KR20220042301A - Image detection method and related devices, devices, storage media, computer programs - Google Patents

Abstract

An embodiment of the present invention discloses an image detection method and related apparatus, apparatus and storage medium, the image detection method comprising: acquiring a first image including a target to be detected; detecting the first image to obtain a detection result of the first image, the detection result including whether or not a target to be detected in the first image is blocked by the preset object; and performing a preset operation matching the detection result.

Description

Image detection method and related devices, devices, storage media, computer programs

[Cross-reference to related applications]

The present invention is filed based on a Chinese patent application with an application number of 202011002322.1, an application date of September 22, 2020, entitled "Image detection method and related apparatus, apparatus, and storage medium", said Chinese patent Claiming the priority of the application, all contents of the above Chinese patent application are incorporated herein by reference.

The present invention relates to the field of image processing technology, and more particularly, to an image detection method and related apparatus, apparatus, storage medium, and computer program.

Currently, image processing, in particular, detection and recognition of a target in an image, is widely applied to each use scene. Taking the face as an example, face detection and recognition in an image is widely applied in fields such as finance, border crossing, government, aerospace, power, factories, education, and medical care. In the prior art, an image or video stream including a face is collected by using a photographing device, the face is automatically detected from the image, and the detected face is subjected to face recognition, and the recognition result is processed accordingly.

An embodiment of the present invention provides at least an image detection method and a related apparatus, apparatus, storage medium, and computer program.

An embodiment of the present invention provides an image detection method, comprising: acquiring a first image including a target to be detected; detecting the first image to obtain a detection result of the first image, the detection result including whether or not a target to be detected in the first image is blocked by the preset object; and performing a preset operation matching the detection result.

Therefore, by detecting the first image including the target to be detected to obtain a detection result that is whether the target to be detected is occluded or not, and then performing a preset operation matching the detection result, the target to be detected is occluded By judging whether or not there is a target, a preset operation matching the detection result is subsequently performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

In some embodiments, the detection result is obtained by detecting the first image using a neural network.

Therefore, it is detected through a pre-trained neural network, so that the detection result is more accurate and the detection speed is faster.

In some embodiments, performing the preset operation matching the detection result includes sending a first alert when the target to be detected is not shielded by the preset object, wherein the first alert is To prompt to use an object to shield the target to be detected.

Therefore, when the target to be detected is not shielded, the first notification is sent, so that the target to be detected is not blocked by the preset object in a timely manner, so that the notification recipient can also take a corresponding action in a timely manner.

In some embodiments, the detection result further includes whether the shielding manner in which the target to be detected is shielded by the preset object is the preset shielding manner; The performing the preset operation matching the detection result includes sending a second notification when the target to be detected is shielded by a preset object and the shielding method does not belong to the preset shielding method, The notification is for prompting to adjust the shielding method of the preset object.

Therefore, when the shielding method is not correct, a second notification is sent to adjust the shielding method of the target to be detected in time.

In some embodiments, the performing the preset operation matching the detection result includes: when the target to be detected is occluded by the preset object, at least a first characteristic of the unoccluded portion of the target to be detected in the first image extracting and characterizing the target to be detected as a recognizable feature; and recognizing a target to be detected by using the feature to be recognized, and obtaining a recognition result.

Therefore, when the target to be detected is shielded by the preset object, it is realized by extracting and recognizing the features of the unshielded portion, recognizing some features of the target to be detected, and since the some features are not shielded, detection It can represent the target to be targeted, and guarantees some degree of recognition accuracy.

In some embodiments, the step of extracting in the first image at least a first feature of an unoccluded portion of the target to be detected to be a recognized feature of the target to be detected comprises: the unoccluded portion of the target to be detected in the first image extracting a first feature of , and obtaining a second feature of the occluded portion of the target to be detected; characterizing the first characteristic and the second characteristic as a recognized characteristic of the target to be detected.

Therefore, in addition to using the characteristics of the unshielded portion of the target to be detected, it is possible to improve the feature diversity of the target to be detected by further matching the characteristics of the shielded portion.

In some embodiments, obtaining the second characteristic of the occluded portion of the target to be detected comprises: extracting a feature of the occluded portion from the first image and characterizing it as the second characteristic; or, obtaining a preset characteristic of the occluded portion and characterizing it as a second characteristic, wherein the preset characteristic includes a characteristic obtained based on at least one reference characteristic, and each reference characteristic includes the occluded portion not present. It is obtained by extracting the area corresponding to the shielded part from the non-existent reference target.

Therefore, with respect to the method of determining the characteristics of the occluded portion, through direct extraction of the characteristics of the occluded portion, the characteristics of the occluded portion may be different to some extent according to the difference of the target to be detected, so that the method is accurate in recognition can improve; It is also possible to obtain a preset feature as a shielded part feature, and since it is not necessary to perform feature extraction on the shielded part in the above method, processing efficiency can be always improved by reducing processing resource consumption.

In some embodiments, the step of recognizing a target to be detected, using the feature to be recognized, and obtaining a recognition result, comprises: when the preset target includes one, a second value between the feature to be recognized and a pre-stored feature of the preset target. obtaining one degree of similarity, and when the first degree of similarity satisfies a first preset condition, determining that the recognition result includes a target to be detected has passed identity authentication; When the preset target includes a plurality, a second degree of similarity between the feature to be recognized and a pre-stored feature of each preset target is respectively obtained, and the identity of the target for which the recognition result is to be detected satisfies the second preset condition. and determining, including determining the identity of the preset target corresponding to the second similarity.

Therefore, by calculating the first similarity between the pre-stored features of a specific preset target or calculating the similarity between the pre-stored features of a plurality of preset targets, according to the actual scene demand, the target to be detected and the specified one It can be implemented to compare a preset target of , or a preset target in any one database.

In some embodiments, the method includes: the first preset condition comprising a first degree of similarity greater than a first similarity threshold; The second preset condition includes at least one of including that the second similarity is greater than a second similarity threshold.

Therefore, the first similarity threshold is set respectively in different scenes, so that the recognition result is more accurate.

In some embodiments, the method includes: a first similarity threshold when the feature to be recognized comprises a second feature of the occluded portion of the target to be detected is a first similarity threshold when the feature to be recognized does not comprise the second feature less than the similarity threshold; The second similarity threshold when the to-be-recognized feature includes the second feature includes at least one of the steps being less than the second similarity threshold when the to-be-recognized feature does not include the second feature.

Therefore, if the feature to be recognized includes the second feature, the second feature may be different from the actual feature of the keypoint of the target occluded portion to be detected, and in this case, appropriately reducing the similarity threshold value reduces the recognition accuracy. can be improved

In some embodiments, prior to obtaining a first degree of similarity between the feature to be recognized and the pre-stored feature of the preset target, the method includes, in response to the account registration request, registering an account for the user; determining a second image that satisfies a preset quality requirement from a second image of at least one frame obtained by photographing the user, and extracting a feature of a preset part of the user from the determined second image; The method further includes the steps of establishing an account association with the characteristics of the preset part, and storing the features of the preset part as a pre-stored feature of the preset target.

Accordingly, the second image that satisfies the quality requirement is determined to extract the features of the preset region so that the extracted features are more accurate.

In some embodiments, when the target to be detected is occluded by the predefined object, prior to the step of extracting in the first image at least a first feature of an unoccluded portion of the target to be detected, the method comprises: determining a first image satisfying a preset quality requirement as a first image from which subsequent features are extracted from the multi-frame first image; pre-processing the first image to extract subsequent features; performing biometric detection on the first image from which subsequent features are extracted, and when the biometric detection result is that the target to be detected is a living body, extracting at least a first feature of an unshielded portion of the target to be detected from the first image and determining to perform the following steps.

Therefore, before proceeding with feature extraction, pre-processing is performed so that the extracted features are more accurate, the target to be detected is recognized only when the target to be detected is a living body, and the safety of recognition is strengthened. can be prevented

In some embodiments, in the multi-frame first image including the target to be detected, determining the first image satisfying the preset quality requirement as the first image for extracting subsequent features comprises: the first image of each frame correspondingly obtaining a quality score of the first image of each frame based on the quality coefficient of including at least one of parameter information reflecting , brightness information of the first image; and determining, based on the quality score, a first image that satisfies the preset quality requirement as a first image for extracting subsequent features, wherein a quality score of the selected first image is higher than a quality score of another first image; includes

Thus, the quality score determines the image that satisfies the needs and extracts features, so that the extracted features can better represent the target to be detected.

In some embodiments, the pre-processing of the first image for extracting subsequent features comprises: when the first image includes a plurality of targets to be detected, a target to be detected that satisfies a preset extraction requirement is selected from the first image. determining a target area of , and removing an image portion other than the target area from the first image; and/or detecting that the inclination angle of the target to be detected in the first image is greater than the preset angle, and rotating the first image so that the inclination angle of the target to be detected is less than or equal to the preset angle.

Therefore, when there are a plurality of targets to be detected in the first image, only a target to be detected that satisfies a preset extraction request is determined, a target to be detected that does not satisfy the request is discarded, and a target to be detected that does not satisfy the request is discarded. This reduced the impact on recognition outcomes; Then, when it is detected that the target to be detected is inclined in the first image, it is placed correctly to reduce the influence caused by the tilt of the target to be detected.

In some embodiments, the preset extraction request includes that an area of a corresponding region of the target to be detected is larger than an area of a corresponding region of another target to be detected, and the other target to be detected includes a target other than the target to be detected.

Therefore, the larger the area of the target to be detected, the more accurate the extracted features, so that the target to be detected with a larger area is selected, so that the result to be detected is more accurate.

In some embodiments, the target to be detected comprises a face and the predefined object comprises a mask.

Therefore, it is determined whether the mask is worn on the face, and the corresponding operation is performed. For example, if the mask is not worn on the face or the wearing method of the mask is incorrect, a corresponding notification is sent, allow the user to make adjustments in time; If a face mask is worn, it recognizes the face.

An embodiment of the present invention provides an image detection apparatus, comprising: an image acquisition module configured to acquire a first image including a target to be detected; a target detection module that detects the first image and obtains a detection result of the first image, the detection result including whether or not a target to be detected in the first image is blocked by a preset object; and an operation performing module that performs a preset operation matching the detection result.

An embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the processor executes a program command stored in the memory to implement the image detection method.

An embodiment of the present invention provides a computer-readable storage medium, in which program instructions are stored, and when the program instructions are executed by a processor, the image detection method is implemented.

An embodiment of the present invention provides a computer program, including a computer readable code, wherein when the computer readable code is operated in an electronic device, a processor in the electronic device includes an image detection method for realizing the contents carry out

An embodiment of the present invention provides an image detection method and related apparatus, apparatus, storage medium, and computer program, detecting a first image including a target to be detected to obtain whether the target to be detected is shielded, and then detecting By performing a preset operation matching the result, it is determined whether the target to be detected is shielded, and a subsequent preset operation matching the detection result can be performed, and flexibly based on the shielding state of the target to be detected in the image processing has been implemented.

It should be understood that the general description above and the description of the procedure below are merely illustrative and interpretative and do not limit the present invention.

The drawings herein are incorporated in and constitute a part of the specification, and these drawings correspond to embodiments of the present invention, and together with the specification, explain the technical solutions of the present invention.
1 is a flowchart of an embodiment of an image detection method according to an embodiment of the present invention.
2 is a schematic diagram of a first image among an embodiment of an image detection method according to an embodiment of the present invention.
3 is a schematic diagram of a first image subjected to pre-processing in an embodiment of an image detection method according to an embodiment of the present invention.
4 is a structural schematic diagram of an embodiment of an image detection apparatus according to an embodiment of the present invention.
5 is a structural schematic diagram of an embodiment of an electronic device according to an embodiment of the present invention.
6 is a structural schematic diagram of an embodiment of a computer-readable storage medium according to an embodiment of the present invention.

Hereinafter, the technical solutions of the embodiments of the present invention will be described in detail in conjunction with the drawings in the specification.

In the following description, specific procedures, such as, for example, specific system structures, interfaces, and techniques, are set forth for purposes of explanation and not limitation, so that the present invention may be more thoroughly understood.

The term “and/or” in the text is merely for describing the relation of the related object, and means that there are three types of relation. For example, A and/or B means three cases in which only A exists, A and B exist simultaneously, and only B exists. In addition, in the text, the symbol “/” generally means that the related object before and after is “or”. In addition, “many” in the text means two or more than two. In addition, the term “at least one” in the text means any one of several or any combination of at least two of several, for example, including at least one of A, B, and C, and A , can indicate any one or more elements selected from the set consisting of B and C.

The image detection method in the embodiment of the present invention can be applied to a device having an image processing capability. In addition, the device may have an image collection or video collection function, for example, the device may include a member for collecting images or videos, such as a camera. Alternatively, the device may acquire a video stream or image required by the other device through data transmission or data interaction with the other device, or access the video stream or image required from the storage resource of the other device. there is. Here, the other device has an image collection or video collection function, and a communication connection between the devices, for example, the device can perform data transmission or data interaction with another device through Bluetooth, wireless network, etc. , the embodiment of the present invention is not limited here to the communication method between the two, including but not limited to the cases listed above. In one embodiment, the device includes, but is not limited to, a mobile phone, a tablet PC, an interactive screen, and the like.

Referring to FIG. 1, FIG. 1 is a flow diagram of an embodiment of an image detection method according to an embodiment of the present invention. Here, the image detection method is performed by an electronic device, and the method may include the following steps.

Step S11: Acquire a first image including a target to be detected.

Here, the first image including the target to be detected may directly call the original image including the target object collected by the camera of the present device, and may of course be an image obtained from another device, select frame, adjust brightness , may be an image that has been subjected to resolution, etc. The target to be detected may also include the face, the face or limbs of another animal, and the like. Therefore, it does not limit the manner of acquiring the first image including the target to be detected. Here, the other devices refer to devices that can be operated only by using different central processors.

Step S12: Detect the first image, to obtain a detection result of the first image, wherein the detection result includes whether or not a target to be detected in the first image is blocked by a preset object.

Here, the preset object refers to an arbitrary object in which the target to be detected is shielded, for example, a visible object such as a mask, a scarf, glasses or an arm, or paper.

Detecting the first image, it is necessary to detect whether the first image includes the image to be detected in the detection process, and if there is a target to be detected, it is determined whether the target to be detected is shielded by a preset object . The method of determining whether the target to be detected is occluded by the preset object is to first train one occlusion detection model before detecting the first image, and apply the first image to the occlusion detection model (eg, using the occlusion detection function). It may be a neural network provided with), it may be known whether a target to be detected in the first image is blocked by a preset object. Of course, in some embodiments, the manner of detecting whether or not the target to be detected is shielded by the preset object is also determined whether the preset detection position is shielded by the target to be detected, and whether the object blocking the preset detection position is detected Whether or not the condition of the set object is satisfied may be detected. Here, it is possible to obtain a detection result including whether a target to be detected is shielded by extracting the features of the shielding object and the preset shielding object, determining the degree of similarity.

Step S13: Perform a preset operation matching the detection result.

Here, the detection result may be that the target to be detected is shielded but not shielded by the preset object, the manner in which the target to be detected is shielded by the preset object but is shielded is not the preset scheme, and the target to be detected is not shielded by the preset object The method of being shielded and shielded by the object is the same as the preset method, or the target to be detected is not completely shielded. In the embodiment of the present invention, the situation of whether or not the target to be detected in the first image is blocked by the preset object should be considered. Of course, in some embodiments, the preset object may be installed as an arbitrary object, and only when it is determined whether the target to be detected is shielded using the occlusion detection model, a corresponding preset operation may be performed. The preset operation may be any operation associated with target detection, such as recognition.

In the above method, the first image including the target to be detected is detected to obtain whether the target to be detected is occluded, and then a preset operation matching the detection result is performed to determine whether the target to be detected is occluded By determining , a preset operation matching the detection result is subsequently performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

In some embodiments, the detection result is obtained by detecting the first image using a neural network. By first training one preset object occlusion model before detecting the first image, the trained preset object occlusion model can detect whether a target to be detected in the first image is occluded by the preset object. . Here, the preset object may be one or a plurality of objects, for example, two or three different objects. It is possible to detect whether or not the target to be detected is blocked by any one preset object at the same time. Here, the target to be detected may be a face, and the preset object may be a mask. Correspondingly, the preset object shielding model is a mask detection model. The mask detection model may detect whether the target to be detected wears a mask, and, of course, in some embodiments, when the target to be detected wears a mask, whether the wearing manner of the mask is correct may also be detected . It is detected through a pre-trained neural network, so that the detection result is more accurate and the detection speed is faster.

if the target to be detected is not obscured by the preset object, sending a first alert, wherein the first alert is for prompting to use the preset object to shield the target to be detected. Here, the first notification includes several notification methods, including a frame selection method of a face frame, and when a target to be detected is detected as not being blocked by a preset object, selecting a frame in the form of a face frame for the face area In this case, the face frame may be displayed in a warning color such as red or yellow, for example, the first notification may be a combination of the face frame and the prompt character, and the prompt character is, for example, It may be “You are not wearing a mask” or “Please wear a mask”, and of course, it may be in the form of a voice notification or a flashing indicator. Of course, these formats may be used by matching a plurality of them, or may be used alone, which is not further defined herein. For example, when the mask detection model detects that the mask is not worn on the face, a first notification may be sent to inform that the mask is worn on the face to cover the mouth and nose of the face. When the target to be detected is not shielded, the first notification is sent, so that the target to be detected is not blocked by the preset object in a timely manner, so that the notification recipient can take a corresponding action in a timely manner.

In some embodiments, the detection result may further include whether the shielding scheme in which the target to be detected is shielded by the preset object is the preset shielding scheme. Here, when training this neural network called a preset object shielding model, a preset shielding method may be tagged to the training sample, where the preset shielding method may be an accurate shielding method, whereby the preset object shielding model is trained to When the trained preset object shielding model detects that the target to be detected is shielded by the preset object, it is determined whether the shielding method of the preset object is the preset shielding method. When the target to be detected is shielded by a preset object and the shielding method does not belong to the preset shielding method, a second notification is sent. The second notification is for prompting to adjust the shielding method of the preset object. For example, if the target to be detected is a face and the preset object is a mask, the preset shielding method is an accurate mask wearing method. If it is detected that the mask is worn on the face, it is determined whether the mask wearing method is the correct mask wearing method, and if not, a second notification is sent to adjust the shielding method of the preset object on the face. Of course, in some embodiments, the preset shielding method may have several shielding methods, for example, an accurate shielding method, a first erroneous shielding method, a second erroneous shielding method, etc., and the target to be detected is If it is detected that the shielding method shielded by the preset object is the first erroneous shielding method, a notification corresponding to the first erroneous shielding method is sent, and the target to be detected is shielded by the preset object. If it is detected that it is the second erroneous shielding method, a notification corresponding to the second erroneous shielding method is sent to prompt the target to be detected to adjust the shielding method to the correct shielding method. For example, likewise, if the target to be detected is a face, and the preset object is a mask, the correct masking method is that the mask covers the nose and mouth at the same time, and the first erroneous masking method is that the mask covers the nose but not the mouth, The notification corresponding to the first erroneous masking method is that the face prompts to cover the mouth at the same time, the second false masking method is that the mask covers the mouth but not the nose, and the notification corresponding to the second false masking method is Your face will prompt you to cover your nose at the same time. If the shielding method is not correct, a second notification is sent to adjust the shielding method of the target to be detected in time. Here, the method of the second notification is similar to the method of the first notification, and may be a form of matching a face frame and text notification, matching a face frame and a voice notification, or a single text notification or an independent voice notification or flashing of an indicator light Of course, if there are several different preset shielding methods, text notification or voice can be set to correspond in various ways, for example, if the preset shielding method is the first erroneous shielding method, the text notification is the first It corresponds to the erroneous shielding method.

In some embodiments, when a target to be detected is not blocked by a preset object in some work scenes, the target to be detected is recognized. In some other work scenes, if the target to be detected is not shielded by a preset object, the target to be detected is not recognized, for example, it is necessary to wear a mask in a public place such as a high-speed train or an airplane at a certain time or special time. In this case, if it is detected that the mask is not worn on the face, face recognition is not performed on the face, and the face not wearing the mask cannot pass through face recognition to enter the station. Of course, according to the demand of the work scene, the target to be detected may be recognized even if it is detected that the target to be detected is not shielded by the preset object. It is necessary to recognize the target to be detected, to extract the target to be detected, and before extracting the feature, from the multi-frame first image including the target to be detected, the first image that satisfies the preset quality requirement in the subsequent feature It may be determined by the extracted first image. Here, the method of determining the first image that satisfies the preset quality requirement as the first image for extracting subsequent features is based on the quality factor of the first image of each frame. A quality score may be obtained, wherein the quality factor of the first image is selected from among pose information about the photographing member that the target to be detected is to be detected, parameter information reflecting the size of the target to be detected in the first image, and brightness information of the first image. contains at least one. Here, the target to be detected pose information about the photographing member may be angle information of the to-be-detected target for the photographing member. Among them, the angle information of the target to be detected with respect to the photographing member may be angle information of the target to be detected with respect to the lens during the photographing period. For example, with the lens as the origin, a three-dimensional coordinate system is constructed, where the connecting line between the lens and the core is the X-axis, the front extension of the lens and perpendicular to the X-axis is the Y-axis, and the X-axis and The Z axis is perpendicular to the Y axis. The three-dimensional coordinate system is only for indicating the angles of the target to be detected and the photographing member, and in some embodiments, the selection of the origin of the three-dimensional coordinate system or the selection of three directions may be different from the embodiments of the present invention. Here, the angle may be divided into an angle corresponding to the lens XYZ direction. For example, the target to be detected faces the lens and the angle along the XYZ direction is 0° (degrees), and the target to be detected is the first image collecting member , the target to be detected has an angle in the X direction corresponding to the first image collection member is 90°, the angle along the Y direction is 0°, and the angle along the Z direction is also 0°, where the target to be detected is Since the X-axis is rotated by 90°, the target to be detected and the X-axis also have an angle corresponding to the X-direction of 90°. Of course, the smaller the angle in each direction, the better. The parameter information reflecting the size of the target to be detected in the first image includes an area size occupied by the target to be detected in the first image, wherein the area size is indicated by the area size occupied by the target to be detected in the first image. Of course, the premise is that the target to be detected must be completely contained in the first image, and if the first image contains only a part containing the target to be detected, then this quality is the size of the target to be detected in the first image of this frame. The score of the coefficient is relatively low. The higher the brightness information of the first image, the better, but the closer it is to the brightness of natural light at the current time, the better, where the score of this quality factor is relatively high. Here, the weight occupied by the three quality factors is set according to the relationship between the degree of influence of the three quality factors on the image quality. For example, set the weight of the angle to 0.4, and set the other two parts to 0.3 each, of course, as this is step-listed, the weight between each quality factor can be set by itself according to the demand, some In an embodiment, in addition to these three quality factors, a factor such as ambiguity of the first image may be included, and all factors affecting the image quality may be used to calculate a quality score of the image. Extract features by selecting images whose quality score satisfies the requirements, and allow the extracted features to better represent the target to be detected. Of course, the setting of the weight may take into consideration the actual demand for image detection precision, the processing capability of the image detection device, the resource occupancy situation, and the like. For example, in some embodiments, if the processing power of the image detection device is relatively high and the resource occupancy is relatively small, the quality score may be calculated in consideration of a plurality of quality factors, and the processing power of the image detection device is too low. , the quality score can be calculated by considering only a few quality factors. For example, the quality factor is selected by calculating the time or memory space required for each quality factor. Therefore, it is possible to flexibly choose how many quality factors or how many quality factors to use. Of course, in some embodiments, it is also possible to determine one relatively low quality score threshold, and if the quality score of the first image is lower than the quality score threshold, it is excluded, and the second image whose quality score is greater than the quality score threshold is excluded. 1 Hold the image.

In some embodiments, a subsequent feature extracted first image may be pre-processed prior to feature extraction and recognition of the first image. Here, the pre-processing method is that, when the first image includes a plurality of targets to be detected, a target to be detected that satisfies a preset extraction request determines a target area in the first image, and the target area other than the target area in the first image is determined. You can remove the image part of Among them, the target area may be an area including one target to be detected. In other words, when the first image includes a plurality of targets to be detected, it does not recognize the complete first image, but only recognizes the target area of the target to be detected that satisfies the preset extraction requirement, and, thus, By reducing to some extent the noise generated in the target recognition process, the target to be detected that does not satisfy the requirement reduces the influence on the recognition result. Here, the preset extraction request may be that an area of a region corresponding to a target to be detected is larger than an area of a region corresponding to another target to be detected, wherein the other target to be detected includes a target other than the target to be detected. If there are a plurality of targets to be detected in the first image, the areas occupied by the plurality of targets to be detected may not match, and a target to be detected having a larger area has a relatively higher resolution in the recognition process, thus, A target with a larger area to be detected is selected and recognized. Here, when the areas of a plurality of targets to be detected are the same, a target to be detected having a center of the target to be detected closer to the center of the first image may be recognized, or, in some other embodiments, corresponding to all targets to be detected may be recognized. Acquire a target area to detect a target, and of course all the targets to be detected of the latter point to a target to be detected whose area size is parallel to the first or whose area is greater than one preset area of extraction threshold.

In some embodiments, prior to feature extraction and recognition of the first image, the pre-processing of the subsequent feature extraction first image includes: detecting that the inclination angle of the target to be detected in the first image is greater than the preset angle; 1 The image may be rotated so that the inclination angle of the target to be detected is less than or equal to a preset angle. In some embodiments, the manner of rotation may also rotate a target to be detected, or a target area including the target to be detected, in addition to rotating the entire first image, so the manner of correctly placing the target to be detected is not limited here. does not Here, the preset angle may be within a range of 0° to 180° in a clockwise or counterclockwise direction, an embodiment of the present invention selects to set the preset angle to 0°, and in some embodiments, the preset angle is 30° , 35°, or the like. Here, the method of determining whether the target to be detected is inclined more than the preset angle is to obtain an included angle between a connecting line and a vertical line of a preset first key point and a preset second key point among the target to be detected, and the included angle is greater than the preset angle It is determined whether it is large, if it is large, the first image is rotated so that the included angle is less than or equal to the preset angle, and the first preset keypoint after rotation is positioned on the preset second keypoint, where the upper It may be determined to correspond to the base of the first image. Of course, the inclination angle may be an inclination angle of a position at which the target to be detected corresponds to the first image, for example, the inclination angle at which the target to be detected corresponds to the center of the first image. Of course, here, the preset angle can be set according to the demands of different scenes, for example, according to the area occupied by the first image in the area where the target to be detected is located. For example, when the area of the region where the target to be detected is located is larger than the first area preset value, the preset angle may be set to be greater than 30°, and the area of the region where the target to be detected is positioned is the second area preset value If smaller, the preset angle may be set to be smaller than 30°. Therefore, it is explained that the larger the area of the region where the target to be detected is located, the larger the area of the target to be detected is, that is, the target to be detected is less affected by the angle, and the larger the inclination angle of the target to be detected is, the larger the inclination angle of the target to be detected is, the more the target to be detected is inclined. It explains that it is highly affected by the angle, and it is more strict with respect to the inclination angle of the target to be detected. Of course, these are merely listed, and in some embodiments, a correspondence relationship between another area and a preset angle may be set, and the present invention is not defined herein.

For example, referring to FIGS. 2 and 3 , FIG. 2 is a schematic diagram of a first image among an embodiment of an image detection method according to an embodiment of the present invention, and FIG. 3 is an image detection method according to an embodiment of the present invention. It is a schematic diagram of a first image that has undergone pre-processing in an embodiment. As shown in FIG. 2 , in the first image 20 , the lower half of the target 21 to be detected is shielded by the preset object 22 , and the target 21 to be detected is clearly inclined to the left, that is, the detection The angle between the vertical line and the connecting line between the upper left end point (first preset key point) and the lower left end point (second preset key point) of the target 21 to be detected is 30°, that is, the inclination angle of the target 21 to be detected is 30 If the preset angle as ° is greater than 0 °, the first image 20 is rotated to the right, that is, clockwise by 30 °, and the first image after rotation is as shown in FIG. 3 , which is detected in FIG. The included angle of the vertical line and the connecting line between the upper left end point (first preset key point) and the lower left end point (second preset key point) of the target 21 to be become 0° is equal to the preset angle 0°.

When the inclination angle of the target to be detected in the first image is greater than the preset angle, the target to be detected is placed right away, and the target to be detected is tilted in the process of biometric detection or target recognition for the target to be detected subsequently. The effect was reduced.

In some embodiments, before feature extraction and recognition of the first image, the subsequent feature extraction of the first image may be biometrically detected. It is determined to extract a first characteristic of the unoccluded portion of the target to be detected and perform subsequent steps. Here, when a plurality of targets to be detected exist in the first image, a target to be detected having the largest area is selected for biometric detection. Here, a target area corresponding to a target to be detected may be input to a biometric detection model to be biodetected, wherein the biometric detection model is obtained by training an image including a target to be detected that is shielded by a predetermined predetermined object. By recognizing the target to be detected only when the target to be detected is a living body, the safety of recognition is enhanced, and attacks of virtual objects can be prevented to some extent.

Here, in the process of recognizing the target to be detected, at least a first feature of the unshielded portion of the target to be detected is extracted from the first image to be a recognized feature of the target to be detected. Here, the first characteristic refers to the characteristic of an unshielded keypoint in the target to be detected. Here, it is possible to extract a first feature of an unshielded portion of the target to be detected from the first image, obtain a second feature of the shielded portion of the target to be detected, and obtain the first feature and the second feature from the target to be detected Characteristics to be recognized of Here the second characteristic is the characteristic of the target occluded partial keypoint to be detected. Here, there may be two methods for acquiring the second feature of the shielded portion, one of which is to extract the feature of the shielded portion from the first image and use it as the second feature. In other words, although this part is shielded by the preset object, the second feature of the shielded part is extracted according to the extraction method of the unshielded first feature, that is, whether the target to be detected is shielded by the preset object all use the same processing mechanism, that is, whether or not they are shielded by a preset object does not affect the feature extraction process. Of course, in some embodiments, keypoint features can be extracted from the target to be detected using this method, unless an object that shields the target to be detected is preset. For example, the target to be detected is a face, the preset object is a mask, and the extraction method of the second feature is that the face is not covered by the mask, extracting features of each key point of the face, that is, using the mask Wearing the mask does not affect the feature extraction process, as we use the same processing mechanism for the worn face and the non-masked face. Another method is to obtain a preset feature of the shielded portion and use it as a second feature, wherein the preset feature may be a feature obtained based on at least one reference feature, and each reference feature is a feature of the shielded portion. It is obtained by extracting the area corresponding to the shielded part from the non-existent reference target. In other words, before recognizing the first image, a reference feature is preset for the keypoint of the occluded part, that is, the feature of the occluded part is supplemented. For example, features corresponding to a preset part are extracted in advance from the target to be detected in which the detection result does not have an occluded part, and then the average value of the extracted features is supplemented as a reference feature of the occluded part in the object. Here, features corresponding to a preset part are extracted in advance from the target to be detected whose detection result is not shielded by the preset object, and then the average value of the extracted features is supplemented to refer to the shielded part of the preset object characterized. For example, features corresponding to a preset part are extracted in advance from a face not wearing a mask, that is, features of parts wearing a mask, such as, for example, nose and mouth, and supplementing the average value of the extracted features to provide a mask It is characterized as a preset reference feature of the portion obscured by As for the method of determining the characteristics of the occluded portion, through direct extraction of the characteristics of the occluded portion, the characteristics of the occluded portion are different to some extent according to the difference of the target to be detected, so this method can improve the accuracy of recognition can; It is also possible to obtain a preset feature as a shielded part feature, and in this way, since it is not necessary to perform feature extraction for the shielded part, processing efficiency can be always improved by reducing processing resource consumption.

In some embodiments, after obtaining the to-be-recognized feature of the to-be-detected target, the to-be-recognized feature is used to recognize the to-be-detected target. Here, the recognized scene can be divided into a 1:1 scene and a 1:N scene, where 1:1 is a ratio between two features, and 1:N is a ratio between one feature and a plurality of features. points to In a scene of 1:1, that is, when the preset target includes one, a first degree of similarity between the feature to be recognized and the pre-stored feature of the preset target is obtained, and the first degree of similarity satisfies the first preset condition , it is determined that the recognition result includes that the target to be detected has passed identity authentication. Here, the first preset condition may be that the first similarity is greater than the first similarity threshold. Here, the first similarity threshold when the feature to be recognized includes the second feature of the occluded portion of the target to be detected is smaller than the first similarity threshold when the feature to be recognized does not include the second feature. If the feature to be recognized includes the second feature, the second feature may be different from the true feature of the keypoint of the part where the target to be detected is occluded, and therefore, in this case, if the similarity threshold is appropriately reduced, the accuracy of recognition can improve Here, when the feature to be recognized does not include the second feature, the selection of the first similarity threshold may be determined according to a ratio of the number of the masked keypoints occupying all the target keypoints to be detected. For example, if the keypoints of the occluded portion are one-third of the total number of keypoints to be detected, it may be determined that the first similarity threshold is one-third of the similarity threshold for recognizing the unshielded target to be detected. . In this case, when it is set that the feature to be recognized includes the second feature, the first similarity threshold is 0.1 less than the first similarity threshold when the feature to be recognized does not include the second feature, or is smaller by another value; , not specified here. For example, the similarity threshold for recognizing an unoccluded target to be detected may be between 0.6 and 1. Of course, this is by way of example only, and in some embodiments, the first similarity threshold when the feature to be recognized comprises a second feature of the occluded portion of the target to be detected is that the feature to be recognized includes the second feature. It may be the same as the first similarity threshold value in the case where it does not occur, and when the second characteristic is included in the characteristic to be recognized, it may be determined according to the actual situation in the above method.

In some embodiments, prior to obtaining a first degree of similarity between the feature to be recognized and the pre-stored feature of the preset target, an association between the user account and the pre-stored feature of the preset target is first established. An embodiment is to register an account for the user in response to the account registration request. Among them, the account may be an account with some electronic payment, and if it is an application that can perform target recognition, it can respond to an account registration request and register an account for the user. A user can register with a mobile phone number in a corresponding application, and after successful registration, the user can obtain information such as an ID and password. A second image satisfying a preset quality requirement is determined from the second image of at least one frame obtained by photographing the user, and features of the user's preset part are extracted from the determined second image. Among them, the preset region is the same as the preset region of the target to be detected. Here, the step of selecting the second image satisfying the preset quality requirement is the same as the step of selecting the first image satisfying the preset quality requirement. Finally, the association between the characteristics of the preset site and the account is established, and the characteristics of the preset site are stored as a pre-stored feature of the preset target. That is, the user's preset portion is a preset target.

Here, in a scene of 1:N, that is, when a preset target includes a plurality, a second degree of similarity between a feature to be recognized and a pre-stored feature of each preset target is respectively obtained, and the recognition result is a target to be detected. and determining the identity as the identity of a preset target corresponding to a second degree of similarity that satisfies a second preset condition. Here, the second preset condition may be that the second similarity is greater than the second similarity threshold. Here, in a general case, satisfying the second preset condition indicated herein is a parameter that is not only greater than the second similarity threshold, but also often has a maximum value in all second similarities. That is, a preset target identity corresponding to the maximum second degree of similarity is selected as the identity of the target to be detected. Here, the second similarity threshold when the feature to be recognized includes the second feature is smaller than the second similarity threshold when the feature to be recognized does not include the second feature. If the feature to be recognized includes a second feature, the second feature may be different from the true feature of the keypoint of the target occluded portion to be detected, and thus, appropriately reducing the similarity threshold in this case improves the accuracy of recognition can do. The method of determining the second similarity threshold is the same as the method of determining the first similarity threshold.

In a scene involving several faces, for example, a face recognition automatic ticket gate is installed at the entrance of one office building or one company, and in this scene, within the scope of the building or company, each It is necessary to register and form one face database, and when the registered person appears in front of the automatic ticket gate, the camera of the automatic ticket gate detects and captures the face, compares the captured face with that in the face database, and if the comparison is successful, , when the automatic ticket gate is opened, and an unregistered person appears in front of the automatic ticket gate, the comparison will fail and the automatic ticket gate will not respond.

In the above method, the first image including the target to be detected is detected to obtain whether the target to be detected is occluded, and then a preset operation matching the detection result is performed to determine whether the target to be detected is occluded By determining , a subsequent preset operation matching the detection result can be performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

Based on the above-described embodiment, an embodiment of the present invention further provides an image detection method, which uses a model detection capability based on a deep learning algorithm, and also starts from a facial examination scene, 1:1 and A 1:N scene image detection process was implemented, a mask wearing test method was provided, and an embodiment of face comparison and face search was provided in the mask wearing process.

(1) The facial examination scene mainly includes 1:1 and 1:N two scenes.

For example, 1:1 of the payment scene performs 1:1 verification of the face photo captured in real time and the library photo bound by the member, and authentication passes if it is confirmed that they are the same user. Also, for example, if a 1:N scene is related to a face search, for example, an office building or company has installed a facial recognition automatic ticket gate at the entrance, and in this scene, within the building or company range, a user who needs to enter and exit is registered and one face You need to create a database. When the user appears in front of the automatic ticket gate, the automatic ticket gate camera detects and captures the face, compares the captured facial image with the photo in the facial database, and opens the automatic ticket gate if the comparison succeeds, and if the comparison fails, the automatic ticket gate does not respond In other words, the automatic ticket gate remains closed.

In some embodiments, whether in a 1:1 scene or a 1:N scene, the defense of the living body must proceed. For example, attacks by false persons, such as photographs (including photographs obtained by photographing people, electronic composites, etc.), masks, etc.

(2) The image detection method in a 1:1 scene may be implemented as follows.

First, the user has one account, for example, an account for electronic payment. A user registers with a mobile phone number in the application system corresponding to the electronic payment account, and after successful registration, the user obtains information such as an ID and password, that is, obtains one electronic payment account. Then, the application system guides the user to perform a face binding operation through some activity, performs face recognition when the user goes through biometric authentication, and when the face quality in the video meets the face collection demand, the user's face Collect images. Here, a frame having the highest quality may be selected during the photographing process, and the quality determination criterion includes one or more dimensions such as the angle of the face, the intensity of illumination, and the size of the face.

The collected user's facial image may be associated with the account, specifically, it may be associated with the account's tag, and a comparison picture of the face library may be used. When the user consumes online, after confirming the order form, the payment procedure is entered. At this time, if the user has already bound a facial image and selects facial payment, the face capture procedure is entered. If a face is detected, the mask wear test is performed (that is, facial attribute detection), and if a mask is worn, the subsequent process is continued. can inform

When the user wears a mask, face detection is performed, and at this time during the shooting process, the frame with the highest quality is selected, and facial symmetry (for example, there is a certain inclination angle on the face) Then, the process of placing the face correctly) is performed, and facial feature values are extracted from the symmetrical face photo.

Here, there are two methods for extracting and contrasting facial feature values, the first is extracting facial feature values including a mask part, for example, the facial feature value without a mask in the photo of the facial library is A, and the same method is used to extract the features of the face wearing the mask, take the feature value A1, and compare the feature vector of the feature value A1 with the mask on and the feature value A without the mask again. The second is to implement extraction of facial feature values from the visible part on the mask. If all faces are 128 keypoints and the mask is 64 keypoints, then 64 keypoints extracted feature values from the visible part on the mask. is compared with the feature values extracted from the corresponding 64 keypoints from the photo of the face library. Furthermore, the generated feature value is compared with the feature value of the photo in the face library of the user, and when it deviates from the comparison threshold (for example, the comparison threshold is 0.8, and the similarity is out of 0.8, the same user is considered ), the comparison is passed, and the face examination process is finished.

(3) An image detection method in a 1:N scene may be implemented as follows.

First, it is necessary to input each picture of the user with the right to enter the building (to form a comparison picture of the facial library in the scene through the building) into the office building personnel management work system, wherein the picture in the facial library and the building user are one-to-one form a corresponding relationship of Then, binding association between the photo in the face library and the face recognition device of the automatic ticket gate allows the face recognition device of the automatic ticket gate to read the photo in the face library. When personnel need to pass through the automatic ticket gate, the facial recognition device of the automatic ticket gate detects the facial information, and then enters the face capture state (at this time, the camera module of the facial recognition device is turned on and can track the face and the facial frame moves along with the movement of the face all the way).

Then, when the face is detected, the mask wearing inspection is performed (ie, facial attribute detection), and different operations are performed according to different inspection results, for specific operation details, refer to the description of the content in the 1:1 scene above hope

Here, there are two methods of extracting and searching for facial feature values in a 1:N scene. The first is extracting facial feature values including a mask part, and A, B, C, D and E are included in the comparison photos of the facial library. If there are 5 facial feature values without a mask, feature extraction is performed on the mask-wearing face using the same method to extract 5 feature values A1, B1, C1, D1, and E1. In the case of searching for the value A1, if the distance between the feature value A1 and the feature vector of the feature value A is the shortest, the corresponding feature value A may be searched for. The second is to implement extraction of facial feature values of the visible part on the mask, where the feature value generation part is the same as the method in the 1:1 scene, so the above description can be referred to. After generating the feature value, further, the generated feature value performs a facial feature search in the photos of all users of the entire building (1:N search), and if there is a feature value out of the comparison threshold, search considered success If the search is successful, the face recognition device sends an open signal, and the automatic ticket gate is opened and the face examination process is ended.

In an embodiment of the present invention, through the above method, before recognizing a face, it is possible to check whether the user wears a mask and notify this, and when the user wears a mask, precise face recognition is performed to By completing the inspection process, traffic efficiency was greatly improved, and the risk of taking off the mask was reduced.

Here, the performing subject of the image detection method may be an image detection apparatus, for example, the image detection method may be performed by a terminal or a server or other processing device, where the terminal is a user equipment (UE); It may be a mobile device, a terminal, a cellular phone, a wireless telephone, a personal digital assistant (PDA), a hand-held device, a computing device, an in-vehicle device, a wearable device, and the like. In some possible embodiments, the image detection method may be implemented in a way that invokes computer readable instructions stored in a memory via a processor.

Referring to FIG. 4 , FIG. 4 is a structural schematic diagram of an embodiment of an image detection apparatus according to an embodiment of the present invention. The image detection apparatus 40 includes an image acquisition module 41 , a target detection module 42 , and an operation performing module 43 . The image acquisition module 41 acquires a first image including a target to be detected; the target detection module 42 detects the first image to obtain a detection result of the first image, wherein the detection result includes whether or not a target to be detected in the first image is shielded by the preset object; The operation performing module 43 performs a preset operation matching the detection result.

In the above method, the first image including the target to be detected is detected to obtain whether the target to be detected is occluded, and then a preset operation matching the detection result is performed to determine whether the target to be detected is occluded By determining , a subsequent preset operation matching the detection result can be performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

In some embodiments, the detection result is obtained by the target detection module 42 detecting the first image using a neural network.

In the above method, detection is performed through a pre-trained neural network, so that the detection result is more accurate and the detection speed is faster.

In some embodiments, the operation performing module 43 performs a preset operation matching the detection result, and includes, when the target to be detected is not blocked by the preset object, sending a first notification; Here, the first notification is for prompting to use a preset object to shield the target to be detected.

In the above method, when the target to be detected is not shielded, the first notification is sent, so that the target to be detected is not blocked by the preset object in a timely manner, so that the notification recipient can take the corresponding action in a timely manner .

In some embodiments, the detection result further includes whether the shielding manner in which the target to be detected is shielded by the preset object is the preset shielding manner; When the operation performing module 43 performs a preset operation matching the detection result, when the target to be detected is shielded by a preset object and the shielding method does not belong to the preset shielding method, sending a second notification includes; Here, the second notification is for prompting to adjust the shielding method of the preset object.

In the above method, when the shielding method is not correct, a second notification is sent to adjust the shielding method of the target to be detected in time.

In some embodiments, when the operation performing module 43 performs the preset operation matching the detection result, when the target to be detected is occluded by the preset object, at least the unshielded target of the to be detected in the first image extracting a first feature of the portion to be a recognized feature of the target to be detected; A target to be detected is recognized by using the feature to be recognized, and a recognition result is obtained.

In the above method, when the target to be detected is shielded by a preset object, features of the unshielded portion are extracted and recognized, and some features of the target to be detected are recognized, and some features are not shielded. In this case, it can represent the target to be detected, and guarantee the accuracy of recognition to some extent.

In some embodiments, the operation performing module 43 extracts, in the first image, at least a first feature of the unoccluded portion of the target to be detected, so that the to-be-detected feature of the target to be detected is the to be detected in the first image. extracting a first characteristic of the unoccluded portion of the target, and obtaining a second characteristic of the occluded portion of the target to be detected; characterizing the first characteristic and the second characteristic as a recognized characteristic of the target to be detected.

In the above method, in addition to using the characteristics of the unshielded portion of the target to be detected, it is possible to improve the feature diversity of the target to be detected by further attaching to the characteristics of the shielded portion.

In some embodiments, the acquiring the second characteristic of the occluded portion of the target to be detected by the operation performing module 43 may include extracting a feature of the occluded portion from the first image to be the second characteristic; or, obtaining a preset characteristic of the occluded portion and setting it as a second characteristic, wherein the preset characteristic includes a characteristic obtained based on at least one reference characteristic, and each reference characteristic includes the presence of the occluded portion It is obtained by extracting the area corresponding to the shielded part from the reference target that is not.

In the above method, for the characteristic determination method of the occluded portion, through direct extraction of the characteristic of the occluded portion, the characteristic of the occluded portion is different to some extent according to the difference of the target to be detected, so the method is recognized can improve the accuracy of; It is also possible to obtain a preset feature as a shielded part feature, and since it is not necessary to perform feature extraction on the shielded part in the above method, processing efficiency can be always improved by reducing processing resource consumption.

In some embodiments, the operation performing module 43 recognizing the target to be detected by using the feature to be recognized, and obtaining a recognition result, may include: when the preset target includes one, the to-be-recognized feature and the preset target obtain a first degree of similarity between the pre-stored features, and when the first degree of similarity satisfies a first preset condition, determine that the recognition result includes that the target to be detected has passed identity authentication; When the preset target includes a plurality, a second degree of similarity between the feature to be recognized and a pre-stored feature of each preset target is respectively obtained, and the identity of the target for which the recognition result is to be detected satisfies the second preset condition. and determining the identity of the preset target corresponding to the second degree of similarity.

In the above method, by calculating the first degree of similarity between the pre-stored features of a specific preset target or calculating the similarity between the pre-stored features of a plurality of preset targets, the target to be detected and the specific It can be implemented to compare any one of the preset targets, or to compare with a preset target in any one database.

In some embodiments, the first preset condition includes that the first similarity is greater than the first similarity threshold; The second preset condition includes that the second similarity is greater than the second similarity threshold.

In the above scheme, the first similarity threshold is respectively set in different scenes, so that the recognition result is more accurate.

In some embodiments, the first similarity threshold when the feature to be recognized comprises a second feature of the occluded portion of the target to be detected is a first similarity threshold when the feature to be recognized does not comprise the second feature smaller than; The second similarity threshold when the to-be-recognized feature includes the second feature is smaller than the second similarity threshold when the to-be-recognized feature does not include the second feature.

In the above scheme, if the characteristic to be recognized comprises a second characteristic, the second characteristic may be different from the true characteristic of the keypoint of the target occluded portion to be detected, and thus, in this case, if the similarity threshold is appropriately reduced The recognition accuracy can be improved.

In some embodiments, the image detection device 40 further includes a dictionary storage module (not shown). Before the operation performing module 43 acquires a first degree of similarity between the feature to be recognized and the preset target feature, the preset target module is configured to: in response to the account registration request, register an account for the user; determining, from the second image of at least one frame obtained by the user, a second image satisfying a preset quality requirement, and extracting a feature of the user's preset part from the determined second image; Establish a relationship between the characteristics of the preset site and the account, and store the characteristics of the preset site as a pre-stored feature of the preset target.

In the above method, a second image that satisfies the quality requirement is determined and features of a preset region are extracted so that the extracted features are more accurate.

In some embodiments, when the target to be detected is occluded by the preset object, prior to the step of extracting from the first image at least a first feature of an unoccluded portion of the target to be detected, the operation performing module 43 may also , determining, in the multi-frame first image including the target to be detected, a first image satisfying a preset quality requirement as a first image from which a subsequent feature is extracted; pre-process the first image to extract subsequent features; Biometric detection is performed on the first image from which subsequent features are extracted, and when the biometric detection result is that the target to be detected is a living body, extracting at least a first feature of an unshielded portion of the target to be detected from the first image; and determining to perform the subsequent step.

In the above method, prior to feature extraction, pre-processing is performed so that the extracted features are more accurate, the target to be detected is recognized only when the target to be detected is a living body, the safety of recognition is strengthened, and the attack of the virtual object can be prevented to some extent.

In some embodiments, determining, by the operation performing module 43 , in the multi-frame first image including the target to be detected, a first image satisfying a preset quality requirement as a first image for extracting subsequent features, includes: Based on the quality factor of the first image of each frame, a quality score of the first image of each frame is correspondingly obtained, wherein the quality factor of the first image includes: pose information for a photographing member in which a target to be detected is a first image; at least one of parameter information reflecting the size of a target to be detected in the image and brightness information of the first image; Based on the quality score, a first image satisfying a preset quality requirement is determined as a first image for extracting subsequent features, wherein the quality score of the selected first image is higher than the quality scores of other first images include

In the above scheme, an image whose quality score satisfies the requirement is extracted and features are extracted, so that the extracted features can better represent the target to be detected.

In some embodiments, the pre-processing of the first image for extracting subsequent features by the operation performing module 43 may include: when the first image includes a plurality of to-be-detected targets, to be detected that satisfy a preset extraction requirement the target determines a target area in the first image, and removes an image portion other than the target area in the first image; and/or detecting that the inclination angle of the target to be detected in the first image is greater than the preset angle, and rotating the first image so that the inclination angle of the target to be detected is smaller than the preset angle.

In the above method, when there are a plurality of targets to be detected in the first image, only a target to be detected that satisfies a preset extraction request is determined, a target to be detected that does not satisfy the request is discarded, and a target that does not satisfy the request is discarded. The influence of the target to be detected on the recognition result was reduced. Then, when it is detected that the target to be detected is inclined in the first image, it is placed correctly to reduce the influence caused by the tilt of the target to be detected.

In some embodiments, the preset extraction request includes that an area of a corresponding region of the target to be detected is larger than an area of a corresponding region of another target to be detected, and the other target to be detected includes a target other than the target to be detected.

In the above scheme, the larger the area of the target to be detected, the more accurate the extracted features, so that the target to be detected with a larger area is selected so that the result to be detected is more accurate.

In some embodiments, the target to be detected comprises a face and the predefined object comprises a mask.

In the above method, it is determined whether the mask is worn on the face and a corresponding operation is performed. For example, if the mask is not worn on the face or the wearing method of the mask is not correct, a corresponding notification is issued send, so that the user can make adjustments in time; If a face mask is worn, it recognizes the face.

In the above method, the first image including the target to be detected is detected to obtain whether the target to be detected is occluded, and then a preset operation matching the detection result is performed to determine whether the target to be detected is occluded By determining , a subsequent preset operation matching the detection result can be performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

Referring to FIG. 5, FIG. 5 is a structural schematic diagram of an embodiment of an electronic device according to an embodiment of the present invention. The electronic device 50 includes a memory 51 and a processor 52 , and the processor 52 executes program instructions stored in the memory 51 to implement the steps in any of the above image detection method embodiments. In one embodiment, the electronic device 50 may include, but is not limited to, a microcomputer and a server, and in addition, the electronic device 50 may further include, but is not limited to, a mobile device such as a notebook computer and a tablet PC. does not

Here, the processor 52 controls itself and the memory 51 to implement the steps in any of the above image detection method embodiments. The processor 52 may also be referred to as a central processing unit (CPU). The processor 52 may be an integrated circuit chip with signal processing capabilities. Processor 52 is a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), a dedicated integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic device, It may be a discrete gate or transistor logic device, a discrete hardware component, or the like. In addition, the processor 52 may be jointly implemented through an integrated circuit chip.

In the above method, the first image including the target to be detected is detected to obtain whether the target to be detected is occluded, and then a preset operation matching the detection result is performed to determine whether the target to be detected is occluded By determining , a subsequent preset operation matching the detection result can be performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

Referring to FIG. 6, FIG. 6 is a structural schematic diagram of an embodiment of a computer-readable storage medium according to an embodiment of the present invention. The computer-readable storage medium 60 stores program instructions 61 that are executed by a processor, and the program instructions 61 implement steps in any of the above-described image detection method embodiments.

In the above method, the first image including the target to be detected is detected to obtain whether the target to be detected is occluded, and then a preset operation matching the detection result is performed to determine whether the target to be detected is occluded By determining , a subsequent preset operation matching the detection result can be performed, and flexible processing is implemented based on the shielding state of the target to be detected in the image.

In some embodiments, an embodiment of the present invention provides a computer program, comprising computer readable code, wherein when the computer readable code is executed in an electronic device, a processor in the electronic device is configured to implement the method is executed

In some embodiments, a function provided by an apparatus provided in an embodiment of the present invention or a module included is for performing the method described in the method embodiment, and for the sake of brevity, the implementation refers to the description of the method embodiment hope

The description of each embodiment above emphasizes the differences between the embodiments, and for the sake of brevity, the same or similar points may be cross-referenced.

In the various embodiments provided by the present invention, it should be understood that the disclosed devices and methods may be implemented in other ways. For example, the device embodiments described above are merely exemplary, for example, the division of a module or a unit is only a division of a logic function, and may have other division manners in actual implementation, for example, a unit or an assembly may be combined or integrated into other systems, or some features may be omitted or not implemented. In addition, the mutual coupling or direct coupling or communication connection indicated or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical, or other form.

In addition, in each embodiment of the present invention, each functional unit may be integrated into one processing unit, or each unit may exist physically alone, or two or two or more units may be integrated into one unit. The integrated unit may be implemented in the form of hardware or may be implemented in the form of a software functional unit. If the integrated unit is realized in the form of a software function module and sold or used as a separate product, it may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention may be implemented in the form of a software product essentially or a part contributing to the prior art, and the computer software product is stored in a single storage medium, The instructions may be included to cause one computing device (such as a personal computer, a server, or a network device) or a processor to perform all or part of the method according to each embodiment of the present invention. The above-described storage medium includes various media capable of storing a program code, such as a USB, a removable disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disk, and the like.

An embodiment of the present invention provides an image detection method and related apparatus, apparatus, storage medium, and computer program, the method comprising: acquiring a first image including a target to be detected; detecting the first image to obtain a detection result of the first image, the detection result including whether a target to be detected in the first image is blocked by a preset object; and performing a preset operation matching the detection result. The image detection method provided according to an embodiment of the present invention determines whether a target to be detected is shielded, and subsequently performs a preset operation matching the detection result to determine the shielding state of the target to be detected in the image. based on flexible processing was implemented.

Claims (20)

An image detection method comprising:
acquiring a first image comprising a target to be detected;
detecting the first image to obtain a detection result of the first image, the detection result including whether a target to be detected in the first image is blocked by a preset object; and
and performing a preset operation matching the detection result. According to claim 1,
The detection result is an image detection method obtained by detecting the first image using a neural network. 3. The method of claim 1 or 2,
The step of performing a preset operation matching the detection result includes:
if the target to be detected is not shielded by the preset object, sending a first alert, wherein the first alert is for prompting to use the preset object to shield the to-be-detected target An image detection method. 4. The method according to any one of claims 1 to 3,
the detection result includes whether the shielding scheme in which the target to be detected is shielded by a preset object is a preset shielding scheme; The step of performing a preset operation matching the detection result includes:
and sending a second notification when the target to be detected is shielded by the preset object and the shielding method does not belong to the preset shielding method, wherein the second notification is the shielding method of the preset object An image detection method for prompting to adjust 5. The method according to any one of claims 1 to 4,
The step of performing a preset operation matching the detection result includes:
when the target to be detected is occluded by the preset object, extracting from the first image at least a first feature of an unoccluded portion of the target to be detected to be a recognized feature of the target to be detected; and
and recognizing the target to be detected by using the to-be-recognized feature to obtain a recognition result. 6. The method of claim 5,
extracting a first feature of at least an unoccluded portion of the target to be detected from the first image, and characterizing the to-be-detected target as a recognizable feature,
extracting a first feature of an unoccluded portion of the target to be detected from the first image, and obtaining a second feature of the occluded portion of the target to be detected; and
characterizing the first characteristic and the second characteristic as a recognized characteristic of the target to be detected. 7. The method of claim 6,
obtaining a second characteristic of the occluded portion of the target to be detected comprises:
extracting a feature of the shielded portion from the first image and setting it as the second feature; or,
obtaining a predetermined characteristic of the occluded portion and characterizing it as the second characteristic, wherein the predetermined characteristic includes a characteristic obtained based on at least one reference characteristic, each reference characteristic comprising the occluded portion An image detection method obtained by extracting an area corresponding to the shielded portion from the non-existent reference target. 8. The method according to any one of claims 5 to 7,
Recognizing the target to be detected by using the to-be-recognized feature, and obtaining a recognition result,
When the preset target includes one, a first degree of similarity between the feature to be recognized and a pre-stored feature of the preset target is obtained, and when the first degree of similarity satisfies a first preset condition, the recognition result is determining that the target to be detected includes that it has passed identity authentication; and
When the preset target includes a plurality, a second degree of similarity between the to-be-recognized feature and a pre-stored feature of each preset target is respectively obtained, and the identification result of the recognition result determines the identity of the to-be-detected target as a second group An image detection method comprising at least one of determining, including determining the identity of a preset target corresponding to a second degree of similarity that satisfies a preset condition. 9. The method of claim 8,
The method is
wherein the first preset condition includes that the first similarity is greater than a first similarity threshold; and
and the second preset condition includes at least one of the step of including that the second similarity is greater than a second similarity threshold. 10. The method of claim 9,
The method is
The first similarity threshold when the feature to be recognized includes a second feature of the occluded portion of the target to be detected is the first similarity threshold when the feature to be recognized does not include the second feature less than the value; and
wherein the second similarity threshold when the to-be-recognized feature includes the second feature is smaller than the second similarity threshold when the to-be-recognized feature does not include the second feature. An image detection method comprising. 11. The method according to any one of claims 8 to 10,
Prior to obtaining a first degree of similarity between the feature to be recognized and the pre-stored feature of the preset target, the method comprises:
in response to the account registration request, registering an account for the user;
determining the second image that satisfies a preset quality requirement from a second image of at least one frame obtained by photographing the user, and extracting a feature of a preset part of the user from the determined second image; and
The image detection method further comprising the step of establishing an association between the characteristic of the preset part and the account, and storing the characteristic of the preset part as a pre-stored feature of the preset target. 12. The method according to any one of claims 5 to 11,
When the target to be detected is occluded by the preset object, before extracting from the first image at least a first feature of an unoccluded portion of the target to be detected, the method comprises:
determining, in the multi-frame first image including the target to be detected, the first image satisfying a preset quality requirement as the first image from which a subsequent feature is extracted;
pre-processing the first image to extract subsequent features; and
When biometric detection is performed on the first image from which subsequent features are extracted, and the biometric detection result is that the target to be detected is a living body, in the first image, at least a first portion of the unshielded portion of the target to be detected is performed. The method of detecting an image further comprising at least one of extracting the feature and determining to perform a subsequent step. 13. The method of claim 12,
In the multi-frame first image including the target to be detected, determining the first image satisfying a preset quality requirement as the first image from which a subsequent feature is extracted includes:
correspondingly, based on the quality factor of the first image of each frame, obtaining a quality score of the first image of each frame; , including at least one of parameter information reflecting the size of a target to be detected in the first image and brightness information of the first image; and
determining, based on the quality score, the first image satisfying a preset quality requirement as the first image for extracting subsequent features, wherein the quality score of the selected first image is different from that of another first image Higher than quality score - an image detection method comprising . 14. The method of claim 12 or 13,
The pre-processing of the first image for extracting the subsequent features comprises:
When the first image includes a plurality of the targets to be detected, the target to be detected that satisfies a preset extraction requirement determines a target area in the first image, and the target area other than the target area in the first image removing the image portion; and
detecting that the inclination angle of the to-be-detected target in the first image is greater than a preset angle, and rotating the first image so that the inclination angle of the to-be-detected target is less than or equal to the preset angle; An image detection method comprising at least one of 15. The method of claim 14,
The preset extraction request includes that an area of a corresponding area of the target to be detected is larger than an area of a corresponding area of another target to be detected, and the other target to be detected includes a target other than the target to be detected. . 11. The method of claim 10,
The target to be detected includes a face, and the preset object includes a mask. An image detection device comprising:
an image acquisition module for acquiring a first image including a target to be detected;
a target detection module that detects the first image and obtains a detection result of the first image, wherein the detection result includes whether a target to be detected in the first image is blocked by a preset object; and
and an operation performing module that performs a preset operation matching the detection result. An electronic device comprising a memory and a processor, comprising:
The processor executes a program command stored in the memory to implement the image detection method according to any one of claims 1 to 16. A computer-readable storage medium having program instructions stored thereon, comprising:
17. A computer-readable storage medium embodying the image detection method according to any one of claims 1 to 16 when the program instructions are executed by a processor. A computer program comprising:
17. A computer program comprising computer readable code, wherein when the computer readable code is executed in an electronic device, a processor in the electronic device is executed to implement the image detection method according to any one of claims 1 to 16. .

Images