Visitor registration all-in-one machine based on face recognition

Abstract

The invention relates to the technical field of visitor management, in particular to a visitor registration all-in-one machine based on face recognition. The background analysis module is used for determining a background area state according to a background number reference value and a background distance reference value of a background area of the face image of the person to be recognized; the background processing module is used for selecting a background region image adjusting mode according to the background region state of the face image of the person to be recognized; the face analysis module is used for determining the state of a pre-estimated face area according to the face number reference value and the distance deviation reference value of the pre-estimated face area of the face image of the person to be recognized; the face processing module is used for selecting an image adjusting mode of an estimated face area according to the state of the estimated face area of the face image of the person to be recognized; and a matching module. According to the method, the face target in a complex environment can be stably and accurately recognized, and the accuracy of facial feature extraction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of visitor management, and in particular to an all-in-one visitor registration machine based on face recognition. Background Art

[0002] Face recognition technology faces many challenges. Face images captured in complex environments are often affected by factors such as blur and noise, which may lead to mutations in local pixel values, and then misjudge the mutation features as key information of the face, seriously affecting the accuracy and stability of face target recognition. Therefore, how to deal with the impact of local pixel value mutations on face images and achieve stable and accurate recognition of face targets in complex environments is a technical problem that technicians in this field need to solve urgently.

[0003] Chinese patent publication number CN107680213A discloses a visitor registration system based on face recognition, including an access control system, a face recognition system and a property management system; the access control system is equipped with a face recognition system; the face recognition system includes a camera, an image processing module and an image storage module; the image processing module is connected to the camera and the image storage module respectively; the property management system is connected to the access control system and the image storage module respectively through a communication module. It can be seen that the above technical solution has the following problems: no measures are taken to stably and accurately identify face targets in complex environments, and the accuracy of facial feature extraction is not improved. Summary of the invention

[0004] To this end, the present invention provides a visitor registration integrated machine based on face recognition to overcome the problem in the prior art that no measures are taken to stably and accurately identify face targets in complex environments and the accuracy of facial feature extraction is not improved.

[0005] To achieve the above object, the present invention provides a visitor registration integrated machine based on face recognition, comprising: A collection module, used to collect and store facial images of persons to be identified; A background analysis module, which is connected to the acquisition module and is used to determine the background area state according to a background quantity reference value and a background distance reference value of the background area of ​​the face image of the person to be identified; A background processing module, which is connected to the background analysis module, and is used to select a background area image adjustment method according to the background area state of the face image of the person to be identified, such as performing a sub-region pooling process on the background area or cutting the face image of the person to be identified; A face analysis module, which is connected to the acquisition module and is used to determine the estimated face area state according to a face quantity reference value and a distance deviation reference value of the estimated face area of ​​the face image of the person to be identified; A face processing module connected to the face analysis module is used to select an estimated face region image adjustment mode according to the estimated face region state of the face image of the person to be identified, such as adjusting the face image acquisition frequency of the person to be identified or performing pooling processing on the estimated face region; A matching module is connected to the background processing module and the face processing module, and is used to match the face image after pooling processing with the pre-stored face image, and when the face image matches the pre-stored face image successfully, the access time of the person to be identified and the face image corresponding to the access time are stored.

[0006] Furthermore, the background processing module determines a background area image adjustment method according to the background area state, and the background area image adjustment method includes a first background area image adjustment method for performing regional pooling processing on the background area, and a second background area image adjustment method for cutting a facial image of a person to be identified.

[0007] Furthermore, the background analysis module determines the background area state according to the background quantity reference value and the background distance reference value, the background area state including a first background area state in which the background quantity reference value is greater than a preset background quantity reference value or the background distance reference value is less than a preset background distance reference value, and a second background area state in which the background quantity reference value is less than or equal to the preset background quantity reference value and the background distance reference value is greater than or equal to the preset background distance reference value.

[0008] Furthermore, the background processing module responds to the first background area image adjustment mode, and performs regional pooling processing on the background area, including: The pooling window area of ​​a single sub-region is adjusted according to the recognition influence coefficient of the mutation pixel point of a single sub-region; If the recognition influence coefficient of the mutation pixel point in a single sub-region is within the first preset recognition influence coefficient range, the pooling window area of ​​the single sub-region is reduced and adjusted; If the recognition influence coefficient of the mutation pixel point in a single sub-region is within the second preset recognition influence coefficient range, the pooling window area of ​​the single sub-region is increased and adjusted; The pooling window area of ​​the single sub-region is positively correlated with the mutation pixel recognition influence coefficient of the single sub-region; The calculation formula of the recognition influence coefficient ε of the mutation pixel point of the single sub-region is: , where m is the reference value of the number of mutation pixels in a single sub-region, α is the weight coefficient of the reference value of the number of mutation pixels in a single sub-region, n is the distribution uniformity of each mutation pixel in a single sub-region, and β is the weight coefficient of the distribution uniformity of each mutation pixel in a single sub-region.

[0009] Furthermore, the background processing module responds to the second background area image adjustment mode, and performs segmentation on the face image of the person to be identified, including: Identify the shortest distance from each mutation pixel point in the background area to the estimated face area, and record the minimum value of the shortest distance as the reference distance, establish a similar face area with the same contour as the estimated face area and an area larger than the estimated face area, and the shortest distance from each point of the similar face area contour to the estimated face area is equal to the reference distance, and cut the area outside the similar face area in the face image of the person to be identified, and the face image after cutting is the image of the similar face area in the initial face image.

[0010] Furthermore, the face analysis module determines the estimated face region state according to the face quantity reference value and the distance deviation reference value; The estimated face area status includes a first estimated face area status in which the face number reference value is less than or equal to a preset face number reference value and the distance deviation reference value is less than or equal to a preset distance deviation reference value, and a second estimated face area status in which the face number reference value is greater than the preset face number reference value or the distance deviation reference value is greater than the preset distance deviation reference value.

[0011] Furthermore, the face processing module determines an estimated face region image adjustment method according to the estimated face region state; If the estimated face region state is the first estimated face region state, the estimated face region image adjustment mode is the first estimated face region image adjustment mode for increasing the frequency of face image acquisition of the person to be identified; If the estimated face region state is the second estimated face region state, the estimated face region image adjustment method is the second estimated face region image adjustment method of performing pooling processing on the estimated face region; The relationship between the increase value of the facial image acquisition frequency of the person to be identified and the reference value of the number of faces is positively correlated.

[0012] Furthermore, the face processing module responds to the second estimated face region image adjustment mode, and performs pooling processing on the estimated face region, including: calculating the discreteness of each mutation pixel in the estimated face region, and increasing or decreasing the step size of the pooling window according to the discreteness of each mutation pixel; The calculation formula of the dispersion γ is: , where N is the number of mutation pixels, is the minimum value of the shortest distances from the i-th mutation pixel to each of the adjacent mutation pixels corresponding to the mutation pixel; The discreteness of each mutation pixel point is negatively correlated with the pooling window step size.

[0013] Furthermore, the acquisition module includes a camera, a transmitter and an image storage.

[0014] Furthermore, the acquisition module determines the initial frequency of acquiring facial images of the person to be identified according to the actual environmental impact coefficient; The relationship between the actual environmental impact coefficient and the initial image acquisition frequency is a positive correlation; The calculation formula of the actual environmental impact coefficient P is: ,in is the current light intensity coefficient, is the weight coefficient of the current light intensity, is the current dust level, is the weight coefficient of the current dustiness.

[0015] Compared with the prior art, the beneficial effect of the present invention lies in that, in the technical scheme of the present invention, the background analysis module determines the background area state according to the background quantity reference value and the background distance reference value, and effectively reflects the distribution of mutation pixel points in the background area in the face image of the current person to be identified through the background area state, and the background processing module adaptively selects the background area image adjustment method, so that the selection of the background area image adjustment method is more in line with the actual working scene, avoiding the problem of inaccurate face target recognition caused by the mutation pixel points in the background area having too much influence on the estimated face area, thereby improving the accuracy of image processing.

[0016] Furthermore, the background processing module in the present invention performs regional pooling processing on the background area, and effectively reflects the mutation pixel situation of the current single sub-region through the recognition influence coefficient of the mutation pixel point corresponding to the single sub-region, and flexibly adjusts the pooling window area of ​​the single sub-region, so that the pooling window can effectively meet the pooling requirements of the actual image, improve the pooling effect, and thereby improve the efficiency of face recognition.

[0017] Furthermore, the background processing module in the present invention performs cutting on the facial image of the person to be identified, and by determining the similar facial area, ensures that the area outside the similar facial area in the facial image of the person to be identified can be cut instead of being pooled, thereby avoiding the interference of mutant pixels in the cut area on facial recognition, thereby improving the accuracy of image processing.

[0018] Furthermore, the face analysis module in the present invention determines the estimated face area state according to the face quantity reference value and the distance deviation reference value, and effectively reflects the mutation pixel points of the estimated face area in the face image of the person to be identified through the estimated face area state, and the face processing module adaptively selects the estimated face area image adjustment method, so that the selection of the estimated face area image adjustment method is more in line with the actual working scene, avoiding the problem, and thus improving the image processing efficiency.

[0019] Furthermore, the face processing module in the present invention increases the frequency of facial image acquisition of the person to be identified. By increasing the image acquisition frequency, more facial images can be captured, avoiding the situation where the face cannot be accurately identified when there are fewer mutant pixels, thereby ensuring the stability and accuracy of face recognition.

[0020] Furthermore, the face processing module in the present invention performs pooling processing on the estimated face area, calculates the discreteness of each mutation pixel point in the estimated face area, and adaptively adjusts the pooling window step size, thereby ensuring the processing accuracy in the area with dense mutation pixels and the processing efficiency in the area with sparse mutation pixels. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a module connection diagram of the visitor registration integrated machine based on face recognition of the present invention; Figure 2 A flow chart of determining a background area state according to a background quantity reference value and a background distance reference value in the present invention; Figure 3 A flow chart of the present invention for determining a background area image adjustment method according to a background area state; Figure 4 The present invention is a flow chart of determining an estimated face region image adjustment method according to an estimated face region state. DETAILED DESCRIPTION

[0022] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0023] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0024] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0025] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0026] See also Figures 1 to 4 As shown, the present invention provides a visitor registration integrated machine based on face recognition, comprising: A collection module, used to collect and store facial images of persons to be identified; A background analysis module, which is connected to the acquisition module and is used to determine the background area state according to a background quantity reference value and a background distance reference value of the background area of ​​the face image of the person to be identified; A background processing module, which is connected to the background analysis module, and is used to select a background area image adjustment method according to the background area state of the face image of the person to be identified, such as performing a sub-region pooling process on the background area or cutting the face image of the person to be identified; A face analysis module, which is connected to the acquisition module and is used to determine the estimated face area state according to a face quantity reference value and a distance deviation reference value of the estimated face area of ​​the face image of the person to be identified; A face processing module connected to the face analysis module is used to select an estimated face region image adjustment mode according to the estimated face region state of the face image of the person to be identified, such as adjusting the face image acquisition frequency of the person to be identified or performing pooling processing on the estimated face region; A matching module is connected to the background processing module and the face processing module, and is used to match the face image after pooling processing with the pre-stored face image, and when the face image matches the pre-stored face image successfully, the access time of the person to be identified and the face image corresponding to the access time are stored.

[0027] The application scenario of the present invention is face recognition in the visitor registration process. The face image of the person to be identified includes a background area and an estimated face area. The method for confirming the estimated face area is: the area with the largest area among the minimum rectangular areas that can contain a complete face in the face images of each successfully identified person to be identified in the historical recognition records, and it is ensured that the center of the minimum circumscribed circle of the area with the largest area coincides with the center of the minimum circumscribed circle of the face image of the person to be identified; the background area is other areas in the face image of the person to be identified excluding the estimated face area; the background analysis module can realize the recognition of the background area and the estimated face area in the face image of the person to be identified through machine vision and deep learning networks. This is content that is easy for technicians in this field to understand and will not be elaborated here.

[0028] For a single pixel, if the minimum pixel value difference between the pixel and the pixel adjacent to it is greater than the preset difference, the pixel is a mutation pixel. The preset difference value can be set by the user according to the actual situation. The greater the user's demand for image processing accuracy, the smaller the preset difference value is. A preset difference value is provided, and the preset difference is the minimum value of the minimum pixel value difference between each mutation pixel and the pixel adjacent to the corresponding mutation pixel in the historical recognition process.

[0029] The pre-stored facial images are stored in a database corresponding to the identity information of the personnel. The present invention uses a cosine similarity algorithm to compare the features of the facial images after the pooling process with the features in the pre-stored facial image library. According to the result of the matching algorithm, one or more pre-stored facial images with the most similar features to the facial images after the pooling process are selected as candidate recognition results. When the similarity between the facial images after the pooling process and one or some of the pre-stored facial images is greater than the preset similarity, it is considered that the facial images after the pooling process are successfully matched with the pre-stored facial images. When the facial images are successfully matched with the pre-stored facial images, the access time of the person to be identified and the facial image corresponding to the access time are stored. This is content that is easy for technicians in this field to understand and will not be elaborated here. The value of the preset similarity can be set by the user according to the user's needs. The greater the user's demand for facial recognition accuracy, the smaller the preset similarity value is. A preset similarity value is provided. The preset similarity is the average value of the similarity between the facial images of the person to be identified that were successfully identified in the historical recognition process and the corresponding pre-stored facial images.

[0030] Specifically, the background processing module determines a background area image adjustment method according to the background area status. The background area image adjustment method includes a first background area image adjustment method for performing regional pooling processing on the background area, and a second background area image adjustment method for cutting a facial image of a person to be identified.

[0031] Specifically, the background analysis module determines the background area state according to the background quantity reference value and the background distance reference value, the background area state including a first background area state in which the background quantity reference value is greater than a preset background quantity reference value or the background distance reference value is less than a preset background distance reference value, and a second background area state in which the background quantity reference value is less than or equal to the preset background quantity reference value and the background distance reference value is greater than or equal to the preset background distance reference value.

[0032] When the background area state is the first background area state, the background area image adjustment mode is the first background area image adjustment mode; when the background area state is the second background area state, the background area image adjustment mode is the second background area image adjustment mode.

[0033] The reference value of the background quantity is the number of mutation pixels in the background area of ​​the face image of the person to be identified; Identify the shortest distances from each mutation pixel point in the background area of ​​the face image of the person to be identified to the estimated face area, and the shortest distance among the shortest distances is the background distance reference value; The values ​​of the preset background quantity reference value and the preset background distance reference value can be set by the user according to actual conditions. The greater the user's demand for image processing accuracy, the smaller the value of the preset background quantity reference value and the larger the value of the preset background distance reference value. In order to reduce the influence of the background area on face image recognition, a preset background quantity reference value and a preset background distance reference value are provided. The preset background quantity reference value is the average value of the number of mutation pixels in the background area of ​​the face image of the person to be recognized that can be normally recognized in the historical face recognition process. The preset background distance reference value is the average value of the background distance reference values ​​of the background area in the face image of the person to be recognized that can be normally recognized in the historical face recognition process.

[0034] Specifically, the background processing module responds to the first background area image adjustment mode, and performs regional pooling processing on the background area, including: The pooling window area of ​​a single sub-region is adjusted according to the recognition influence coefficient of the mutation pixel point of a single sub-region; If the recognition influence coefficient of the mutation pixel point in a single sub-region is within the first preset recognition influence coefficient range, the pooling window area of ​​the single sub-region is reduced and adjusted; If the recognition influence coefficient of the mutation pixel point in a single sub-region is within the second preset recognition influence coefficient range, the pooling window area of ​​the single sub-region is increased and adjusted; The pooling window area of ​​the single sub-region is positively correlated with the mutation pixel recognition influence coefficient of the single sub-region; The calculation formula of the recognition influence coefficient ε of the mutation pixel point of the single sub-region is: , where m is the reference value of the number of mutation pixels in a single sub-region, α is the weight coefficient of the reference value of the number of mutation pixels in a single sub-region, n is the distribution uniformity of each mutation pixel in a single sub-region, and β is the weight coefficient of the distribution uniformity of each mutation pixel in a single sub-region.

[0035] Extract any two connected sides of the facial image of the person to be identified, record the longer side of the two sides as the long side, and the shorter side as the short side, divide the long side of the facial image of the person to be identified into A equal parts to obtain A-1 equal parts, divide the short side of the facial image of the person to be identified into B equal parts to obtain B-1 equal parts, then connect the equal parts points of the same order corresponding to the two long sides in pairs, and connect the equal parts points of the same order corresponding to the two short sides in pairs, so as to divide the background area into several sub-areas, wherein the values ​​of A and B can be set by the user according to the facial image of the person to be identified, and provide a value of A and B, A=10, B=6, select the maximum pixel value in each sub-area as the representative value of the sub-area during the pooling process for each sub-area, and then calculate the average value of the representative value of each sub-area to replace each pixel value in the background area of ​​the facial image of the person to be identified, so as to effectively smooth the noise and mutation in the image and reduce the interference of local pixel value changes on subsequent processing. The pooling window area preliminarily designed in the present invention is 2×2, which is easily understood by those skilled in the art and will not be elaborated here.

[0036] The values ​​within the range of the first preset recognition influence coefficient are all greater than the first preset recognition influence coefficient, and the values ​​within the range of the second preset recognition influence coefficient are all less than the second preset recognition influence coefficient. The values ​​of the first preset recognition influence coefficient and the second preset recognition influence coefficient can be determined by the user based on historical usage records. The greater the user's demand for face recognition accuracy, the smaller the value of the first preset recognition influence coefficient, and the larger the value of the second preset recognition influence coefficient. A first preset recognition influence coefficient and a second preset recognition influence coefficient are provided. The first preset recognition influence coefficient and the second preset recognition influence coefficient are respectively the maximum and minimum values ​​of the recognition influence coefficients of the mutation pixels in each sub-region that can meet the face recognition accuracy when the pooling window area is 2×2 in the historical recognition process.

[0037] The weight coefficient α of the reference value of the number of mutation pixels in a single sub-region and the weight coefficient β of the uniformity of distribution of each mutation pixel in a single sub-region are both determined in advance by the user. The user can freely set them according to actual needs. Here, one value of α and β is provided: α=0.5, β=0.5.

[0038] Specifically, the background processing module responds to the second background area image adjustment mode, and performs segmentation on the face image of the person to be identified, including: Identify the shortest distance from each mutation pixel point in the background area to the estimated face area, and record the minimum value of the shortest distance as the reference distance, establish a similar face area with the same contour as the estimated face area and an area larger than the estimated face area, and the shortest distance from each point of the similar face area contour to the estimated face area is equal to the reference distance, and cut the area outside the similar face area in the face image of the person to be identified, and the face image after cutting is the image of the similar face area in the initial face image.

[0039] The estimated face region contour is the estimated boundary shape of the face region.

[0040] Specifically, the face analysis module determines the estimated face region state according to the face quantity reference value and the distance deviation reference value; The estimated face area status includes a first estimated face area status in which the face number reference value is less than or equal to a preset face number reference value and the distance deviation reference value is less than or equal to a preset distance deviation reference value, and a second estimated face area status in which the face number reference value is greater than the preset face number reference value or the distance deviation reference value is greater than the preset distance deviation reference value.

[0041] The reference value of the number of faces is the number of mutation pixels in the estimated face area in the face image of the person to be identified; the face image in the historical recognition record that can perform face recognition normally and has the same estimated face area in the face image of the person to be identified is recorded as the historical face image, the estimated face area in the face image of the person to be identified is overlapped with the estimated face area of ​​the kth historical face image, and the sum of the distances from each mutation pixel in the estimated face area in the face image of the person to be identified to the nearest mutation pixel in the estimated face area in the kth historical face image is calculated, and the sum of the distances is recorded as , then the calculation formula of the distance deviation reference value σ is: , where Z is the number of historical face images and k is the kth historical face image.

[0042] The values ​​of the preset face quantity reference value and the preset distance deviation reference value can be set by the user according to actual conditions. The higher the face recognition accuracy required by the user, the larger the value of the preset face quantity reference value and the smaller the value of the preset distance deviation reference value. A value of a preset face quantity reference value and a preset distance deviation reference value are provided. The preset face quantity reference value is the minimum value of the mutation pixel points in the estimated face area in the face image of the person to be identified that can be normally identified in the historical face recognition process. The preset distance deviation reference value is the maximum value of the distance deviation reference values ​​of each historical face image.

[0043] Specifically, the face processing module determines the estimated face region image adjustment method according to the estimated face region state; If the estimated face region state is the first estimated face region state, the estimated face region image adjustment mode is the first estimated face region image adjustment mode for increasing the frequency of face image acquisition of the person to be identified; If the estimated face region state is the second estimated face region state, the estimated face region image adjustment method is the second estimated face region image adjustment method of performing pooling processing on the estimated face region; The relationship between the increase value of the facial image acquisition frequency of the person to be identified and the reference value of the number of faces is positively correlated.

[0044] The frequency of facial image acquisition of the person to be identified is the number of times the camera in the acquisition module captures facial images per unit time (such as per second).

[0045] Specifically, the face processing module responds to the second estimated face region image adjustment mode, and performs pooling processing on the estimated face region, including: calculating the discreteness of each mutation pixel point in the estimated face region, and increasing or decreasing the step size of the pooling window according to the discreteness of each mutation pixel point; The calculation formula of the dispersion γ is: , where N is the number of mutation pixels, is the minimum value of the shortest distances from the i-th mutation pixel to each of the adjacent mutation pixels corresponding to the mutation pixel; The discreteness of each mutation pixel point is negatively correlated with the pooling window step size.

[0046] When the discreteness of each mutation pixel point is within the first preset discreteness range, the step size of the pooling window is increased; when the discreteness of each mutation pixel point is within the second preset discreteness range, the step size of the pooling window is decreased; The values ​​within the first preset discreteness range are all greater than the first preset discreteness, and the values ​​within the second preset discreteness range are all less than the second preset discreteness. The values ​​of the first preset discreteness and the second preset discreteness can be determined by the user based on historical usage records. The greater the user's demand for face recognition accuracy, the smaller the value of the first preset discreteness and the larger the value of the second preset discreteness. A first preset discreteness and a second preset discreteness are provided. The first preset discreteness and the second preset discreteness are respectively the maximum value and the minimum value of the discreteness of each mutation pixel point in the estimated face area in each historical face image.

[0047] It can be understood that the first preset discreteness is greater than the second preset discreteness.

[0048] Specifically, the acquisition module includes a camera, a transmitter and an image storage.

[0049] The acquisition module uses a camera to capture the facial image of the person to be identified in real time. There is no restriction on the camera model used. What needs to be ensured is that the camera can clearly capture the contour and feature points of the face, and the transmitter can transmit the information of the acquisition module to other units. The model and position of the transmitter are not specifically limited. The user makes adaptive settings according to the actual scene and actual needs. There is no specific restriction on the model of the image storage device, and the user can choose it according to the actual situation. The method of using the camera, transmitter and image storage device used in the present invention is easy to understand for technicians in this field and will not be elaborated here.

[0050] Specifically, the acquisition module determines the initial frequency of acquiring facial images of the person to be identified according to the actual environmental impact coefficient; The relationship between the actual environmental impact coefficient and the initial image acquisition frequency is a positive correlation; The calculation formula of the actual environmental impact coefficient P is: ,in is the current light intensity coefficient, is the weight coefficient of the current light intensity, is the current dust level, is the weight coefficient of the current dustiness.

[0051] The current light intensity is measured by a light intensity meter, providing a current light intensity coefficient When the light intensity is greater than or equal to 100 lux and less than or equal to 900 lux, the current light intensity coefficient =0, when the light intensity is less than 100lux or greater than 900lux, the current light intensity coefficient =1, current dust level The measurement is carried out by a dust meter. There is no specific limit on the model and quantity of infrared sensors, light intensity measuring instruments and dust meters. Customers can choose according to actual conditions. It is understandable that the weight coefficient of the current light intensity And the weight coefficient of the current dustiness All are determined by the user in advance, and the user can freely set them according to actual needs, providing a and The value of =0.5, =0.5.

[0052] The acquisition module determines the initial facial image acquisition frequency of the person to be identified based on the actual environmental impact coefficient. When the actual environmental impact coefficient is large, effective facial information may not be obtained. Increasing the initial image acquisition frequency can capture more facial details and improve the accuracy of portrait recognition.

[0053] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A visitor registration machine based on face recognition, characterized in that: include: A collection module, used to collect and store facial images of persons to be identified; A background analysis module, which is connected to the acquisition module and is used to determine the background area state according to a background quantity reference value and a background distance reference value of the background area of ​​the face image of the person to be identified; A background processing module, which is connected to the background analysis module, and is used to select a background area image adjustment method according to the background area state of the face image of the person to be identified, such as performing a sub-region pooling process on the background area or cutting the face image of the person to be identified; A face analysis module, which is connected to the acquisition module and is used to determine the estimated face area state according to a face quantity reference value and a distance deviation reference value of the estimated face area of ​​the face image of the person to be identified; A face processing module connected to the face analysis module is used to select an estimated face region image adjustment mode according to the estimated face region state of the face image of the person to be identified, such as adjusting the face image acquisition frequency of the person to be identified or performing pooling processing on the estimated face region; A matching module is connected to the background processing module and the face processing module, and is used to match the face image after pooling processing with the pre-stored face image, and when the face image matches the pre-stored face image successfully, the access time of the person to be identified and the face image corresponding to the access time are stored.

2. The visitor registration integrated machine based on face recognition according to claim 1 is characterized in that: The background processing module determines a background area image adjustment method according to the background area state. The background area image adjustment method includes a first background area image adjustment method for performing regional pooling processing on the background area, and a second background area image adjustment method for cutting a facial image of a person to be identified.

3. The visitor registration integrated machine based on face recognition according to claim 2 is characterized in that: The background analysis module determines the background area state according to the background quantity reference value and the background distance reference value, the background area state including a first background area state in which the background quantity reference value is greater than a preset background quantity reference value or the background distance reference value is less than a preset background distance reference value, and a second background area state in which the background quantity reference value is less than or equal to the preset background quantity reference value and the background distance reference value is greater than or equal to the preset background distance reference value.

4. The visitor registration integrated machine based on face recognition according to claim 3 is characterized in that: The background processing module responds to the first background area image adjustment mode, and performs regional pooling processing on the background area, including: The pooling window area of ​​a single sub-region is adjusted according to the recognition influence coefficient of the mutation pixel point of a single sub-region; If the recognition influence coefficient of the mutation pixel point in a single sub-region is within the first preset recognition influence coefficient range, the pooling window area of ​​the single sub-region is reduced and adjusted; If the recognition influence coefficient of the mutation pixel point in a single sub-region is within the second preset recognition influence coefficient range, the pooling window area of ​​the single sub-region is increased and adjusted; The pooling window area of ​​the single sub-region is positively correlated with the mutation pixel recognition influence coefficient of the single sub-region; The calculation formula of the recognition influence coefficient ε of the mutation pixel point of the single sub-region is: , where m is the reference value of the number of mutation pixels in a single sub-region, α is the weight coefficient of the reference value of the number of mutation pixels in a single sub-region, n is the distribution uniformity of each mutation pixel in a single sub-region, and β is the weight coefficient of the distribution uniformity of each mutation pixel in a single sub-region.

5. The visitor registration integrated machine based on face recognition according to claim 3 is characterized in that: The background processing module responds to the second background area image adjustment mode, and performs segmentation on the face image of the person to be identified, including: Identify the shortest distance from each mutation pixel point in the background area to the estimated face area, and record the minimum value of the shortest distance as the reference distance, establish a similar face area with the same contour as the estimated face area and an area larger than the estimated face area, and the shortest distance from each point of the similar face area contour to the estimated face area is equal to the reference distance, and cut the area outside the similar face area in the face image of the person to be identified, and the face image after cutting is the image of the similar face area in the initial face image.

6. The visitor registration integrated machine based on face recognition according to claim 3 is characterized in that: The face analysis module determines the estimated face area state according to the face quantity reference value and the distance deviation reference value; The estimated face area status includes a first estimated face area status in which the face number reference value is less than or equal to a preset face number reference value and the distance deviation reference value is less than or equal to a preset distance deviation reference value, and a second estimated face area status in which the face number reference value is greater than the preset face number reference value or the distance deviation reference value is greater than the preset distance deviation reference value.

7. The visitor registration integrated machine based on face recognition according to claim 6 is characterized in that: The face processing module determines an estimated face region image adjustment method according to the estimated face region state; If the estimated face region state is the first estimated face region state, the estimated face region image adjustment mode is the first estimated face region image adjustment mode for increasing the frequency of face image acquisition of the person to be identified; If the estimated face region state is the second estimated face region state, the estimated face region image adjustment method is the second estimated face region image adjustment method of performing pooling processing on the estimated face region; The relationship between the increase value of the facial image acquisition frequency of the person to be identified and the reference value of the number of faces is positively correlated.

8. The visitor registration integrated machine based on face recognition according to claim 6 is characterized in that: The face processing module responds to the second estimated face region image adjustment mode, and performs pooling processing on the estimated face region, including: calculating the discreteness of each mutation pixel point in the estimated face region, and increasing or decreasing the step size of the pooling window according to the discreteness of each mutation pixel point; The calculation formula of the dispersion γ is: , where N is the number of mutation pixels, is the minimum value of the shortest distances from the i-th mutation pixel to each of the adjacent mutation pixels corresponding to the mutation pixel; The discreteness of each mutation pixel point is negatively correlated with the pooling window step size.

9. The visitor registration integrated machine based on face recognition according to claim 1 is characterized in that: The acquisition module includes a camera, a transmitter and an image storage.

10. The visitor registration integrated machine based on face recognition according to claim 9, characterized in that: The acquisition module determines the initial face image acquisition frequency of the person to be identified according to the actual environmental impact coefficient; the relationship between the actual environmental impact coefficient and the initial image acquisition frequency is a positive correlation; The calculation formula of the actual environmental impact coefficient P is: ,in, is the current light intensity coefficient, is the weight coefficient of light intensity, is the current dust level, is the weight coefficient of dustiness.

Images