CN107346426A - A kind of face information collection method based on video camera recognition of face - Google Patents

Abstract

A face information collection method based on camera face recognition, the method is as follows: the main controller collects video data and caches it in the main memory, the main controller transmits the segmented video data in the main memory to the image computing chip, the main controller The device receives the face information in the segmented video data analyzed by the image computing chip and the quality parameters of the face image, and the main controller captures the face photos according to the face information and the quality parameters of the face image reach the preset index and stores and uploads them , the present invention undertakes a simple but large amount of face recognition work by adding an image computing chip, and uses a general-purpose processor to acquire and upload face photos according to the results of face recognition, which can replace the original direct upload of the original video code stream for the server to perform The waste of bandwidth caused by face recognition and the waste of hardware resources of the server are suitable for monitoring and tracking more faces at the same time to improve the monitoring effect.

Description

A face information collection method based on camera face recognition

technical field

The invention relates to a method for collecting face information, in particular to a method for collecting face information based on camera face recognition.

Background technique

In today's society, due to security considerations, video surveillance equipment is more and more applied to the monitoring of safe cities, as well as the monitoring of various places such as communities and factories. In most cases, the monitoring process is aimed at the identification of people, and with the maturity of face recognition technology, it is possible to automatically digest the faces in the monitoring video data by artificial intelligence. As monitoring equipment becomes more and more popular, and face recognition has higher requirements for video quality, the corresponding monitoring data is also increasing.

At present, when using face recognition in surveillance systems, in most cases, the camera sends video data to the background server, and the background server relies on its powerful hardware and complex software for intelligent analysis. The data throughput of such a monitoring network is astonishing; under the condition of more and more cameras, the storage capacity and computing power of the background server are also put to the test, which greatly increases the cost; in addition, the monitoring environment is complex and diverse, and the parameter settings of the cameras cannot Ensuring real-time integrity will lead to failure to recognize and collect portraits. Using remote servers to analyze images to judge image quality and return camera parameter modification requirements, and camera adjustments to take pictures will cause great pressure on hardware costs and time. too late.

In recent years, the improvement of algorithms and the improvement of hardware performance have made it possible for cameras to recognize faces. Traditionally, general-purpose processors using common mobile architectures such as cortex or X86-based atom are not suitable for simultaneous video processing and hardware management. Perform large-scale simple calculations required for artificial intelligence.

Contents of the invention

In order to overcome the large network overhead and severely uneven distribution of hardware overhead in the prior art monitoring system, the purpose of the present invention is to provide a face information collection method based on camera face recognition.

The method provided by the invention is as follows:

A face information collection method based on camera face recognition, the method is: in the process of collecting video data by the camera, face information recognition is performed on each frame of video data at the same time to obtain the face images and corresponding faces of all people in the current frame Finally, determine the person whose face information and quality parameters of the face image of all people in the current frame reach the preset standard, and capture a photo of the person's face.

The face information recognition of each frame of video data to obtain the face images of all people in the current frame and the quality parameters of the corresponding face images is specifically: the camera according to the brightness distribution in one frame of the video data Pick up the face information, and then determine the quality parameters of the face image according to the brightness and clarity of the face image in the face information and the matching degree of face features under the front face.

After the camera picks up the face information according to the brightness distribution in one frame of the video data, it also includes: the camera judges whether the current face image is a new person according to the comparison between the current face information and the face information picked up in several adjacent frames, If it is determined to be a new person, assign a face ID to the current face image; otherwise, establish a corresponding relationship between the current face image and an existing face ID.

The determination of the person whose face information and quality parameters of the face image of all the characters in the current frame reach the preset standard, and taking a photo of the face of the person, specifically: when it is determined to be a new person and the quality of the face image When the parameter is greater than or equal to the preset threshold, the camera captures and stores a face photo.

After performing face information recognition on each frame of video data to obtain the face images of all people in the current frame and the quality parameters of the corresponding face images, the method also includes: When the threshold is set, the camera adjusts the camera parameters according to the face information, discards the face image and continues to judge the quality parameter of the face image generated by subsequent video data until the quality parameter of the face image is greater than or equal to the preset threshold.

Determine the person whose face information and quality parameters of the face image of all the characters in the current frame reach the preset standard, and capture a photo of the face of the person, specifically: when it is determined to be the original face and the quality of the face image When the parameter is greater than or equal to the preset threshold, the camera judges that the quality parameter of the face image or the frontal degree of the face is greater than the original face photo, and finally captures the face photo and replaces the original face photo.

After performing face information recognition on each frame of video data to obtain the face images of all people in the current frame and the quality parameters of the corresponding face images, the method also includes: When it is less than the preset threshold, the camera adjusts the camera parameters according to the face information, discards the face image and continues to judge the quality parameter of the face image generated by subsequent video data until the quality parameter of the face image is greater than or equal to the preset threshold.

After the quality parameter of the face image is greater than or equal to the preset threshold, the method further includes: the camera records the exposure parameter when the quality parameter of the face image is greater than or equal to the threshold, and then judges that the camera does not detect For face information, the camera is adjusted according to the exposure parameters recorded when the last face image parameter is greater than or equal to the preset threshold.

Determine the person whose face information and quality parameters of the face images of all the characters in the current frame reach the preset standard, and capture the face photo of the person at the same time, the method also includes: the camera detects that the storage capacity is less than the warning value , and then delete some face photos, specifically: the camera determines the priority of the face ID according to the access status, upload status, and tracking status of the face photo corresponding to the face ID and makes real-time adjustments, and deletes the face photos according to the priority. The face photo corresponding to the face ID and log out the face ID, and finally detect that the storage capacity is higher than the preset threshold and stop the deletion work.

According to the face information and the quality parameters of the face image reaching the preset index, the face photo is captured and uploaded. Specifically, the camera selects the face photo capture method according to the requirements of the remote server.

A method for collecting face information based on camera face recognition according to claim 1, characterized in that: according to the face information and the quality parameters of the face image reaching the preset index, the photo of the face is captured and uploaded, specifically, The camera selects the upload strategy of face photos according to the requirements of the remote server.

Compared with the prior art, the present invention undertakes simple but massive face recognition work by adding an image computing chip, and uses a general-purpose processor to acquire and upload face photos according to the results of face recognition, instead of directly uploading the original video The bandwidth waste caused by the code stream for the server to perform face recognition, as well as the waste of hardware resources of the server, are suitable for monitoring and tracking more faces at the same time to improve the monitoring effect.

Description of drawings

FIG. 1 is a flow chart of a face information collection method based on camera face recognition.

FIG. 2 is a flow chart of a specific embodiment of a face information collection method based on camera face recognition.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

An embodiment of the present invention provides a camera-based face recognition method, the steps include: the camera collects video data, and performs face information recognition on each frame of video data while collecting data to obtain face information of a face image and the quality parameters of the face image, determine that the face image reaches the preset index, and take a photo of the face whose face reaches the preset index.

In order to implement this method specifically, the camera includes a main controller for controlling camera operation, image encoding and decoding, and communication, as well as a corresponding main memory, and also includes a face recognition algorithm for face image recognition. The image computing chip and its corresponding memory and other external components, the camera can also be equipped with an external memory to store the captured photos and reduce the storage pressure of the main memory. On this basis, the specific steps of the method include:

Step 100: the main controller collects the video data of the sensor and caches it in the main memory.

Step 200: The main controller transmits the segmented video data to the image computing chip.

Step 300: The image computing chip analyzes the face in the segmented video data, recognizes the face and returns the face information to the main controller.

Step 301: The image computing chip picks up face information according to the brightness distribution of the face image in a certain frame of the received segmented video data.

Step 302: The image computing chip compares the currently detected face information with the face information of the face images detected in several frames forward, judges the similarity between the two, and judges according to the similarity and the set threshold Whether the currently detected face is a new person, according to the judgment result, the face information of the face image that has appeared is corresponding to the corresponding face ID, or a new face ID is set for the new person.

Step 303: The image computing chip provides the quality parameters of the current face image according to the brightness and clarity of the current face image and the degree of matching with the features of the face under the front face.

Step 304: Transmitting the face image and quality parameter information to the main controller.

Step 400: The main controller adjusts the environmental peripherals according to the face information.

Step 401: The main controller judges whether the exposure parameter needs to be adjusted and the required adjustment step according to the quality parameter of the face image and the brightness of the current face.

Step 402: When the quality parameter of the face image reaches the set threshold, the main controller records the exposure parameters within a period of time.

Step 403: When no human face is detected, the main controller uses the recorded exposure parameters to set the camera.

Step 500: The main controller captures the face and stores it in the main memory or external memory.

Step 501: According to the face ID, when the quality parameter of the received face image of a new person is higher than the preset threshold, the main controller controls the camera to capture and store the photo of the face.

Specifically, the main controller controls the camera to capture a face image, a human body image or an original image, or a free combination of the three, according to settings.

Step 502: According to the face ID, when the main controller judges that the current face image is not a frontal face when receiving the face image information of the saved face photo, the current face image and the saved face photo The face area is relatively larger, and the quality parameter of the current face image is higher than the threshold, or the quality parameter of the current face image is higher than the saved face photo, then capture the face photo and store it to replace the original face photo and record the face The quality parameter of the photo.

Step 503: According to the face ID, when receiving the face image information of the saved face photo, the main controller judges that when the saved face photo is a positive face, the quality parameter of the current face image is greater than the saved face Then capture the photo of the face, store and replace the original photo of the face, and record the quality parameters of the photo of the face.

Step 600: Upload and clean up face photos.

Step 601: When the quality of the face photo exceeds the set value, the main controller uploads the face photo.

Specifically, the main controller preferentially uploads photos whose photo quality parameters meet the uploading standard and whose time interval is shorter than the current face photo, and the face photo whose quality parameter is higher among all the face photos.

Specifically, according to the settings, the main controller controls the camera to perform three upload methods: real-time upload, upload after people leave, and upload at intervals. When the person leaves the detection area, upload the best photo, and the interval upload means that when the face image corresponding to the face ID appears in the detection area, the best photo is uploaded regularly.

Step 602: When the usage capacity of the main memory or the external storage reaches the set value, the main controller sets the ID for each face ID according to the current access situation of the face photo, the current upload situation, the current tracking situation and the last tracking time according to the face ID. priority.

Specifically, the face photo currently being accessed has the highest priority, the face photo that has not been uploaded currently has a higher priority, the face photo that is currently being tracked has a higher priority, and the last tracking time and the current time interval Longer face photos get lower priority.

Step 603: According to the priority of the face ID, priority is given to deleting and covering face photos with lower priority, and canceling the face ID corresponding to the covered face photo.

Embodiment one:

The camera records the video of the detection area and stores it in the memory, and the ARM processor transmits the segmented video data to the FPGA computing chip through the BT1120 standard video data interface.

The FPGA computing chip caches the segmented video data in the memory of the FPGA computing chip and analyzes the segmented data frame by frame, picks up the brightness distribution of the face area in the image, captures the face image and records the face information. The FPGA computing chip uses the currently recorded face information to compare with the face information recorded in the last few frames to determine whether the face is consistent with the previously recorded face. If the face is judged to be a new face, a new face is assigned. Face ID, if the face is judged to be the original face, the currently recorded face information is associated with the face ID corresponding to the original face.

The FPGA computing chip gives the quality parameters of the current face image according to the brightness and clarity of the current face information and the matching degree of the face features under the front face, and transmits the quality parameters and face information to the ARM processor.

The ARM processor judges whether exposure parameters and other camera parameters need to be adjusted according to the face brightness and quality parameters of the current face information, and confirms the adjusted step size, controls the camera to complete the modification of the shooting parameters, and monitors the follow-up person transmitted by the FPGA computing chip Face information in the face image until the brightness is sufficient and the quality parameters reach the standard.

On the premise that the face image quality parameter reaches the preset threshold, the ARM processor captures the face photo and stores it in the memory.

Specifically, the ARM processor extracts the face image in the cache as a face photo, adds description information to the file, and records the shooting time and face ID.

The ARM processor uploads the face photos stored in the memory to the remote server.

The advantages of this solution are: using face ID to judge the acquisition of face images, it is convenient to accurately control and record the collection of face information, and there is no need to acquire and upload a large number of pictures, which greatly saves hardware costs.

According to the face information and quality parameters of the face photo, the photo shooting situation is obtained, and the camera parameters are adjusted in real time, which is conducive to taking photos with higher quality.

Embodiment two:

On the basis of Embodiment 1, when the FPGA computing chip cannot detect any human face, the ARM processor traces back to when the most recent standard human face photo was taken, and the adjusted exposure parameters are used to set the camera.

This solution is convenient to set the exposure state of the camera to the most reasonable state when there is no face in the detection area, and it is convenient to adjust to the appropriate exposure position more quickly after the face is detected again.

Embodiment three:

On the basis of Embodiment 1, the ARM controller judges the face ID after receiving the face image information. When the face ID has been captured and stored, the quality parameters of the corresponding stored face photo and the front face area and the quality parameters and face area contained in the current face image information, if the quality parameter is higher or the frontal face area is larger, the face photo is captured again and the original face photo is replaced.

If the stored face photo is a frontal face, only the quality parameters of the face photo are compared with the quality parameters of the current face image. If the quality parameter of the current face image is higher than the quality parameter of the stored face photo, a new face photo is captured and stored to replace the old face photo.

On average, photos with a larger frontal area have correspondingly higher quality parameters of the face image, so this solution is suitable for filtering out images and photos that are more recognizable to people. The scheme is convenient for fast comparison and obtaining a higher-quality face photo, and is convenient for saving a better-quality face photo.

Embodiment four:

Based on the third embodiment, the ARM processor also receives instructions from the remote server, and adjusts the grasping rules according to the requirements of the remote server.

Specifically, including but not limited to the following image styles:

The original image of the current scene, the human body image to which the face is tracked or the face image of the tracked face, or any combination of the three;

It also includes but is not limited to the following capture strategies:

When the quality parameter of the face image reaches the threshold, capture and upload it immediately;

When the person whose face belongs to leaves the detection area, upload the optimal face photo corresponding to the face;

When the person to whom the face belongs appears in the detection area, the optimal face picture is regularly transmitted.

This solution makes the camera more suitable for daily monitoring tasks, and retains more reasonable information when there is no video material as an archive.

Embodiment five:

On the basis of Embodiment 1, a TF card or a ROM memory is added as an external storage to connect with the ARM processor, and the ARM processor stores the photo of the face in the TF card or the ROM memory.

This solution optimizes the storage environment of face photos to prevent untimely uploads, resulting in photos that are not uploaded and discarded, or newly captured photos that cannot be stored.

Embodiment six:

On the basis of Embodiment 1 or 5, when the ARM processor detects that the storage space of the internal memory or the external storage is less than the preset threshold, the stored face photos are prioritized.

Specifically, according to the face ID, it is judged whether the face photo corresponding to the face ID is being accessed, and a high priority is set for the face photo being accessed, and the upload status of the face photo corresponding to the face ID is judged as already uploaded. Set a low priority for the face photo, determine the tracking status of the face image corresponding to the face ID, set a high priority for the face ID corresponding to the face image being tracked, and set the sequence from the nearest to the farthest according to the last tracking time High to low priority.

According to the priority information changed in real time, the ARM processor deletes the face photo corresponding to the face ID with a lower priority in the internal memory or external memory, and cancels the face ID. When the storage space of the internal memory or external storage is higher than a preset threshold, the deletion work is stopped.

In the case of extremely bad network conditions, this solution ensures more effective preservation of high-value face photos without hindering the normal progress of monitoring work.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present invention can easily think of changes or Replacement should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (10)

1. A face information collection method based on camera face recognition, characterized in that, the method is: in the process of video data collection by camera, simultaneously carry out face information recognition to each frame of video data to obtain all characters in the current frame The face information of the face image and the quality parameter of the corresponding face image, finally determine the person whose face information and quality parameters of the face image of all the characters in the current frame reach the preset standard, and capture the person face photos. 2. A method for collecting face information based on camera face recognition according to claim 1, characterized in that: said face information recognition is carried out to each frame of video data to obtain the face images of all people in the current frame And the quality parameters of the corresponding face image, specifically: the camera picks up the face information according to the brightness distribution in one frame of the video data, and then according to the brightness, clarity and front face image of the face image in the face information The matching degree of the lower face features determines the quality parameters of the face image. 3. A method for collecting face information based on camera face recognition according to claim 2, characterized in that: after the camera picks up the face information according to the brightness distribution in one frame of the video data, it also includes: The camera judges whether the current face image is a new person according to the comparison between the current face information and the face information picked up in several adjacent frames. If it is determined to be a new person, assign a face ID to the current face image; The existing face ID establishes a corresponding relationship. 4. a kind of face information collection method based on camera face recognition according to claim 3, is characterized in that: the face information and the quality parameter of the face images of all characters in the described current frame are determined to reach a predetermined level. Set a standard person, and capture a face photo of the person, specifically: when it is determined to be a new person and the quality parameter of the face image is greater than or equal to a preset threshold, the camera captures a face photo and stores it. 5. A method for collecting face information based on camera face recognition according to claim 3, characterized in that: carry out face information recognition to each frame of video data to obtain the face images and corresponding images of all characters in the current frame. After the quality parameters of the face image, the method also includes: when it is determined to be a new person but the quality parameter of the face image is less than a preset threshold, the camera adjusts the camera parameters according to the face information, discards the face image and continues to judge The quality parameter of the face image generated from the subsequent video data until the quality parameter of the face image is greater than or equal to a preset threshold. 6. A method for collecting face information based on camera face recognition according to claim 3, characterized in that: it is determined that the face information and quality parameters of the face images of all people in the current frame reach a preset standard person, and take a photo of the face of the person, specifically: when it is determined to be the original face and the quality parameter of the face image is greater than or equal to the preset threshold, the camera determines whether the quality parameter of the face image or the face is correct The face level is greater than the original face photo, and finally the face photo is captured and the original face photo is replaced. 7. A method for collecting face information based on camera face recognition according to claim 3, characterized in that: carry out face information recognition for each frame of video data to obtain the face images and corresponding images of all people in the current frame. After the quality parameter of the face image, the method also includes: when it is determined to be the original face but the quality parameter of the face image is less than a preset threshold, the camera adjusts the camera parameters according to the face information, discards the face image and Continue to judge the quality parameter of the face image generated by the subsequent video data until the quality parameter of the face image is greater than or equal to a preset threshold. 8. A method for collecting face information based on camera face recognition according to claim 4 or 6, characterized in that: after the quality parameter of the face image is greater than or equal to a preset threshold, the method also includes: The above-mentioned camera records the exposure parameters when the quality parameter of the face image is greater than or equal to the threshold, and then judges that the camera has not detected face information within the preset time, based on the last face image parameter recorded when the face image parameter is greater than or equal to the preset threshold Exposure parameters to adjust the camera. 9. A method for collecting face information based on camera face recognition according to claim 3, characterized in that: it is determined that the face information and quality parameters of the face images of all people in the current frame reach a preset standard person, and at the same time of capturing a face photo of the person, the method also includes: the camera detects that the storage capacity is less than the warning value, and then deletes part of the face photo, specifically: the camera stores the face photo according to the face ID Take status, upload status and tracking status to determine the priority of the face ID and make real-time adjustments. At the same time, delete the face photos corresponding to the face ID according to the priority order and log out the face ID. Finally, it is detected that the storage capacity is higher than the preset Threshold stops deletion from working. 10. A method for collecting face information based on camera face recognition according to claim 1, characterized in that: according to the quality parameters of the face information and the face image, the preset index is reached to capture a photo of the face and upload it, specifically For, the camera selects the upload strategy of face photos according to the requirements of the remote server.