in the image, calculate the recognition confidence of each target. The confidence represents the probability that the target is correctly recognized, and evaluate the consistency of the target recognition results;

[0012] S5: Conduct a comprehensive analysis of the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score, which is used to evaluate the final recognition accuracy of the target under different lighting conditions;

[0013] S6: Based on the comprehensive analysis results, ensure fast and accurate target recognition under different lighting conditions by dynamically adjusting the image preprocessing, feature extraction, and target recognition algorithms.

[0014] As a further solution of the present invention: The extraction of the brightness features and texture features of the target region from the preprocessed image specifically includes:

[0015] Extract the brightness information of the image at different resolutions through the multi-scale analysis technology;

[0016] Perform multi-scale decomposition on the image using the Gaussian pyramid and Laplacian pyramid, and extract the brightness contrast information of the image at different scales;

[0017] Based on the local contrast analysis, calculate the brightness difference between the target region and the background region to obtain the brightness features of the target region;

[0018] Use the LBP algorithm to extract the local texture information from the image to obtain the texture encoding of the target region;

[0019] Perform frequency and direction filtering on the image through the Gabor filter, extract the texture structure features of the target region, calculate the overall energy feature value and overall phase feature value of each filtered image to obtain the texture features of the target region.

[0020] As a further solution of the present invention: The calculation of the target visibility coefficient of the image based on the extracted brightness features and texture features specifically includes:

[0021] Analyze the brightness difference between the target area and the background area through the local contrast enhancement algorithm and calculate the brightness feature of the target area based on the brightness difference between the target area and the background area The calculation expression is: ;

[0022] Based on the local contrast analysis result, calculate the brightness feature index of the target area. The calculation expression is: ;

[0023] Among them, represents the brightness of the target area, represents the target area, represents the brightness of the background area, represents the local neighborhood of the target area, represents a constant to avoid division by zero, represents the brightness difference, represents the brightness feature index, represents the target area pixel positions within the surrounding neighborhood area;

[0024] Use the Gabor filter to extract the texture features of the image, including the energy feature and the phase feature and calculate the total values of the energy feature and the phase feature of the filtered image. The calculation expression is: ; ; ; ;

[0025] Among them, represents the response of the Gabor filter, represents the energy feature, represents the phase feature, total value of the energy feature, represents the total value of the phase feature, represents the angular variable, represents the standard deviation of the Gaussian function;

[0026] Calculate the texture feature index. The calculation expression is: ;

[0027] In the formula, represents the texture feature index;

[0028] Through the combination of the energy feature and the phase feature, calculate the texture feature index. The calculation expression is:

[0029] Fuse the comprehensive brightness feature index and the texture feature index to obtain the target visibility coefficient. The calculation expression is: ;

[0030] Among them, is the weight coefficient of brightness, and its calculation expression is: ;

[0031] Among them, is the weight coefficient of texture, and its calculation expression is: ;

[0032] Among them, is the adjustment parameter for controlling the weights of brightness and texture, indicating the target visibility coefficient.

[0033] As a further solution of the present invention: The target in the image is detected by using the target recognition algorithm, and the recognition confidence of each target is calculated, specifically including:

[0034] Extract the features of the target area , including brightness features and texture features;

[0035] Locate the target area through the target detection algorithm and the background area ;

[0036] Calculate the prior probability of the target according to the target features and training data, and the calculation expression is: ;

[0037] In the formula, represents the prior probability of the target, represents the th feature in the target area feature vector, represents the feature of the background area, represents the target area, represents the th target category in the training set, represents that the category is prior probability, is the number of samples;

[0038] Calculate the likelihood function of the target area features, and the calculation expression is: ,

[0039] Among them, represents the likelihood function of the target area features, is the dimension of the feature, is the th feature, represents the th feature mean, represents the th feature standard deviation, Indicates the category of the target feature;

[0040] Calculate the posterior probability of the target, and the calculation expression is: ;

[0041] In the formula, Indicates the posterior probability, Indicates the total observation probability;

[0042] Calculate the confidence coefficient of target recognition, and the calculation expression is: ;

[0043] In the formula, Indicates the confidence coefficient of target recognition, Indicates the preset adjustment coefficient, Indicates the natural logarithm of the base.

[0044] As a further solution of the present invention: The evaluation of the consistency of the target recognition result specifically includes:

[0045] Judge whether the confidence coefficient of the target recognition result is greater than or equal to the preset threshold. If so, it means that the corresponding target recognition result is consistent; if not, it means that the corresponding target recognition result is inconsistent.

[0046] As a further solution of the present invention: The comprehensive analysis of the visibility and recognition confidence of the target to obtain the comprehensive recognition accuracy score specifically includes:

[0047] Obtain the target visibility coefficient and confidence coefficient of the image data of the turnstile channel, perform normalization calculation processing on the target visibility coefficient and confidence coefficient, and calculate the comprehensive recognition accuracy score for evaluating the accuracy of image recognition;

[0048] The calculation expression of the normalization calculation processing is: ;

[0049] In the formula, Indicates the comprehensive recognition accuracy score, and Indicates the preset proportional coefficient, Indicates the confidence coefficient, Indicates the target visibility coefficient.

[0050] As a further solution of the present invention: The evaluation of the final recognition accuracy of the target under different lighting conditions specifically includes:

[0051] Judge whether the comprehensive recognition accuracy score of the target under different lighting conditions is greater than or equal to the preset threshold. If so, it means that the corresponding final recognition is accurate; if not, it means that the corresponding final recognition is inaccurate.

[0052] As a further solution of the present invention: Based on the comprehensive analysis result, by dynamically adjusting the image preprocessing, feature extraction and target recognition algorithms, specifically including:

[0053] When the target visibility coefficient is lower than the preset threshold, increase the intensity of the light compensation algorithm, and automatically adjust the image brightness and contrast to improve the clarity of the target area;

[0054] When the recognition confidence is lower than the preset threshold, dynamically adjust the brightness difference of the target area by increasing the brightness contrast;

[0055] When the texture information of the image is inaccurate, increase the frequency response of the Gabor filter, optimize the texture feature extraction to improve the distinguishability of the target texture.

[0056] A target recognition and detection system for a turnstile channel, including:

[0057] A data acquisition module, which collects turnstile channel image data through a camera, and performs light compensation, noise removal and high-dynamic range imaging processing on the image to improve the image quality and adapt to target recognition under different lighting conditions;

[0058] A feature extraction module, which extracts the brightness features and texture features of the target area from the preprocessed image. The brightness features are obtained through the brightness contrast analysis of the image, and the texture features are obtained through the texture analysis algorithm to describe the brightness difference and texture clarity of the target area;

[0059] A target visibility coefficient calculation module, which calculates the target visibility coefficient of the image based on the extracted brightness features and texture features. The target visibility coefficient is used to evaluate the recognizable degree of the target area under different lighting conditions to ensure the visibility of the target area under changing lighting conditions;

[0060] A confidence recognition module, which uses the target recognition algorithm to detect the target in the image, calculates the recognition confidence of each target. The confidence represents the probability that the target is correctly recognized, and evaluates the consistency of the target recognition result;

[0061] A comprehensive evaluation module, which comprehensively analyzes the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score, which is used to evaluate the final recognition accuracy of the target under different lighting conditions;

[0062] A dynamic regulation module, which based on the comprehensive analysis result, dynamically adjusts the image preprocessing, feature extraction and target recognition algorithms to ensure fast and accurate target recognition under different lighting conditions.

[0063] Advantages of the present invention:

[0064] (1) By comprehensively applying light compensation, noise removal, and high-dynamic range imaging (HDR) techniques, the present invention optimizes the image quality under different lighting conditions in all aspects. First, by dynamically adjusting the exposure time and gain of the camera, it automatically adapts to environmental light changes, avoiding image distortion caused by overexposure or underexposure. Thus, in complex environments such as strong light and backlight, the image can maintain clear details, avoiding difficulties in target recognition due to lighting differences. At the same time, using high-dynamic range imaging technology, by synthesizing images with different exposure degrees, the details of dark and bright areas in the image are enhanced, improving the overall visual quality of the image. In addition, in the image preprocessing stage, combining brightness contrast analysis with advanced texture analysis algorithms (such as local binary pattern LBP and Gabor filtering), the brightness features and texture features of the target area are accurately extracted. Then, through the local contrast enhancement algorithm and Gabor filter, the image texture features are optimized, enabling the target area to still maintain a high visibility coefficient under changing lighting conditions, thereby enhancing the clarity and recognizability of the target area. The combination of this multi-dimensional image processing and feature extraction method improves the image quality, laying a solid foundation for subsequent target recognition and effectively coping with challenges under different lighting conditions;

[0065] (2) The target visibility coefficient precisely combines the brightness contrast and texture features of the image, comprehensively evaluating the recognizability of the target under different lighting conditions, thus providing key information support for subsequent target detection. At the same time, the target recognition algorithm, based on deep learning and training data, accurately locates the target and calculates the recognition confidence coefficient, which reflects the probability of the target being correctly recognized. By combining the target visibility coefficient with the recognition confidence coefficient, the system can dynamically adjust the image preprocessing and target recognition algorithms, ensuring that when the lighting environment changes, target recognition can not only be carried out quickly but also maintain high accuracy. This comprehensive analysis method enables the system to efficiently and accurately complete the target recognition task under complex lighting conditions, greatly improving the efficiency and robustness of target recognition in the turnstile channel, not only reducing the misrecognition rate but also ensuring the stability and efficiency of the entire system, meeting the actual application requirements in various complex environments. Brief Description of the Drawings

[0066] The present invention will be further described below with reference to the drawings.

[0067] Figure 1 It is a specific step flow block diagram of a method for target recognition and detection in a turnstile channel according to the present invention;

[0068] Figure 2 It is a flow block diagram of a target recognition and detection system in a turnstile channel according to the present invention. Detailed implementation manners

[0069] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0070] Please refer to Figure 1 As shown, the present invention is a method for detecting and identifying targets in a turnstile passage, including the following steps:

[0071] S1: Collect image data of the turnstile passage through a camera, and perform light compensation, noise removal, and high-dynamic-range imaging processing on the image to improve the image quality and adapt to target recognition under different lighting conditions;

[0072] S2: Extract the brightness features and texture features of the target area from the preprocessed image. The brightness features are obtained through the analysis of the brightness contrast of the image, and the texture features are obtained through a texture analysis algorithm to describe the brightness difference and texture clarity of the target area;

[0073] S3: Based on the extracted brightness features and texture features, calculate the target visibility coefficient of the image. The target visibility coefficient is used to evaluate the recognizable degree of the target area under different lighting conditions to ensure the visibility of the target area under changing lighting conditions;

[0074] S4: Use a target recognition algorithm to detect the targets in the image, calculate the recognition confidence of each target. The confidence represents the probability that the target is correctly recognized, and evaluate the consistency of the target recognition results;

[0075] S5: Conduct a comprehensive analysis of the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score, which is used to evaluate the final recognition accuracy of the target under different lighting conditions;

[0076] S6: Based on the comprehensive analysis results, ensure fast and accurate target recognition under different lighting conditions by dynamically adjusting image preprocessing, feature extraction, and target recognition algorithms.

[0077] In S1, collecting image data of the turnstile passage through a camera, and performing light compensation, noise removal, and high-dynamic-range imaging processing on the image to improve the image quality and adapt to target recognition under different lighting conditions includes:

[0078] In the process of target recognition in the turnstile channel, real-time image acquisition of the channel is first carried out through a camera. The camera is configured with high resolution to ensure that details of the target area can be clearly captured under complex lighting conditions. During the acquisition process, the exposure time and gain settings of the camera are automatically adjusted according to the lighting changes to ensure that the image will not be distorted due to overexposure or underexposure. At the same time, the dynamic range control configured in the camera can provide relatively balanced image brightness under different lighting conditions for subsequent restoration of image details in the processing stage.

[0079] The acquired images are immediately preprocessed, including light compensation, noise removal, and high dynamic range imaging (HDR) processing. Light compensation eliminates image deviation caused by uneven lighting by adjusting the brightness and contrast of the image, especially improving the image quality under strong light or backlight conditions. Noise removal uses filtering algorithms such as Gaussian filtering or bilateral filtering to remove noise caused by sensor noise, motion blur, or environmental interference. High dynamic range imaging processing reconstructs the high dynamic range of the image by combining multiple images with different exposure levels to ensure that details can still be retained in overly dark or overly bright areas, providing a clear image basis for subsequent target recognition.

[0080] In S2, the brightness features and texture features of the target area are extracted from the preprocessed images. The brightness features are obtained through the brightness contrast analysis of the image, and the texture features are obtained through texture analysis algorithms to describe the brightness difference and texture clarity of the target area, specifically including:

[0081] Through multi-scale analysis technology, the brightness information of the image at different resolutions is extracted;

[0082] The image is decomposed into multiple scales using Gaussian pyramids and Laplacian pyramids, and the brightness contrast information of the image at different scales is extracted;

[0083] Based on local contrast analysis, the brightness difference between the target area and the background area is calculated to obtain the brightness features of the target area;

[0084] The local binary pattern (LBP) algorithm is used to extract local texture information from the image to obtain the texture coding of the target area;

[0085] The image is filtered in terms of frequency and direction through Gabor filters to extract the texture structure features of the target area, and the overall energy feature value and overall phase feature value of each filtered image are calculated to obtain the texture features of the target area.

[0086] In S3, based on the extracted brightness features and texture features, the target visibility coefficient of the image is calculated. The target visibility coefficient is used to evaluate the recognizable degree of the target area under different lighting conditions to ensure the visibility of the target area under changing lighting conditions, specifically including:

[0087] Analyze the brightness difference between the target area and the background area through the local contrast enhancement algorithm and calculate the brightness feature of the target area based on the brightness difference between the target area and the background area The calculation expression is: ;

[0088] Based on the local contrast analysis result, calculate the brightness feature index of the target area. The calculation expression: ;

[0089] Among them, represents the brightness of the target area, represents the target area, represents the brightness of the background area, represents the local neighborhood of the target area, represents a constant to avoid division by zero, represents the brightness difference, represents the brightness feature index, represents the target area pixel positions within the surrounding neighborhood area;

[0090] Use the Gabor filter to extract the texture features of the image, including energy features and phase features and calculate the total values of the energy feature and the phase feature of the filtered image. The calculation expression is: ; ; ; ;

[0091] Among them, represents the response of the Gabor filter, represents the energy feature, represents the phase feature, total value of the energy feature, represents the total value of the phase feature, represents the angular variable, represents the standard deviation of the Gaussian function;

[0092] Calculate the texture feature index. The calculation expression is: ;

[0093] In the formula, represents the texture feature index;

[0094] Through the combination of the energy feature and the phase feature, calculate the texture feature index. The calculation expression is:

[0095] Fuse the comprehensive brightness feature index and the texture feature index to obtain the target visibility coefficient. The calculation expression is as follows: ;

[0096] where, is the weight coefficient of brightness, and the calculation expression is: ;

[0097] where, is the weight coefficient of texture, and the calculation expression is: ;

[0098] where, is the adjustment parameter for controlling the weights of brightness and texture, represents the target visibility coefficient.

[0099] Judge whether the target visibility coefficient of the target area under different lighting conditions is greater than or equal to the preset threshold. If so, it means that the image of the corresponding target area is recognizable; if not, the corresponding target area is affected by light and is unrecognizable.

[0100] It should be noted that: The target visibility coefficient reflects whether the target area of image recognition can be clearly recognized, and when the value of the target visibility coefficient is larger, the corresponding image recognition is more difficult.

[0101] In S4, use the target recognition algorithm to detect the targets in the image, calculate the recognition confidence of each target. The confidence represents the probability that the target is correctly recognized, and evaluate the consistency of the target recognition results, specifically including:

[0102] Extract the features of the target area , including brightness features and texture features;

[0103] Locate the target area through the target detection algorithm and the background area ;

[0104] Calculate the prior probability of the target according to the target features and the training data. The calculation expression is: ;

[0105] In the formula, represents the prior probability of the target, represents the th feature in the target area feature vector, represents the features of the background area, represents the target area, represents the th target category in the training set, represents that the category is prior probability, is the number of samples;

[0106] Calculate the likelihood function of the target region features, and the calculation expression is: ,

[0107] where, represents the likelihood function of the target region features, is the dimension of the feature, is the th feature, represents the th feature mean, represents the th feature standard deviation, represents the category of the target feature;

[0108] Calculate the posterior probability of the target, and the calculation expression is: ;

[0109] In the formula, represents the posterior probability, represents the total observation probability;

[0110] Calculate the confidence coefficient of target recognition, and the calculation expression is: ;

[0111] In the formula, represents the confidence coefficient of target recognition, represents the preset adjustment coefficient, represents the natural logarithm of the base.

[0112] Judge whether the confidence coefficient of the target recognition result is greater than or equal to the preset threshold. If so, it means that the corresponding target recognition results are consistent. If not, it means that the corresponding target recognition results are inconsistent;

[0113] It should be noted that: The confidence coefficient reflects the consistency of the recognition results in the process of target recognition in the target region, and the larger the value of the confidence coefficient, the higher the recognition consistency.

[0114] In S5, comprehensively analyze the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score for evaluating the final recognition accuracy of the target under different lighting conditions. Specifically, it includes:

[0115] Obtain the target visibility coefficient and confidence coefficient of the turnstile channel image data, perform normalization calculation processing on the target visibility coefficient and confidence coefficient, and calculate the comprehensive recognition accuracy score for evaluating the accuracy of image recognition;

[0116] The calculation expression of the normalization calculation processing is: ;

[0117] In the formula, represents the comprehensive recognition accuracy score, and represents the preset proportional coefficient, represents the confidence coefficient, represents the target visibility coefficient.

[0118] It should be noted that: The comprehensive recognition accuracy score reflects the accuracy of target recognition in the image recognition process of the turnstile channel. And when the comprehensive recognition accuracy score is larger, it indicates that the corresponding recognition process is more accurate.

[0119] In S6, based on the comprehensive analysis result, by dynamically adjusting the image preprocessing, feature extraction, and target recognition algorithms, it is ensured to achieve fast and accurate target recognition under different lighting conditions, specifically including:

[0120] According to the changes in the comprehensive recognition accuracy score and the target visibility coefficient, automatically adjust the lighting compensation parameters, noise removal methods, and high dynamic range imaging processing in the image preprocessing steps;

[0121] When the target visibility coefficient is lower than the preset threshold, enhance the intensity of the lighting compensation algorithm, and automatically adjust the image brightness and contrast to improve the clarity of the target area;

[0122] Based on the evaluation results of the target visibility coefficient and the confidence coefficient, automatically adjust the local contrast analysis parameters in the brightness feature extraction. Specifically: When the recognition confidence is low, dynamically adjust the brightness difference of the target area by increasing the brightness contrast;

[0123] When the texture information of the image is inaccurate, increase the frequency response of the Gabor filter, optimize the texture feature extraction, to improve the distinguishability of the target texture;

[0124] After calculating the comprehensive recognition accuracy score, through the feedback mechanism, dynamically adjust the image preprocessing, feature extraction, and target recognition algorithms to achieve self-update of the image preprocessing, feature extraction, and target recognition algorithms.

[0125] Please refer to Figure 2 shown, a target recognition and detection system for a turnstile channel, including:

[0126] A data acquisition module, the data acquisition module collects turnstile channel image data through a camera, and performs lighting compensation, noise removal, and high dynamic range imaging processing on the image to improve the image quality and adapt to target recognition under different lighting conditions;

[0127] Feature extraction module, which extracts the brightness features and texture features of the target area from the preprocessed image. The brightness features are obtained through the analysis of the brightness contrast of the image, and the texture features are obtained through texture analysis algorithms to describe the brightness difference and texture clarity of the target area;

[0128] Target visibility coefficient calculation module, which calculates the target visibility coefficient of the image based on the extracted brightness features and texture features. The target visibility coefficient is used to evaluate the recognizability of the target area under different lighting conditions to ensure the visibility of the target area under changing lighting conditions;

[0129] Confidence recognition module, which uses target recognition algorithms to detect the targets in the image, calculates the recognition confidence of each target. The confidence represents the probability that the target is correctly recognized, and evaluates the consistency of the target recognition results;

[0130] Comprehensive evaluation module, which comprehensively analyzes the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score, used to evaluate the final recognition accuracy of the target under different lighting conditions;

[0131] Dynamic regulation module, which based on the comprehensive analysis results, dynamically adjusts the image preprocessing, feature extraction and target recognition algorithms to ensure fast and accurate target recognition under different lighting conditions.

[0132] The working principle of the present invention: High-resolution images of the turnstile passage are collected in real time through a camera, combined with dynamically adjusted exposure and gain control to ensure the stability of the image quality in complex lighting environments. The collected images are processed by light compensation, noise removal and high-dynamic range imaging to improve the contrast and detail clarity of the images, ensuring that the target area can be clearly recognized under different lighting conditions such as strong light and backlight. Subsequently, techniques such as multi-scale analysis, local contrast analysis and Gabor filtering are used to extract the brightness features and texture features of the target area, and the target visibility coefficient is calculated, which is used to evaluate the recognizability of the target under different lighting conditions. At the same time, the targets are detected through target recognition algorithms and their recognition confidence is calculated, reflecting the consistency of the target recognition results. The target visibility coefficient and the confidence coefficient are comprehensively analyzed to obtain a comprehensive recognition accuracy score, used to evaluate the final recognition accuracy of the target under different lighting conditions. According to the comprehensive analysis results, the parameters of the image preprocessing, feature extraction and target recognition algorithms are dynamically adjusted to achieve fast and accurate target recognition. The present invention can optimize the image processing process under complex lighting conditions, improve the accuracy of target recognition, has strong adaptability and practicability, and is widely applicable to target recognition tasks in intelligent turnstile systems.

[0133] The above formulas are all dimensionless and only take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulations to obtain a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0134] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, or magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0135] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context before and after.

[0136] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0137] The above has described in detail one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. Any equivalent changes and improvements made within the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A gate channel target recognition and detection method, characterized in that: The following steps are involved: S1: Collect gate channel image data through the camera, and perform illumination compensation, noise removal and high dynamic range imaging processing on the image to improve image quality and adapt to target recognition under different lighting conditions; S2: extracting brightness features and texture features of the target area from the preprocessed image, wherein the brightness features are obtained by brightness contrast analysis of the image, and the texture features are obtained by a texture analysis algorithm to describe the brightness difference and texture clarity of the target area; S3: Based on the extracted brightness features and texture features, the target visibility coefficient of the image is calculated. The target visibility coefficient is used to evaluate the degree of recognition of the target area under different lighting conditions and ensure the visibility of the target area under changing lighting conditions; The step of calculating the target visibility coefficient of the image based on the extracted brightness features and texture features specifically includes: Analyze the brightness difference between the target area and the background area through the local contrast enhancement algorithm , and calculate the brightness feature of the target area based on the brightness difference between the target area and the background area , the calculation expression is: ; Based on the local contrast analysis results, the brightness characteristic index of the target area is calculated, and the calculation expression is: ; in, Indicates the brightness of the target area, Indicates the target area, Indicates the brightness of the background area, represents the local neighborhood of the target area, represents a constant to avoid division by zero, Represents the brightness difference, represents the brightness characteristic index, Indicates the target area The pixel position in the surrounding neighborhood area; Use Gabor filter to extract texture features of the image, including energy features and phase characteristics , and calculate the total value of the filtered image energy feature and phase feature. The calculation expression is: ; ; ; ; in, represents the response of the Gabor filter, Represents energy characteristics, Represents the phase characteristics, The overall value of energy characteristics, represents the overall value of phase characteristics, represents the angle variable, represents the standard deviation of the Gaussian function; Calculate the texture feature index, the calculation expression is: ; In the formula, represents the texture feature index; By combining the energy feature and the phase feature, the texture feature index is calculated. The calculation expression is: The comprehensive brightness feature index and the texture feature index are fused to obtain the target visibility coefficient. The calculation expression is: ; in, is the weight coefficient of brightness, and the calculation expression is: ; in, is the weight coefficient of the texture, and the calculation expression is: ; in, To control the adjustment parameters of brightness and texture weight, represents the target visibility coefficient; S4: Detecting the targets in the image using a target recognition algorithm, calculating the recognition confidence of each target, wherein the confidence represents the probability that the target is correctly recognized, and evaluating the consistency of the target recognition results; S5: Comprehensively analyze the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score, which is used to evaluate the final recognition accuracy of the target under different lighting conditions; S6: Based on the comprehensive analysis results, the image preprocessing, feature extraction and target recognition algorithms are dynamically adjusted to ensure fast and accurate target recognition under different lighting conditions.

2. A gate channel target recognition and detection method according to claim 1, characterized in that: The step of extracting the brightness feature and the texture feature of the target area from the preprocessed image specifically includes: Through multi-scale analysis technology, the brightness information of the image at different resolutions is extracted; Use Gaussian pyramid and Laplacian pyramid to decompose the image at multiple scales and extract the brightness contrast information of the image at different scales; Based on local contrast analysis, the brightness difference between the target area and the background area is calculated to obtain the brightness characteristics of the target area; Use LBP algorithm to extract local texture information from the image and obtain the texture coding of the target area; The image is filtered in frequency and direction by Gabor filter to extract the texture structure features of the target area. The overall value of energy features and the overall value of phase features of each filtered image are calculated to obtain the texture features of the target area.

3. A gate channel target recognition and detection method according to claim 1, characterized in that: The evaluation of the identifiability of the target area under different lighting conditions specifically includes: It is determined whether the target visibility coefficient of the target area under different lighting conditions is greater than or equal to a preset threshold. If so, it means that the image of the corresponding target area is recognizable. If not, the corresponding target area is affected by the lighting and cannot be recognized.

4. A gate channel target recognition and detection method according to claim 1, characterized in that: The target recognition algorithm is used to detect the target in the image and calculate the recognition confidence of each target, which specifically includes: Extract features of the target area , including brightness features and texture features; Locate the target area through the target detection algorithm and background area ; Calculate the prior probability of the target based on the target features and training data ; Calculate the likelihood function of the target area features, and the calculation expression is: ; in, represents the likelihood function of the target region features, is the dimension of the feature, It is Features, Indicates The mean of the features, Indicates The standard deviation of the feature, Indicates the category of the target feature; Calculate the posterior probability of the target, the calculation expression is: ; In the formula, represents the posterior probability, represents the total observation probability; Calculate the confidence coefficient of target recognition, the calculation expression is: ; In the formula, represents the confidence coefficient of target recognition, Indicates the preset adjustment coefficient, Represents the logarithm of the base of a natural number.

5. A gate channel target recognition and detection method according to claim 1, characterized in that: The consistency of the evaluation target recognition result specifically includes: It is determined whether the confidence coefficient of the target recognition result is greater than or equal to a preset threshold. If so, it means that the corresponding target recognition results are consistent. If not, it means that the corresponding target recognition results are inconsistent.

6. A gate channel target recognition and detection method according to claim 1, characterized in that: The target visibility and recognition confidence are comprehensively analyzed to obtain a comprehensive recognition accuracy score, which specifically includes: Obtain the target visibility coefficient and confidence coefficient of the gate channel image data, perform normalized calculation on the target visibility coefficient and confidence coefficient, and calculate a comprehensive recognition accuracy score for evaluating the accuracy of image recognition; The calculation expression of the normalized calculation process is: ; In the formula, represents the comprehensive recognition accuracy score, and Indicates the preset scale factor, represents the confidence coefficient, Represents the target visibility coefficient.

7. A gate channel target recognition and detection method according to claim 1, characterized in that: The final recognition accuracy of the evaluation target under different lighting conditions specifically includes: It is determined whether the comprehensive recognition accuracy score of the target under different lighting conditions is greater than or equal to the preset threshold. If so, it means that the corresponding final recognition is accurate; if not, it means that the corresponding final recognition is inaccurate.

8. A gate channel target recognition and detection method according to claim 1, characterized in that: Based on the comprehensive analysis results, the image preprocessing, feature extraction and target recognition algorithms are dynamically adjusted, specifically including: If the target visibility coefficient is lower than the preset threshold, the intensity of the illumination compensation algorithm is enhanced to automatically adjust the image brightness and contrast to improve the clarity of the target area; If the recognition confidence is lower than the preset threshold, the brightness difference of the target area is dynamically adjusted by increasing the brightness contrast; When the texture information of an image is inaccurate, the frequency response of the Gabor filter is increased and the texture feature extraction is optimized to improve the discrimination of the target texture.

9. A gate channel target recognition and detection system, characterized in that: A gate channel target recognition and detection method as claimed in any one of claims 1 to 8, comprising: A data acquisition module, which collects gate channel image data through a camera and performs illumination compensation, noise removal and high dynamic range imaging processing on the image to improve image quality and adapt to target recognition under different lighting conditions; A feature extraction module, wherein the feature extraction module extracts brightness features and texture features of the target area from the preprocessed image, wherein the brightness features are obtained by brightness contrast analysis of the image, and the texture features are obtained by a texture analysis algorithm to describe brightness differences and texture clarity of the target area; A target visibility coefficient calculation module, wherein the target visibility coefficient calculation module calculates the target visibility coefficient of the image based on the extracted brightness features and texture features, and the target visibility coefficient is used to evaluate the identifiability of the target area under different lighting conditions to ensure the visibility of the target area under changing lighting conditions; A confidence recognition module, which detects targets in the image using a target recognition algorithm, calculates the recognition confidence of each target, wherein the confidence represents the probability that the target is correctly recognized, and evaluates the consistency of the target recognition results; A comprehensive evaluation module, which comprehensively analyzes the visibility and recognition confidence of the target to obtain a comprehensive recognition accuracy score for evaluating the final recognition accuracy of the target under different lighting conditions; A dynamic control module, based on the comprehensive analysis results, dynamically adjusts the image preprocessing, feature extraction and target recognition algorithms to ensure fast and accurate target recognition under different lighting conditions.

Images