Camera-based 3D pinna recognition method, system and camera

Abstract

The application relates to the technical field of 3D auricle recognition, in particular to a 3D auricle recognition method and system based on a camera and the camera. The application first acquires ambient light intensity and adjusts near-infrared dot array projector parameters to ensure that a high-quality structured light reflection image is obtained; secondly, a binocular infrared camera is used to collect images and reconstruct an auricle 3D model; then, a historical recognition database is queried to obtain the recognition success rate and spatial distribution characteristics of feature points under different illuminations, and feature point weights are dynamically adjusted accordingly; finally, feature points are extracted on the 3D model, and a matching score is calculated by combining position, definition and weight. The recognition robustness is improved through real-time environmental adaptation and dynamic feature point weight adjustment, the recognition strategy is adaptively optimized by using historical data, and the 3D feature point weighted matching is adopted to fully utilize the auricle anatomical feature difference, so that the recognition accuracy is improved.

Description

Technical Field

[0001] This application relates to the technical field of 3D auricle recognition, and in particular to a camera-based 3D auricle recognition method, system and camera. Background Technology

[0002] With the rapid development of biometric identification technology, its application in fields such as security authentication and identity verification is becoming increasingly widespread. The auricle, as one of the important biometric features of the human body, has advantages such as strong uniqueness, good stability, and convenient data collection, and is gradually becoming a new type of biometric identification method following fingerprints and faces.

[0003] Currently, mainstream ear recognition technologies mainly employ feature extraction and matching methods based on 2D images, or obtain 3D information about the ear through laser scanning. These methods identify individuals by extracting ear contours and texture features, but in practical applications, they are easily affected by external environmental factors such as lighting conditions and shooting angles.

[0004] Existing auricle recognition technology suffers from poor stability in feature point recognition, with significant fluctuations in recognition performance under different lighting conditions, and cannot adaptively optimize for different scenarios; this situation needs further improvement. Summary of the Invention

[0005] To address the issues of poor stability in feature point recognition, significant fluctuations in recognition performance under different lighting conditions, and the inability to adaptively optimize for different scenarios in existing auricular recognition technologies, this application provides a camera-based 3D auricular recognition method, system, and camera, employing the following technical solution: In a first aspect, this application provides a 3D auricle recognition method based on a camera, comprising the following steps: The system acquires the target auricle acquisition requirements and ambient light intensity information. Based on the ambient light intensity information, it automatically adjusts the luminescence parameters of the near-infrared dot matrix projector and determines the distribution information of auricle feature points based on the target auricle acquisition requirements. Near-infrared structured light dot arrays are emitted toward the target auricle according to the luminescence parameters, and reflected images are acquired by a binocular infrared camera. A 3D auricle model is constructed based on the left and right parallax information of the reflected images. Query the auricle feature database to obtain historical recognition information, which includes the success rate of feature point recognition under different lighting conditions and the spatial distribution characteristics of feature points. Based on the historical recognition information, the weight coefficients of the auricular feature points are dynamically adjusted; A preset number of auricular feature points are extracted from the 3D auricular model of the target auricle. The recognizability of each feature point is calculated based on its spatial position and clarity. A weighted matching score is calculated by combining the weight coefficients. The weighted matching score is matched with a preset stored template. When the total weighted matching degree is lower than a preset threshold, the verification is deemed to have failed.

[0006] By adopting the above technical solution, in practical applications, the reflection intensity of structured light varies greatly under different lighting conditions, leading to unstable feature point extraction. Simultaneously, due to significant differences in the anatomical features of different regions of the auricle, uniformly processing all feature points would reduce the reliability of recognition. This application first acquires ambient light intensity information and adjusts the luminescence parameters of the near-infrared dot matrix projector in real time to ensure high-quality structured light reflection images are obtained under different lighting conditions. Second, it uses a binocular infrared camera to acquire reflection images and reconstructs a 3D model of the target auricle using parallax information. Then, it queries a historical recognition database to obtain the recognition success rate and spatial distribution characteristics of feature points under different lighting conditions, and dynamically adjusts the feature point weight coefficients accordingly. Finally, it extracts feature points from the 3D model, calculates a weighted matching score based on spatial location, clarity, and weight coefficients, and performs matching judgment with the stored template. Through real-time environmental adaptation and dynamic weight adjustment of feature points, the robustness of auricle recognition is significantly improved. Through in-depth mining of historical recognition data, adaptive optimization of the recognition strategy is achieved. The weighted matching mechanism of 3D feature points fully utilizes the differences in auricle anatomical features, effectively improving the recognition accuracy.

[0007] Optionally, the recognizability of each feature point can be calculated based on its spatial location and sharpness, specifically including the following steps: Calculate the depth information integrity and point cloud density of the auricular feature points in the 3D model; Based on the reflected image of the auricular feature points, the imaging sharpness and contrast of the structured light dot array are calculated; The recognizability of the auricular feature points is generated by comprehensively considering the completeness of the depth information, point cloud density, imaging clarity, and contrast.

[0008] By adopting the above technical solution, due to the complex surface structure of the auricle and the influence of factors such as occlusion and reflection during 3D reconstruction, the reconstruction quality of feature points at different locations varies significantly. Simply relying on a single index to evaluate the recognizability of feature points is often inaccurate. This application first calculates the completeness of depth information in the region where the feature point is located from the perspective of 3D reconstruction quality, assesses whether there is missing information or reconstruction error, and statistically analyzes the density distribution of the surrounding point cloud to ensure the accuracy of spatial features. Secondly, it analyzes the reflection image of the structured light dot array at the feature point location, calculates the sharpness index of the local area, assesses whether there is blur or noise in the image, and judges the distinguishability of the feature by calculating the contrast between light and dark. Finally, the evaluation indexes of the above four dimensions are weighted and fused to generate a comprehensive recognizability score for the feature point. The quality of the feature point is comprehensively evaluated. The dual evaluation mechanism of 3D and 2D avoids the limitations of a single index. Through comprehensive weighting, the recognizability of the feature point is accurately quantified.

[0009] Optionally, query the auricle feature database to obtain historical recognition information, specifically including the following steps: Based on the current ambient light intensity information, query historical recognition records under similar lighting conditions; Extract the recognition success rate and 3D reconstruction quality data of feature points from the historical recognition records; Based on the anatomical location distribution of feature points, the spatial distribution characteristics of feature points in different regions are statistically analyzed.

[0010] By adopting the above technical solution, in order to fully utilize existing recognition experience and improve system performance, it is necessary to deeply mine historical recognition data. However, existing technologies often simply archive and store historical data without considering the similarity of environmental conditions or the specificity of the auricle's anatomical structure. This application first obtains the current ambient light intensity information and retrieves historical recognition records under similar lighting conditions from the feature database to ensure that the extracted experience data is scene-relevant. Secondly, it performs in-depth analysis on the selected historical records, extracts the recognition success rate of each feature point in this type of scene, and obtains its three-dimensional reconstruction quality index to establish a reliability evaluation benchmark for feature points. Finally, based on the anatomical characteristics of the auricle, the feature points are classified according to their anatomical location, and the spatial distribution characteristics of feature points in different regions are statistically analyzed to reveal the inherent laws of feature point distribution. Through scene relevance analysis, the applicability of historical experience is ensured. Through multi-dimensional data extraction, the quantitative evaluation of feature point performance is achieved.

[0011] Optionally, based on the historical recognition information, the weight coefficients of the auricle feature points are dynamically adjusted, specifically including the following steps: Based on the historical recognition information, the recognition stability of each feature point under different lighting conditions is analyzed, and the reliability index of each feature point is calculated. The basic weight values ​​are determined based on the importance of the anatomical location of the feature points and the historical recognition success rate. The basic weight values ​​are dynamically adjusted based on the identifiability and reliability indicators under the current acquisition conditions. The adjusted weight values ​​are used as the final weight coefficients for the feature points.

[0012] By adopting the above technical solution, the traditional static weight allocation method is difficult to adapt to complex and ever-changing recognition scenarios because the importance of auricular feature points varies significantly in the recognition process, and this variation changes with environmental conditions. For example, some feature points perform well under ideal lighting but may completely fail under strong light interference; anatomical landmarks such as the helix ridge have high distinguishability, but their weights are often fixed and cannot be dynamically adjusted according to the actual acquisition quality. This application first analyzes the recognition performance of each feature point under different lighting conditions based on a historical recognition database, calculates the stability score of the feature point by calculating the fluctuation of the recognition results, and then generates a reliability index. Second, it calculates a basic weight value for each feature point by comprehensively considering the importance of the anatomical location of the feature point and the success rate data in the historical recognition process. Then, it adjusts the basic weight value in real time according to the actual recognizability of the feature point under the current acquisition environment and the calculated reliability index, so that the weight allocation is more in line with the characteristics of the current scenario. Finally, the adjusted value is determined as the final weight coefficient of the feature point. This scheme establishes a feature point reliability evaluation system through historical data analysis; ensures the scientific nature of weight allocation through anatomical structure correlation; and achieves dynamic optimization of weights through real-time status feedback, significantly improving the adaptability and accuracy of feature recognition.

[0013] Optionally, while constructing the 3D auricle model, the method further includes the following steps: The blood oxygen saturation of the target earlobe area is monitored by a blood oxygen sensor integrated into the binocular infrared camera module. Acquire blood oxygen saturation data from multiple consecutive frames and analyze the dynamic changes in blood oxygen saturation. Based on a preset hemodynamic characteristic model, it is determined whether the dynamic change characteristics of the blood oxygen saturation conform to the characteristics of a living organism; The verification fails when the dynamic change characteristics of blood oxygen saturation do not conform to the characteristics of a living organism.

[0014] By adopting the above technical solution, this application first utilizes a blood oxygen sensor integrated into a binocular infrared camera module to continuously monitor the target earlobe area. Due to the rich blood vessels in the earlobe and the stable measurement position, reliable blood oxygen saturation data can be obtained. Secondly, multiple consecutive frames of blood oxygen saturation data are collected, and physiological signals reflecting real blood flow characteristics are obtained by analyzing their dynamic change characteristics. Then, the collected dynamic features are compared with a pre-established hemodynamic feature model, which includes typical characteristics of blood oxygen changes in the normal human earlobe area. Finally, the comparison results are used to determine whether the characteristics conform to those of a living organism. When the detected dynamic change characteristics deviate from the normal physiological range, the system will determine that the verification has failed. By detecting physiological features that are difficult to simulate, the reliability of anti-counterfeiting verification is improved. Through dynamic feature analysis, the defect that static features are easily counterfeited is avoided.

[0015] Optionally, the dynamic changes in blood oxygen saturation can be analyzed, specifically including the following steps: Acquire pulse waveform data in the earlobe area; Calculate the fluctuation amplitude and periodicity of blood oxygen saturation data over multiple consecutive frames; Extract characteristic peaks and valleys from the pulse waveform data; Based on the fluctuation amplitude, periodicity, characteristic peak and trough values, dynamic blood oxygen characteristic indicators are generated.

[0016] By employing the aforementioned technical solutions, simple blood oxygen threshold judgments can be misled by carefully designed electronic devices; similarly, single waveform features are insufficient to effectively describe complex physiological rhythms, easily leading to misjudgments. This application first acquires pulse waveform data from the earlobe region using a blood oxygen sensor to obtain raw signals reflecting local blood flow changes; secondly, it processes continuously acquired multi-frame blood oxygen saturation data, calculates the fluctuation range of the values, and extracts their periodic variation characteristics through time-domain analysis. These indicators can reflect the normal physiological rhythm of the cardiovascular system; then, it performs peak detection on the acquired pulse waveforms, extracting characteristic peaks and troughs in each cardiac cycle. These feature points contain rich hemodynamic information; finally, it fuses features from multiple dimensions such as fluctuation amplitude, periodicity, peak values, and trough values ​​to generate a comprehensive dynamic blood oxygen feature index; comprehensively characterizing the dynamic characteristics of blood oxygen changes, providing a more reliable basis for in vivo detection.

[0017] Optionally, determining whether the dynamic change characteristics of the blood oxygen saturation conform to the characteristics of a living organism specifically includes the following steps: Obtain the preset reference range of normal human earlobe blood oxygen saturation and pulse waveform feature template; The blood oxygen dynamic characteristic index is matched with the pulse waveform characteristic template; When the blood oxygen saturation data does not fall within the reference range, or when the matching degree between the blood oxygen dynamic characteristic index and the pulse waveform characteristic template is lower than a preset threshold, it is determined that it does not meet the characteristics of a living organism.

[0018] By adopting the above technical solution, this application first obtains the reference range of normal human earlobe blood oxygen saturation from a preset physiological parameter database. This range takes into account physiological variations in different ages, genders, and physical conditions. At the same time, it retrieves pulse waveform feature templates optimized for different populations. These templates contain standard waveform morphology and dynamic change characteristics. Secondly, it performs multi-dimensional matching between the real-time collected blood oxygen dynamic feature indicators and the feature templates. By calculating indicators such as waveform similarity and period matching degree, it comprehensively evaluates the degree of conformity between the current data and normal physiological characteristics. Finally, based on a dual judgment mechanism, it checks whether the blood oxygen saturation falls within a reasonable range and verifies whether the matching degree between the dynamic feature indicators and the template reaches a preset threshold. If either condition is not met, it is judged as a non-living feature. This effectively reduces the risk of misjudgment and missed judgment.

[0019] Secondly, this application provides a camera-based 3D auricle recognition system, comprising: The acquisition module is used to obtain information on the target auricle acquisition requirements and ambient light intensity. The light emission control module is used to automatically adjust the light emission parameters of the near-infrared dot matrix projector based on the ambient light intensity information. The feature planning module is used to determine the distribution information of auricular feature points according to the target auricular acquisition requirements; The image acquisition module includes a near-infrared dot projector and a binocular infrared camera. The near-infrared dot projector is used to emit near-infrared structured light dot arrays towards the target auricle according to the emission parameters, and the binocular infrared camera is used to acquire reflected images. The 3D reconstruction module is used to construct a 3D auricle model based on the left and right parallax information of the reflected image; The feature recognition module is used to extract a preset number of auricular feature points from the 3D auricular model of the target auricle; The feature evaluation module is used to calculate the recognizability of each feature point based on its spatial location and clarity. The matching determination module is used to calculate a weighted matching score by combining weight coefficients, and match the weighted matching score with a preset storage template. When the total weighted matching degree is lower than a preset threshold, the verification is determined to be unsuccessful.

[0020] Optionally, the feature evaluation module includes: A depth information evaluation unit is used to calculate the depth information integrity and point cloud density of the auricular feature points in the 3D model. An image quality assessment unit is used to calculate the imaging sharpness and contrast of the structured light dot array based on the reflected image of the auricular feature points; The recognizability generation unit is used to generate the recognizability of the auricular feature points based on the completeness of the depth information, point cloud density, imaging clarity and contrast.

[0021] Thirdly, this application provides a camera, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described camera-based 3D auricle recognition method.

[0022] In summary, this application includes at least one of the following beneficial technical effects: This application first acquires ambient light intensity information and adjusts the luminescence parameters of the near-infrared dot matrix projector in real time to ensure high-quality structured light reflection images under different lighting conditions. Second, it uses a binocular infrared camera to acquire reflection images and reconstructs a 3D model of the target auricle using parallax information. Then, it queries a historical recognition database to obtain the recognition success rate and spatial distribution characteristics of feature points under different lighting conditions, and dynamically adjusts the feature point weight coefficients accordingly. Finally, it extracts feature points from the 3D model, calculates a weighted matching score based on spatial location, sharpness, and weight coefficients, and performs a matching judgment with a stored template. Through real-time environmental adaptation and dynamic weight adjustment of feature points, the robustness of auricle recognition is significantly improved. Through in-depth mining of historical recognition data, adaptive optimization of the recognition strategy is achieved. The weighted matching mechanism of 3D feature points fully utilizes the differences in auricle anatomical features, effectively improving recognition accuracy. Due to the complex surface structure of the auricle and the influence of factors such as occlusion and reflection during 3D reconstruction, the reconstruction quality of feature points at different locations varies significantly. Simply relying on a single indicator to evaluate the recognizability of feature points is often inaccurate. This application first calculates the completeness of depth information in the region where the feature point is located from the perspective of 3D reconstruction quality, assesses whether there is missing information or reconstruction error, and statistically analyzes the density distribution of the surrounding point cloud to ensure the accuracy of spatial features. Second, it analyzes the reflection image of the structured light dot array at the feature point location, calculates the sharpness index of the local area, assesses whether there is blur or noise in the image, and judges the distinguishability of the feature by calculating the contrast between light and dark. Finally, the evaluation indicators of the above four dimensions are weighted and fused to generate a comprehensive recognizability score for the feature point. This comprehensively evaluates the quality of the feature point. It adopts a dual evaluation mechanism of 3D and 2D to avoid the limitations of a single indicator. Through a comprehensive weighting method, it achieves accurate quantification of the recognizability of the feature point. This application first obtains the current ambient light intensity information and retrieves historical recognition records under similar lighting conditions from the feature database to ensure that the extracted experience data is scene-relevant. Secondly, it performs in-depth analysis on the selected historical records, extracting the recognition success rate of each feature point in that type of scene, and simultaneously obtaining its 3D reconstruction quality indicators to establish a reliability evaluation benchmark for the feature points. Finally, based on the anatomical characteristics of the auricle, the feature points are classified according to their anatomical location, and the spatial distribution characteristics of feature points in different regions are statistically analyzed to reveal the inherent patterns in feature point distribution. Scene relevance analysis ensures the applicability of historical experience; and multi-dimensional data extraction enables a quantitative evaluation of feature point performance. Attached Figure Description

[0023] Figure 1 This is a schematic flowchart of a camera-based 3D auricle recognition method according to an embodiment of this application; Figure 2 This is a flowchart illustrating step S300 in a camera-based 3D auricle recognition method according to an embodiment of this application. Figure 3 This is a flowchart illustrating step S400 in a camera-based 3D auricle recognition method according to an embodiment of this application. Figure 4 This is a flowchart illustrating step S500 in a camera-based 3D auricle recognition method according to an embodiment of this application. Figure 5 This is a schematic diagram of the process for monitoring blood oxygen saturation in a camera-based 3D auricle recognition method according to an embodiment of this application; Figure 6 This is a flowchart illustrating step S700 in a camera-based 3D auricle recognition method according to an embodiment of this application. Figure 7 This is a flowchart illustrating step S800 in a camera-based 3D auricle recognition method according to an embodiment of this application. Figure 8 This is a schematic diagram of a camera-based 3D auricle recognition system according to an embodiment of this application; Figure 9 This is an internal structural diagram of a camera according to an embodiment of this application. Detailed Implementation

[0024] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0025] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0026] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0027] Firstly, this application provides a 3D auricle recognition method based on a camera, referring to... Figure 1 It includes the following steps: S100: Obtain the target auricle acquisition requirements and ambient light intensity information; automatically adjust the luminescence parameters of the near-infrared dot matrix projector based on the ambient light intensity information; and determine the distribution information of auricle feature points based on the target auricle acquisition requirements.

[0028] In this embodiment, the target auricle acquisition requirements refer to the conditional parameters that need to be met during the acquisition process, including auricle imaging size, number of feature points, spatial resolution, and depth of field. Ambient light intensity information refers to the ambient light intensity value of the current scene, obtained through measurement by the photosensor built into the camera module. The emission parameters of the near-infrared dot matrix projector include three adjustable parameters: emission power, pulse width, and projection angle. The auricle feature point distribution information includes the required number of feature points and anatomical position constraints.

[0029] Specifically, the system first acquires the ambient light intensity using a photosensor and then queries a preset mapping table of light intensity and luminescence parameters. This mapping table divides the ambient light intensity into three intervals: weak light, normal light, and strong light, and sets corresponding combinations of luminescence power and pulse width for each interval. The projection angle is fixed based on a preset auricular imaging area. Simultaneously, the number of feature points to be extracted is determined according to preset auricular feature acquisition standards. These feature points are mainly distributed in the helix, antihelix, scaphoid fossa, and cymba conchae. The system maintains a feature point anatomical location distribution table, recording the reference anatomical location of each feature point. This ensures high-quality structured light patterns can be obtained under different lighting conditions, while guaranteeing that the feature points are distributed in anatomically critical areas.

[0030] S200: Based on the luminescence parameters, a near-infrared structured light dot array is emitted towards the target auricle. The reflected images are collected by a binocular infrared camera, and a 3D auricle model is constructed based on the left and right parallax information of the reflected images.

[0031] In this embodiment, the near-infrared structured light dot matrix refers to a regular dot matrix pattern formed by near-infrared light passing through a diffraction grating. The binocular infrared camera consists of two infrared cameras with a fixed baseline distance. Left and right parallax information refers to the positional difference of the same feature point on the imaging planes of the left and right cameras. The 3D auricle model refers to the three-dimensional point cloud data of the auricle surface reconstructed using the principle of triangulation.

[0032] Specifically, after the system starts, the near-infrared projector projects a structured light dot matrix onto the target auricle according to the adjusted emission parameters. The binocular cameras simultaneously acquire images using preset acquisition parameters. The system maintains a camera calibration parameter table to establish the transformation relationship between the image coordinate system and the world coordinate system. Feature matching is performed on the acquired left and right images, a disparity map is generated using the disparity calculation module, and the three-dimensional coordinates of feature points are calculated using triangulation methods in conjunction with the camera calibration parameter table, ultimately generating a 3D auricle model.

[0033] S300. Query the auricle feature database to obtain historical recognition information, which includes the success rate of feature point recognition and the spatial distribution characteristics of feature points under different lighting conditions.

[0034] In this embodiment, the auricular feature database refers to a structured dataset that stores historical recognition records, containing two basic data tables: a feature recognition result table and a spatial distribution characteristic table. The feature recognition result table records the recognition status information of each feature point under different lighting conditions. The spatial distribution characteristic table records the distribution patterns of feature points in different anatomical regions. Historical recognition information refers to empirical data extracted from the database to guide the current recognition process.

[0035] Specifically, the system first queries the feature recognition result table for historical data under similar lighting conditions based on the current ambient light intensity. The feature recognition result table uses a multi-level index structure: the first level index is the light intensity range, the second level index is the anatomical region type, and the third level index is the feature point number. The system extracts the feature point recognition success rate data from the table and simultaneously queries the spatial distribution characteristic table to obtain the typical distribution patterns of feature points in each anatomical region. The spatial distribution characteristic table uses a region mapping method to divide the auricle into multiple anatomical regions, recording the distribution density and relative positional relationships of feature points in each region. The feature point recognition success rate under different lighting conditions can be used to determine which feature points are easy to extract and which are prone to failure under the current lighting environment. For example, some feature points are often not accurately recognized under strong light, so their importance should be reduced in strong light environments. The spatial distribution characteristics of feature points can be used to determine the inherent reliability of feature points. For example, feature points on the edge of the helix are relatively stable under various conditions due to their clear outlines; while feature points within the concha are greatly affected by shadows and have lower reliability.

[0036] S400: Based on historical recognition information, dynamically adjust the weight coefficients of auricular feature points.

[0037] In this embodiment, the weight coefficient is a numerical parameter reflecting the importance of feature points, used to assign different levels of importance to different feature points during subsequent matching. Dynamic adjustment refers to the process of updating the weight coefficient in real time based on the current recognition environment and historical recognition experience.

[0038] Specifically, the system maintains a weight calculation mapping table, which defines the conversion rules from feature point performance indicators to weight coefficients. First, the basic weight values ​​are determined based on success rate data from historical recognition information, and then adjusted according to the importance of the anatomical region where the feature point is located. The system updates the weight coefficients using a weighted summation method, and assigns importance coefficients to different influencing factors through a reliability scoring table.

[0039] S500. Extract a preset number of auricular feature points from the 3D auricular model of the target auricle. Calculate the recognizability of each feature point based on its spatial location and clarity. Calculate a weighted matching score by combining the weight coefficients. Match the weighted matching score with a preset stored template. If the total weighted matching score is lower than a preset threshold, the verification is deemed to have failed.

[0040] In this embodiment, the preset number of auricular feature points refers to the number of feature points to be extracted, determined according to the recognition accuracy requirements. Recognizability refers to the clarity and stability score of the feature points under the current acquisition conditions. The weighted matching score refers to the overall matching result after comprehensively considering the feature point weights. The stored template refers to the pre-collected auricular feature data of registered users.

[0041] Specifically, the system first extracts a predetermined number of feature points from the reconstructed 3D auricle model based on a feature point distribution table. It then calculates the recognizability of each feature point by querying a feature scoring mapping table, which defines the conversion rules from image quality parameters to recognizability scores. The recognizability score is multiplied by a dynamic weighting coefficient to obtain a weighted score for each feature point. The system maintains a matching threshold table, setting corresponding threshold requirements based on different application scenarios. When the total weighted score is lower than the threshold obtained from the table lookup, the system determines that the verification fails.

[0042] In one embodiment, refer to Figure 2 In step S300, the auricle feature database is queried to obtain historical recognition information, specifically including the following steps: S310. Based on the current ambient light intensity information, query historical identification records under similar lighting conditions.

[0043] In this embodiment, similar lighting conditions refer to historical scenes with the same light intensity range as the current environment. Historical recognition records refer to the result data of previous recognition processes stored in the database, including three basic attributes: lighting information, recognition result, and recognition time.

[0044] Specifically, the recognition effect of the current environment largely depends on the lighting conditions. By querying historical records under similar lighting conditions, potential problems can be predicted. The system divides the ambient light intensity into multiple ranges. First, it reads the light intensity value collected by the current photosensor and determines its range by looking up a table. The system uses a multi-level index structure to store historical records, establishing a lighting range index table to quickly locate historical data under similar lighting conditions. By setting a lighting similarity threshold table, the range of historical records that can be used for reference is determined. The system establishes a time decay table, assigning different reference weights to historical records in different time periods.

[0045] S320, the recognition success rate and 3D reconstruction quality data of feature points extracted from historical recognition records.

[0046] In this embodiment, the recognition success rate refers to the probability that a feature point is accurately located during the historical recognition process. 3D reconstruction quality data refers to the completeness of depth information and point cloud density records during the historical reconstruction process of feature points.

[0047] Feature points at different locations exhibit significant differences in performance: depth sensors show larger measurement errors in black or highly reflective areas, while remaining relatively stable in areas with rough surfaces. By analyzing the recognition success rate and reconstruction quality of each feature point in historical data, the most reliable combination of feature points can be identified. For example, feature points at the helix, due to their clear outlines, typically have high recognition success rates and reconstruction quality.

[0048] S330. Based on the anatomical location distribution of feature points, statistically analyze the spatial distribution characteristics of feature points in different regions.

[0049] In this embodiment, anatomical location distribution refers to the positional information of feature points in different anatomical regions of the auricle. Spatial distribution characteristics refer to the distribution pattern of feature point groups in three-dimensional space.

[0050] Specifically, the anatomical structure of the auricle determines the regularity of the distribution of feature points. For example, the junction of the helix and antihelix forms significant anatomical landmarks, and the feature point groups at these locations exhibit stable spatial distribution characteristics; while feature points in concave areas such as the concha are easily affected by posture changes, resulting in relatively unstable spatial distribution characteristics. The system establishes an auricle anatomical region mapping table, dividing the auricle into multiple anatomical regions. By maintaining a regional feature statistics table, typical distribution patterns of feature points within each region are recorded. The system constructs a spatial relationship analysis table to assess the relative positional constraints between feature points. A sub-regional statistical method is used to establish an independent feature distribution model for each anatomical region. The system maintains a feature density reference table, recording the standard feature point density requirements for different regions. By establishing a positional constraint table, the spatial positional constraint relationships between feature points are defined.

[0051] In one embodiment, refer to Figure 3 In step S400, the weight coefficients of the auricle feature points are dynamically adjusted based on historical recognition information, specifically including the following steps: S410. Based on historical recognition information, analyze the recognition stability of each feature point under different lighting conditions, and calculate the reliability index of each feature point.

[0052] In this embodiment, recognition stability refers to the ability of feature points to remain reliably recognizable under varying lighting conditions. The reliability index refers to the feature point confidence score obtained by analyzing historical recognition performance. Variations in lighting conditions refer to the transition between different levels of ambient light intensity.

[0053] Specifically, the system first establishes an illumination transformation matrix to record the differences in recognition results between adjacent illumination levels. The system then constructs a feature point stability index table, calculating the dispersion of recognition results under different illumination conditions to obtain a stability score. Simultaneously, a reliability calculation rule table is established to convert the stability score into a reliability index.

[0054] S420. Determine the basic weight value based on the importance of the anatomical location of the feature points and the historical recognition success rate.

[0055] In this embodiment, the importance of anatomical location refers to the contribution level of the auricular region where the feature point is located to the recognition process. The base weight value refers to the initial weight value in the weight calculation process. The historical recognition success rate refers to the proportion of accurate recognition of the feature point in the historical recognition process.

[0056] Specifically, the system establishes an anatomical region importance table, assigning importance coefficients to different auricular anatomical regions. It maintains a recognition success rate statistics table to record the historical recognition accuracy of feature points. The system constructs a weight initialization rule table, converting importance coefficients and recognition success rates into basic weight values. A region-based weighting method is used to calculate weights for the helix, antihelix, scaphoid, and cymba conchae regions separately. The system establishes a weight standardization table to ensure reasonable weight allocation for different feature points.

[0057] S430: Based on the identifiability and reliability indicators under the current acquisition conditions, dynamically adjust the basic weight values.

[0058] In this embodiment, dynamic adjustment refers to the process of updating the weights based on real-time acquisition conditions. Acquisition conditions refer to the various factors affecting feature point recognition in the current environment, including three main aspects: illumination intensity, imaging quality, and reconstruction accuracy.

[0059] Specifically, the system maintains a weight adjustment rule table, defining the mapping relationship from acquisition conditions to weight correction values. An acquisition quality evaluation table is established to comprehensively assess the impact of current acquisition conditions on feature point recognition. The system constructs a weight update matrix, obtaining the adjusted weights by multiplying the base weight values ​​by correction coefficients. A tiered adjustment mechanism is adopted, determining the adjustment magnitude based on different combinations of identifiability and reliability indicators.

[0060] S440. Use the adjusted weight values ​​as the final weight coefficients for the feature points.

[0061] In this embodiment, the final weight coefficient refers to the final weight value used in the feature point matching process. Weight value determination refers to the process of completing weight calculation and fixing the weight parameters.

[0062] Specifically, the system establishes a weight normalization table to ensure that the adjusted weight values ​​meet the normalization requirements. A weight validity verification table is maintained to verify the rationality of the final weight coefficients. A weight allocation record table is constructed to store process data for each weight adjustment. A batch update mechanism is adopted to update the weight coefficient database after completing one round of identification.

[0063] In one embodiment, refer to Figure 4 In step S500, the recognizability of each feature point is calculated based on its spatial location and clarity. This includes the following steps: S510. Calculate the depth information integrity and point cloud density of auricular feature points in the 3D model.

[0064] In this embodiment, depth information integrity refers to the degree of coverage of the effective depth values ​​obtained in the 3D reconstruction process of the area surrounding the feature point. Point cloud density refers to the number of effective point clouds per unit area within the neighborhood of the feature point. The neighborhood of a feature point refers to the local area centered on the feature point.

[0065] Specifically, the system maintains a depth evaluation mapping table, which divides the integrity of depth information into multiple levels. First, the neighborhood of a feature point is determined, and the number of pixels with valid depth values ​​within this range is counted. An integrity score is then obtained by looking up the table. Simultaneously, the system establishes a point cloud density reference table, recording the standard point cloud density requirements for different anatomical regions. A density score is obtained by calculating the ratio of the actual point cloud density within the feature point's neighborhood to the standard density. The system employs a region-based evaluation method, scoring the helix, antihelix, scaphoid fossa, and cymba conchae separately. Finally, a weighted combination is performed by consulting a region weight table to obtain the depth reconstruction quality score for that feature point.

[0066] S520. Based on the reflection image of auricular feature points, calculate the imaging sharpness and contrast of the structured light dot array.

[0067] In this embodiment, image sharpness refers to the image sharpness level of the structured light dot array at the feature point location. Contrast refers to the degree of brightness difference between the structured light dots and the background. Reflected image refers to the image captured by the binocular camera after the structured light illuminates the auricle.

[0068] Specifically, the system constructs an image quality assessment table, containing two sub-tables: a sharpness calculation rule and a contrast scoring standard. For sharpness assessment, the system extracts image gradient information within the neighborhood of feature points and maps the gradient values ​​to sharpness scores by looking up the table. For contrast assessment, the system calculates the brightness difference between structured light points and the local background and obtains the contrast score by looking up the table. The system maintains an illumination compensation table to correct the scores under different ambient lighting conditions. Finally, by querying the feature fusion table, the sharpness score and contrast score are combined to form a comprehensive image quality score.

[0069] S530: Based on the completeness of depth information, point cloud density, imaging clarity, and contrast, the recognizability of auricular feature points is comprehensively generated.

[0070] Specifically, the system first inputs the depth reconstruction quality score and the overall image quality score into a fusion table to obtain an initial recognizability value. The system maintains a feature location weight table, adjusting the recognizability based on the anatomical location of the feature points. A tiered scoring mechanism is employed, dividing recognizability into multiple levels, and a quality threshold table is consulted to determine whether a feature point meets the recognition requirements. The system periodically updates the scoring rules and optimizes the parameter configuration of the fusion table by analyzing historical recognition data.

[0071] In one embodiment, refer to Figure 5 In addition to constructing the 3D auricle model, the method also includes the following steps: The S600 monitors the blood oxygen saturation of the target earlobe area using a blood oxygen sensor integrated into a binocular infrared camera module.

[0072] Among them, a blood oxygen sensor refers to a photoelectric sensor device used to measure the oxygen saturation of hemoglobin. Blood oxygen saturation refers to the percentage of oxygen-bound hemoglobin in the blood relative to total hemoglobin. The target earlobe area refers to the tissue area below the auricle that is rich in capillaries.

[0073] Specifically, the system establishes a blood oxygen measurement parameter table and configures the wavelength combination and drive current of the light-emitting diodes. By maintaining a measurement area positioning table, it ensures the sensor is aligned with the densely populated capillary area of ​​the earlobe. The system constructs a signal acquisition rule table, defining the sampling frequency and sampling duration. Blood oxygen measurement is performed using the dual-wavelength ratio method, and a photoelectric signal conversion table is established to convert the photoelectric signal into a blood oxygen saturation value.

[0074] S700: Acquire blood oxygen saturation data from multiple consecutive frames and analyze the dynamic changes in blood oxygen saturation.

[0075] Among them, continuous multi-frame data refers to a set of blood oxygen saturation measurements acquired within a specified time window.

[0076] Specifically, the system maintains a data acquisition configuration table, specifying the sampling time window and sampling interval. A signal feature extraction table is established to extract fluctuation characteristics from the raw data. The system constructs a periodic analysis rule table to identify periodic patterns in blood oxygenation changes.

[0077] S800: Based on a preset hemodynamic characteristic model, determine whether the dynamic change characteristics of blood oxygen saturation conform to the characteristics of a living organism.

[0078] Among them, the hemodynamic characteristic model refers to the mathematical model describing the laws governing changes in blood oxygenation in the human body. In vivo characteristics refer to the typical physiological characteristics exhibited by living tissues during changes in blood oxygenation. The pre-defined model refers to a standard characteristic model obtained through training with a large amount of experimental data.

[0079] Specifically, the system establishes a liveness feature reference table to record the standard range of blood oxygen changes in normal human bodies. It defines the matching method between measured data and the standard model by maintaining a feature matching rule table. The system constructs a decision threshold table to set the critical conditions for liveness determination. A multi-feature fusion approach is used for liveness determination, and a feature weight allocation table is established to determine the importance of each feature.

[0080] S900. When the dynamic change characteristics of blood oxygen saturation do not conform to the characteristics of a living organism, the verification is deemed to have failed.

[0081] Specifically, verification failure refers to a state where the currently collected data does not meet the requirements for liveness verification. Failure to meet liveness characteristics means that blood oxygenation changes exceed the preset normal physiological range.

[0082] In one embodiment, refer to Figure 6 In step S700, the dynamic changes in blood oxygen saturation are analyzed, specifically including the following steps: S710: Acquire pulse waveform data in the earlobe area.

[0083] Specifically, it is the process of acquiring pulse signals using photoplethysmography.

[0084] S720 Calculate the fluctuation amplitude and periodicity of blood oxygen saturation data over multiple consecutive frames.

[0085] Specifically, fluctuation range refers to the range of variation in blood oxygen saturation values. Periodicity refers to the recurring pattern of changes in blood oxygen saturation. Continuous multi-frame data refers to a sequence of measurements collected within a fixed time window.

[0086] S730: Extract characteristic peaks and valleys from pulse waveform data.

[0087] Specifically, the system maintains a feature point sequence table, recording the peak and valley value sequence within a continuous period.

[0088] S740 generates dynamic blood oxygenation characteristic indicators based on fluctuation amplitude, periodicity, characteristic peaks and troughs.

[0089] Among them, the dynamic blood oxygenation characteristic index refers to a comprehensive evaluation index that describes the characteristics of blood oxygenation changes.

[0090] Specifically, the system determines the importance of different features in the comprehensive index by establishing a feature weight allocation table. It then constructs an index standardization table to convert features of different dimensions into a unified scale, and finally establishes an index scoring table to classify and evaluate the feature indicators.

[0091] In one embodiment, refer to Figure 7 In step S800, determining whether the dynamic changes in blood oxygen saturation conform to in vivo characteristics specifically includes the following steps: S810: Obtain the preset reference range of normal human earlobe blood oxygen saturation and pulse waveform feature template.

[0092] Specifically, the system establishes a blood oxygenation reference range table, recording standard blood oxygenation intervals for different age groups and physical conditions. It maintains a waveform feature library, storing various typical pulse waveform templates. The system also constructs a reference data update table, periodically adjusting the reference standards based on newly added clinical data.

[0093] S820: Match the dynamic blood oxygenation characteristic indicators with the pulse waveform characteristic template.

[0094] S830. When the blood oxygen saturation data does not fall within the reference range, or the matching degree between the blood oxygen dynamic characteristic index and the pulse waveform characteristic template is lower than the preset threshold, it is determined that it does not meet the characteristics of a living body.

[0095] Specifically, if the blood oxygen saturation data does not fall within the reference range, or if the matching degree between the blood oxygen dynamic characteristic index and the pulse waveform characteristic template is lower than the preset threshold, it is determined that it does not meet the characteristics of a living organism.

[0096] Specifically, the system establishes a threshold table, sets threshold requirements for different detection scenarios, and categorizes different types of verification failures. When blood oxygen saturation falls within the danger zone, the system will trigger an emergency alert; when the matching degree is close to but does not reach the threshold, the system will require repeated measurement. An emergency handling table is maintained to specify handling procedures for different abnormal situations.

[0097] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0098] Secondly, this application provides a camera-based 3D auricle recognition system. The camera-based 3D auricle recognition system of this application will be described below in conjunction with the above-mentioned camera-based 3D auricle recognition method.

[0099] Reference Figure 8 A camera-based 3D auricle recognition system includes: The acquisition module is used to obtain information on the target auricle acquisition requirements and ambient light intensity. The light emission control module is used to automatically adjust the light emission parameters of the near-infrared dot matrix projector based on the ambient light intensity information. The feature planning module is used to determine the distribution information of auricular feature points according to the target auricular acquisition requirements; The image acquisition module includes a near-infrared dot projector and a binocular infrared camera. The near-infrared dot projector is used to emit near-infrared structured light dot arrays toward the target auricle according to the emission parameters, and the binocular infrared camera is used to acquire reflected images. The 3D reconstruction module is used to construct a 3D auricle model based on the left and right parallax information of the reflected image; The feature recognition module is used to extract a preset number of auricular feature points from the 3D auricular model of the target auricle; The feature evaluation module is used to calculate the recognizability of each feature point based on its spatial location and clarity. The matching determination module is used to calculate a weighted matching score by combining the weight coefficients, and then match the weighted matching score with a preset stored template. When the total weighted matching degree is lower than the preset threshold, the verification is determined to be unsuccessful.

[0100] In one embodiment, the feature evaluation module includes: The depth information evaluation unit is used to calculate the depth information integrity and point cloud density of auricular feature points in the 3D model. The image quality assessment unit is used to calculate the imaging sharpness and contrast of the structured light dot array based on the reflected image of the auricle feature points; The recognizability generation unit is used to comprehensively generate the recognizability of auricular feature points based on the integrity of depth information, point cloud density, imaging clarity, and contrast.

[0101] In one embodiment, this application provides a camera whose internal structure diagram can be as follows: Figure 9 As shown, the camera includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The camera's database stores data. The network interface allows communication with external terminals via a network connection. When executed by the processor, the computer program implements a camera-based 3D auricle recognition method.

[0102] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the solution of this application and does not constitute a limitation on the camera to which the solution of this application is applied. A specific camera may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0103] In one embodiment, a camera is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0104] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0105] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A 3D auricle recognition method based on a camera, characterized in that, Includes the following steps: The system acquires the target auricle acquisition requirements and ambient light intensity information. Based on the ambient light intensity information, it automatically adjusts the luminescence parameters of the near-infrared dot matrix projector and determines the distribution information of auricle feature points based on the target auricle acquisition requirements. Near-infrared structured light dot arrays are emitted toward the target auricle according to the luminescence parameters, and reflected images are acquired by a binocular infrared camera. A 3D auricle model is constructed based on the left and right parallax information of the reflected images. Query the auricle feature database to obtain historical recognition information, which includes the success rate of feature point recognition under different lighting conditions and the spatial distribution characteristics of feature points. Based on the historical recognition information, the weight coefficients of the auricular feature points are dynamically adjusted; A preset number of auricular feature points are extracted from the 3D auricular model of the target auricle. The recognizability of each feature point is calculated based on its spatial position and clarity. A weighted matching score is calculated by combining the weight coefficients. The weighted matching score is matched with a preset stored template. When the total weighted matching degree is lower than a preset threshold, the verification is deemed to have failed.

2. The 3D auricle recognition method based on a camera according to claim 1, characterized in that, The identifiability of each feature point is calculated based on its spatial location and sharpness, specifically including the following steps: Calculate the depth information integrity and point cloud density of the auricular feature points in the 3D model; Based on the reflected image of the auricular feature points, the imaging sharpness and contrast of the structured light dot array are calculated; The recognizability of the auricular feature points is generated by comprehensively considering the completeness of the depth information, point cloud density, imaging clarity, and contrast.

3. The 3D auricle recognition method based on a camera according to claim 1, characterized in that, The process of querying the auricle feature database to obtain historical recognition information includes the following steps: Based on the current ambient light intensity information, query historical recognition records under similar lighting conditions; Extract the recognition success rate and 3D reconstruction quality data of feature points from the historical recognition records; Based on the anatomical location distribution of feature points, the spatial distribution characteristics of feature points in different regions are statistically analyzed.

4. The 3D auricle recognition method based on a camera according to claim 3, characterized in that, Based on the historical recognition information, the weight coefficients of the auricular feature points are dynamically adjusted, specifically including the following steps: Based on the historical recognition information, the recognition stability of each feature point under different lighting conditions is analyzed, and the reliability index of each feature point is calculated. The basic weight values ​​are determined based on the importance of the anatomical location of the feature points and the historical recognition success rate. The basic weight values ​​are dynamically adjusted based on the identifiability and reliability indicators under the current acquisition conditions. The adjusted weight values ​​are used as the final weight coefficients for the feature points.

5. The 3D auricle recognition method based on a camera according to claim 1, characterized in that, While constructing the 3D auricle model, the method also includes the following steps: The blood oxygen saturation of the target earlobe area is monitored by a blood oxygen sensor integrated into the binocular infrared camera module. Acquire blood oxygen saturation data from multiple consecutive frames and analyze the dynamic changes in blood oxygen saturation. Based on a preset hemodynamic characteristic model, it is determined whether the dynamic change characteristics of the blood oxygen saturation conform to the characteristics of a living organism; The verification fails when the dynamic change characteristics of blood oxygen saturation do not conform to the characteristics of a living organism.

6. The 3D auricle recognition method based on a camera according to claim 5, characterized in that, Analyzing the dynamic changes in blood oxygen saturation involves the following steps: Acquire pulse waveform data in the earlobe area; Calculate the fluctuation amplitude and periodicity of blood oxygen saturation data over multiple consecutive frames; Extract characteristic peaks and valleys from the pulse waveform data; Based on the fluctuation amplitude, periodicity, characteristic peak and trough values, dynamic blood oxygen characteristic indicators are generated.

7. The 3D auricle recognition method based on a camera according to claim 6, characterized in that, Determining whether the dynamic changes in blood oxygen saturation conform to in vivo characteristics includes the following steps: Obtain the preset reference range of normal human earlobe blood oxygen saturation and pulse waveform feature template; The blood oxygen dynamic characteristic index is matched with the pulse waveform characteristic template; When the blood oxygen saturation data does not fall within the reference range, or when the matching degree between the blood oxygen dynamic characteristic index and the pulse waveform characteristic template is lower than a preset threshold, it is determined that it does not meet the characteristics of a living organism.

8. A camera-based 3D auricle recognition system, characterized in that, include: The acquisition module is used to obtain information on the target auricle acquisition requirements and ambient light intensity. The light emission control module is used to automatically adjust the light emission parameters of the near-infrared dot matrix projector based on the ambient light intensity information. The feature planning module is used to determine the distribution information of auricular feature points according to the target auricular acquisition requirements; The image acquisition module includes a near-infrared dot projector and a binocular infrared camera. The near-infrared dot projector is used to emit near-infrared structured light dot arrays towards the target auricle according to the emission parameters, and the binocular infrared camera is used to acquire reflected images. The 3D reconstruction module is used to construct a 3D auricle model based on the left and right parallax information of the reflected image; The feature recognition module is used to extract a preset number of auricular feature points from the 3D auricular model of the target auricle; The feature evaluation module is used to calculate the recognizability of each feature point based on its spatial location and clarity. The matching determination module is used to calculate a weighted matching score by combining weight coefficients, and match the weighted matching score with a preset storage template. When the total weighted matching degree is lower than a preset threshold, the verification is determined to be unsuccessful.

9. The camera-based 3D auricle recognition system according to claim 8, characterized in that, The feature evaluation module includes: A depth information evaluation unit is used to calculate the depth information integrity and point cloud density of the auricular feature points in the 3D model. An image quality assessment unit is used to calculate the imaging sharpness and contrast of the structured light dot array based on the reflected image of the auricular feature points; The recognizability generation unit is used to generate the recognizability of the auricular feature points based on the completeness of the depth information, point cloud density, imaging clarity and contrast.

10. A camera having a computer program stored thereon, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the camera-based 3D auricle recognition method according to any one of claims 1-7.

Images