Banking outlet security video time detection method and related equipment
By constructing 2D-PDN and TDN models based on convolutional neural networks, the problem of insufficient robustness and generalization ability of time recognition in bank branch surveillance videos is solved, realizing automated and low-cost time extraction and recognition, adapting to various time types of videos.
Patent Information
- Application Number
- CN202511733615.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-24
- Publication Date
- 2026-02-10
AI Technical Summary
In existing technologies, time recognition algorithms based on heuristic learning suffer from insufficient robustness and generalization ability in bank branch surveillance videos, making it difficult to adapt to various time types, resulting in high operating costs and limited applicability.
By employing 2D-PDN and TDN models based on convolutional neural networks, an end-to-end recognition system with stronger robustness and generalization ability is constructed by detecting the initial time region of video frame images and identifying the current time, directly extracting time information from the video.
It enables automated extraction of video time from various time types, reducing operating costs and system load, expanding the applicability of the algorithm, and reducing the need for manual parameter adjustments.
Smart Images
Figure CN121505518A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of time detection technology, and more specifically, to a method and related equipment for time detection of security videos in bank branches. Background Technology
[0002] In bank security systems, video surveillance systems play a crucial role as a fundamental infrastructure. These systems not only safeguard the property security of bank premises, but their application efficiency has also significantly improved with advancements in artificial intelligence algorithms and equipment performance. Notably, the time information in surveillance videos is a vital component in recording the time of events. Therefore, automating the extraction of time data from bank branch surveillance videos will provide crucial temporal feature support for the intelligent analysis of security surveillance videos.
[0003] In existing technologies, time extraction from videos is achieved through heuristic learning-based time recognition algorithms. Specifically, this involves locating each digit in the time and then extracting the video time based on the time recognition of individual characters. However, because time digits in surveillance videos are a special type of video text, they suffer from low resolution, diverse colors and sizes, and inconsistent time position distribution, making it difficult to locate individual time digits. Furthermore, there are many types of time (e.g., different colors of time digits, background colors, font sizes, and font colors). In practical applications, heuristic learning-based time recognition algorithms exhibit insufficient robustness and generalization ability, only applicable to single types of video time and unable to adapt to multiple time types. When dealing with different time types of video time, extensive manual parameter adjustments are required, which not only increases operational costs but also limits the applicability of the algorithm. Summary of the Invention
[0004] In view of this, the present invention provides a method and related equipment for detecting the time of security videos in bank branches, with the aim of improving scalability, reducing maintenance difficulty, reducing network occupation and reducing system load.
[0005] The first aspect illustrates a method for time detection in security videos at bank branches, the method comprising:
[0006] Retrieve videos with time information;
[0007] The video is detected using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where a second transition occurs;
[0008] The current time of the specified video frame image is identified from the initial time region of the specified video frame image using a pre-built TDN model; wherein the TDN model is a variant based on a convolutional neural network.
[0009] Optionally, the video is detected using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image, including:
[0010] The video is traversed using a 2D-PDN model based on a convolutional neural network to select multiple stable pixels from all pixels; wherein the grayscale value sequence of the stable pixels is periodically stable.
[0011] The 2D-PDN model based on the convolutional neural network identifies multiple second pixels from all pixels based on each of the stable pixels, and determines a specified video frame image and its initial time region based on each of the second pixels.
[0012] Optionally, the 2D-PDN model based on the convolutional neural network identifies multiple second pixels from all pixels based on each of the stable pixels, and determines a specified video frame image and its initial time region based on each of the second pixels, including:
[0013] The 2D-PDN model based on the convolutional neural network is used to determine the neighboring pixels of each stable pixel from all pixels, and multiple second pixels are determined from each neighboring pixel based on the periodic feature map of each neighboring pixel.
[0014] The 2D-PDN model based on convolutional neural networks determines the second-transition frame image and the time region on the second-transition frame from the video based on each second pixel, and determines the previous video frame image of the second-transition frame as the designated video frame image; wherein, the initial time region of the designated video frame image is at the same position as the time region of the second-transition frame image.
[0015] Optionally, the current time of the specified video frame image is identified from the initial time region of the specified video frame image using a pre-built TDN model, including:
[0016] The initial time region of the specified video frame image is adjusted, and the adjusted initial time region is input into a pre-constructed TDN model so that the TDN model can identify the current time of the specified video frame image from the adjusted initial time region.
[0017] Optionally, the method further includes:
[0018] Synchronize the current time of the specified video frame image with the video time of the specified video frame image.
[0019] The second aspect illustrates a security video time detection system for bank branches, the system comprising:
[0020] The video acquisition module is used to acquire videos with time information.
[0021] The time region detection module is used to detect the video using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where the second transition occurs;
[0022] The time recognition module is used to identify the current time of the specified video frame image from the initial time region of the specified video frame image using a pre-built TDN model; wherein the TDN model is a variant based on a convolutional neural network.
[0023] Optionally, a time region detection module is used to detect the video using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image. Specifically, this module is used for:
[0024] The video is traversed using a 2D-PDN model based on a convolutional neural network to select multiple stable pixels from all pixels; wherein the grayscale value sequence of the stable pixels is periodically stable.
[0025] The 2D-PDN model based on the convolutional neural network identifies multiple second pixels from all pixels based on each of the stable pixels, and determines a specified video frame image and its initial time region based on each of the second pixels.
[0026] Optionally, a time region detection module is used to identify multiple second pixels from all pixels based on each of the stable pixels using the 2D-PDN model based on a convolutional neural network, and to determine the specified video frame image and its initial time region based on each of the second pixels. Specifically, this module is used for:
[0027] The 2D-PDN model based on the convolutional neural network is used to determine the neighboring pixels of each stable pixel from all pixels, and multiple second pixels are determined from each neighboring pixel based on the periodic feature map of each neighboring pixel.
[0028] The 2D-PDN model based on convolutional neural networks determines the second-transition frame image and the time region on the second-transition frame from the video based on each second pixel, and determines the previous video frame image of the second-transition frame as the designated video frame image; wherein, the initial time region of the designated video frame image is at the same position as the time region of the second-transition frame image.
[0029] The third aspect discloses an electronic device comprising: a processor and a memory connected via a bus; wherein the processor is configured to call and execute a program stored in the memory; and the memory is configured to store the program for implementing the bank branch security video time detection method as described in any one of the first aspects.
[0030] The fourth aspect discloses a computer-readable storage medium storing computer-executable instructions for performing the bank branch security video time detection method as described in any one of the first aspects.
[0031] This application provides a method and related equipment for time detection in bank branch security videos. The method involves acquiring videos with time information, using a 2D-PDN model based on a convolutional neural network to detect the video and identify the initial time region of a specified video frame image. The specified video frame image is the next frame after the second-transition frame image in the video. A pre-built TDN model is used to identify the current time of the specified video frame image from its initial time region. The TDN model is a variant of a convolutional neural network. Therefore, the technical solution provided in this application utilizes a 2D-PDN model with stronger robustness and generalization ability to construct a 2D-PDN model based on a convolutional neural network. This 2D-PDN model is then used to detect videos with time information, identifying the initial time region of a specified video frame. This solves the problem in existing technologies where algorithms with low robustness and generalization ability cannot adapt to various video time types. Furthermore, this application uses a pre-built TDN model to directly identify the current time of the initial time region end-to-end, eliminating the need to separately identify time digits at different locations or manually adjust parameters. This not only reduces the difficulty of time identification but also lowers operating costs and expands the scope of application. Attached Figure Description
[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0033] Figure 1 This application provides an example diagram of a video feed from a bank branch surveillance system and its duration.
[0034] Figure 2 An example diagram illustrating the periodic variation pattern of the second pixel in a second digit, provided as an embodiment of this application;
[0035] Figure 3 An example diagram of a second-similar grayscale value difference sequence provided in this application embodiment;
[0036] Figure 4 An example diagram of a second-pixel feature map provided in this application embodiment;
[0037] Figure 5 A flowchart illustrating a time detection method for security videos in bank branches, provided as an embodiment of this application;
[0038] Figure 6 A structural diagram of a 2D-PDN model provided in an embodiment of this application;
[0039] Figure 7 A structural diagram of a TDN model provided in an embodiment of this application;
[0040] Figure 8 A schematic diagram of the structure of a bank branch security video time detection method system provided in this application embodiment;
[0041] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0043] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.
[0044] To better understand this application, the technical terms used in this application are explained below:
[0045] Convolutional Neural Networks (CNNs) are deep learning models primarily used for image recognition and processing. They extract local features from images through convolutional layers, reduce data dimensionality through pooling layers, and perform classification through fully connected layers. CNNs leverage local perception and weight sharing to effectively extract image features and are widely used in image classification, object detection, and other fields.
[0046] 2D-PDN model: full name is two-dimensional periodicity detection network, a convolutional neural network used to detect two-dimensional periodic sequences. It can effectively detect the periodic changes in the second digits of the time region in network surveillance videos.
[0047] Video frame rate: Video frame rate refers to the number of frames displayed per second (FPS). The higher the frame rate, the smoother the picture; the lower the frame rate, the more choppy the picture.
[0048] Video Time: The area in the bank branch's surveillance video used to identify the current time is recorded as the video time, such as... Figure 1 As shown, Figure 1 The box in the image represents the video time range. Figure 1 As can be seen from the video, the current time is 11:16:11 on June 20, 2022.
[0049] Seconds digit / Seconds digit box / Seconds pixel: The numbers in the seconds time box in a video are called seconds digits, the box containing the seconds digits is called the seconds digit box, and the pixels in the seconds digit box are called second pixels. For example... Figure 1 The last "1" in the current time shown corresponds to the seconds digit box. The number in this box changes every second, exhibiting a periodic pattern. The periodic pattern is as follows: Figure 2As shown.
[0050] Video frame rate: The frequency (rate) at which bitmap images appear continuously on a display, measured in frames. In this invention, it is denoted as frame rate. Frame rate can also be called frame frequency, and is expressed in Hertz (Hz).
[0051] Second to frame conversion: Figure 1 The video frame corresponding to the change of the seconds digit in the video is called a second-shift frame. In other words, when the video reaches a second-shift frame, the number in the seconds box changes. Theoretically, a second-shift frame corresponds to only one frame in the video, but due to frame skipping during video storage and playback, a second-shift frame may correspond to several consecutive frames in the video.
[0052] Second-pixel grayscale value difference sequence: The second-pixel grayscale value difference sequence is obtained by performing interpolation processing on the second-pixel grayscale value frame sequence, such as... Figure 3 As shown, from Figure 3 As can be seen, this description of the changes in the unit second digit ignores the differences in the color of the time digits, focusing more on and highlighting the periodic changes in the unit second digits. This allows for a more accurate description of the periodic changes in the grayscale values of second pixels for various colors, resulting in a stronger representation of second pixels. From... Figure 3 It can also be seen that the value corresponding to the non-second conversion frame in the grayscale difference sequence is relatively small. Theoretically, this value should be equal to 0, but in reality, noise exists in the video, causing this value to be relatively small, close to 0.2. The value corresponding to the second conversion frame in the grayscale difference sequence is larger, but this value will vary depending on the position of the second pixel in the unit digit box. This value will also vary depending on the video time type, because the color of the time digits will vary depending on the time type.
[0053] Second-position pixel feature map / two-dimensional difference periodic feature: The two-dimensional matrix obtained by flattening the one-dimensional frame sequence of pixel grayscale values according to the frame rate is the pixel feature map, such as... Figure 4 As shown. From below Figure 4 As can be seen, the larger grayscale changes caused by the frame conversion are concentrated in adjacent columns of the pixel feature map, while the grayscale changes in other columns are relatively small, close to zero. This is the characteristic map after the conversion.
[0054] See Figure 5 The diagram illustrates a flowchart of a bank branch security video time detection method according to an embodiment of this application. The method specifically includes the following steps:
[0055] S501: Get video with time information.
[0056] In this application embodiment, it was found that the grayscale changes caused by the second-to-second transition frame are mainly concentrated in the adjacent columns of the pixel feature map. This feature is very suitable for recognition using convolutional neural networks. Therefore, in order to better identify the second-to-second transition frame image in the video, this application acquires the video with time information so that the second-to-second transition frame image and its initial time region can be identified from the video with time information using a 2D-PDN model based on a convolutional neural network with strong robustness and generalization ability. Then, based on the second-to-second transition frame image and its initial time region, a specified video frame image and its initial time region are determined. The specified video frame image is the next frame of the video after the second-to-second transition frame image in the video. The initial time region of the specified video frame image is in the same position as the time region of the second-to-second transition frame image.
[0057] It should be noted that a video with time information can be a video in which the number in the seconds box changes periodically from 0 to 9. The video with time information includes multiple video frames.
[0058] It should also be noted that the video with time information can be security video collected in real time at bank branches.
[0059] S502: Detect the video using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where the second transition occurs.
[0060] In this embodiment, a 2D-PDN model based on a convolutional neural network (CNN) can be constructed. This allows for the determination of multiple second pixels from all pixels in the video after acquiring a video with time information. The grayscale variation patterns of these second pixels are then analyzed to identify the second-conversion frame image and its initial time region from the video. Furthermore, based on the second-conversion frame image and its initial time region, a specified video frame image and its initial time region can be determined. The structure diagram of the 2D-PDN model is shown below. Figure 6 As shown.
[0061] It should be noted that the specified video frame image is the next video frame image after the video frame image where the second transition occurs.
[0062] Optionally, the process of using a 2D-PDN model based on a convolutional neural network to detect the video and identify the initial time region of a specified video frame image is as follows: traverse all pixels of the video using a 2D-PDN model based on a convolutional neural network to select multiple stable pixels from all pixels; wherein the gray value sequence of the stable pixels is periodically stable; identify multiple second pixels from all pixels based on each stable pixel using a 2D-PDN model based on a convolutional neural network, and determine the specified video frame image and its initial time region based on each second pixel.
[0063] In some embodiments, for each pixel, the periodicity of that pixel is obtained; if the periodicity of that pixel is as follows... Figure 2 If the periodicity shown is generally unstable, then the pixel is determined to be a non-stable pixel; if the periodicity of the pixel shows a stable trend, then the pixel can be confirmed as a stable pixel.
[0064] In practical implementation, the specific process of using a two-dimensional periodic detection network to detect video and identify the initial time region of a specified video frame image is as follows:
[0065]
[0066] Specifically, a 2D-PDN model based on a convolutional neural network can be used to determine the neighboring pixels of each stable pixel from all pixels, and multiple second pixels can be determined from each neighboring pixel based on the periodic feature map of each neighboring pixel; the 2D-PDN model based on a convolutional neural network can be used to determine the second-transition frame image and the time region on the second-transition frame from the video based on each second pixel, and the previous video frame image of the second-transition frame can be determined as the specified video frame image; wherein, the initial time region of the specified video frame image is in the same position as the time region of the second-transition frame image.
[0067] In some embodiments, 2D-PDN can be used to tile the one-dimensional frame sequence of grayscale values of adjacent pixels of a stable pixel according to the frame rate to obtain a pixel feature map (periodic feature map), see [reference]. Figure 4 Analyze the pixel feature map. If the analysis results indicate that the gray value of the adjacent similarity changes significantly, then the adjacent pixel can be considered a second pixel; otherwise, the adjacent pixel is considered not a second pixel.
[0068] It should be noted that, due to the robustness and generalization ability of 2D-PDN, it can effectively capture single-second pixel regions in different types of time. Therefore, 2D-PDN can effectively detect second pixels.
[0069] In some embodiments, after determining the second pixel, the second pixel grayscale value difference sequence can be further obtained, and the second pixel grayscale value periodic change can be determined based on the second pixel grayscale value difference sequence; if the fluctuation of the second similarity grayscale value periodic change is large, the video frame image to which the second pixel belongs can be determined as the second conversion frame image where the second conversion occurs.
[0070] In summary, since the grayscale changes caused by the second conversion frame are mainly concentrated in adjacent columns of the periodic feature map, and this periodic feature map also highlights the periodic changes in the unit second digit, it has a stronger representation ability for the second image in the time domain. It can not only eliminate the influence of the color difference of the time digit, but this characteristic is also very suitable for recognition using convolutional neural networks. Therefore, this application constructs a corresponding 2D-PDN model based on convolutional neural networks by combining convolutional neural networks and 2D-PDN models. The 2D-PDN model based on convolutional neural networks can effectively and quickly identify the second pixels, and determine the corresponding second conversion frame image and the corresponding initial time region based on the second pixels, thereby determining the specified video frame image and its initial time region, laying a solid foundation for subsequent time detection and recognition.
[0071] S503: Identify the current time of a specified video frame image from the initial time region of the specified video frame image using a pre-built TDN model; wherein, the TDN model is a variant based on a convolutional neural network.
[0072] In this embodiment, a time-to-end recognition (TDN) model based on a convolutional neural network is obtained by modifying the convolutional neural network. This TDN model is used to identify the current time in an initial time region. Unlike traditional detection methods, this application does not employ character segmentation to separately identify different digits of the time. This not only simplifies the recognition method but also provides good performance for changing times. Furthermore, this method requires no manual intervention, resulting in low cost and expanded applicability. The structure of the constructed TDN model is as follows: Figure 7 As shown.
[0073] In the specific execution step S103, the initial time region of the specified video frame image is adjusted, and the adjusted initial time region is input into the pre-built TDN model so that the TDN model can identify the current time of the specified video frame image from the adjusted initial time region.
[0074] In practice, the specific implementation process of using the TDN model to identify the current time in the initial time region is as follows:
[0075]
[0076] Specifically, see Figure 7 The initial time region of a specified video frame image is adjusted, and the adjusted initial time region is input into the TDN model. The TDN model uses a two-dimensional convolutional layer (Conv 2D) to convolve the adjusted initial time region; a pooling layer (MaxPooling2D) is used to pool the convolution result obtained above; two concatenated two-dimensional convolutional layers are used to convolve the pooling result obtained above, and then a pooling layer is used to pool the convolution result obtained above; finally, a random deactivation layer (Dropout), a flatten layer (Flatten), and a random deactivation layer (Dropout) are used in sequence to process the pooling result obtained above in order to identify the current time within the initial time region.
[0077] This application provides a method for time detection in bank branch security videos, which involves acquiring videos with time information; using a 2D-PDN model based on a convolutional neural network to detect the video and identify the initial time region of a specified video frame image; wherein the specified video frame image is the next video frame image after the second-transition frame image in the video where a second transition occurs; and using a pre-built TDN model to identify the current time of the specified video frame image from the initial time region of the specified video frame image; wherein the TDN model is a variant based on a convolutional neural network. Therefore, the technical solution provided in this application utilizes a 2D-PDN model with stronger robustness and generalization ability to construct a 2D-PDN model based on a convolutional neural network. This 2D-PDN model is then used to detect videos with time information, identifying the initial time region of a specified video frame. This solves the problem in existing technologies where algorithms with low robustness and generalization ability cannot adapt to various video time types. Furthermore, this application uses a pre-built TDN model to directly identify the current time of the initial time region end-to-end, eliminating the need to separately identify time digits at different locations or manually adjust parameters. This not only reduces the difficulty of time identification but also lowers operating costs and expands the scope of application.
[0078] Based on the bank branch security video time detection method provided in the above embodiments of this application, correspondingly, this application provides a bank branch security video time detection system, such as... Figure 8 As shown, the bank branch's security video time detection system includes:
[0079] Video acquisition module 801 is used to acquire videos with time information;
[0080] The time region detection module 802 is used to detect video using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where the second transition occurs;
[0081] The time recognition module 803 is used to identify the current time of a specified video frame image from the initial time region of the specified video frame image using a pre-built TDN model; wherein, the TDN model is a variant based on a convolutional neural network.
[0082] This application provides a security video time detection system for bank branches, which acquires video with time information; uses a 2D-PDN model based on a convolutional neural network to detect the video and identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where a second transition occurs; and uses a pre-built TDN model to identify the current time of the specified video frame image from the initial time region of the specified video frame image; wherein, the TDN model is obtained by a variant based on a convolutional neural network. Therefore, the technical solution provided in this application utilizes a 2D-PDN model with stronger robustness and generalization ability to construct a 2D-PDN model based on a convolutional neural network. This 2D-PDN model is then used to detect videos with time information, identifying the initial time region of a specified video frame. This solves the problem in existing technologies where algorithms with low robustness and generalization ability cannot adapt to various video time types. Furthermore, this application uses a pre-built TDN model to directly identify the current time of the initial time region end-to-end, eliminating the need to separately identify time digits at different locations or manually adjust parameters. This not only reduces the difficulty of time identification but also lowers operating costs and expands the scope of application.
[0083] Optionally, a time region detection module is used to detect the initial time region of a specified video frame image by utilizing a 2D-PDN model based on a convolutional neural network. Specifically, this module is used for:
[0084] A 2D-PDN model based on a convolutional neural network is used to traverse all pixels in the video to select multiple stable pixels from all pixels; among them, the gray value sequence of stable pixels is periodically stable.
[0085] The 2D-PDN model based on convolutional neural networks identifies multiple second pixels from all pixels based on each stable pixel, and determines the specified video frame image and its initial time region based on each second pixel.
[0086] Optionally, a time region detection module is used to identify multiple second pixels from all pixels based on each stable pixel using a 2D-PDN model based on a convolutional neural network, and to determine the specified video frame image and its initial time region based on each second pixel. Specifically, this module is used for:
[0087] The neighboring pixels of each stable pixel are determined from all pixels using a 2D-PDN model based on a convolutional neural network, and multiple second pixels are determined from each neighboring pixel based on the periodic feature map of each neighboring pixel.
[0088] The 2D-PDN model based on convolutional neural networks is used to determine the second-transition frame image and the time region on the second-transition frame from the video based on each second pixel, and the previous video frame image of the second-transition frame is determined as the specified video frame image; wherein, the initial time region of the specified video frame image is at the same position as the time region of the second-transition frame image.
[0089] Optionally, a time recognition module is used to identify the current time of a specified video frame image from the initial time region of the specified video frame image using a pre-built TDN model, specifically for:
[0090] The initial time region of the specified video frame image is input into the pre-built TDN model so that the TDN model adjusts the initial time region and identifies the current time of the specified video frame image from the adjusted video time region.
[0091] Optionally, the time recognition module is also used to synchronize the current time of a specified video frame with the video time of the specified video frame.
[0092] This application also provides a storage medium storing program instructions, which, when loaded and executed by a processor, implement any of the above-described embodiments of the bank branch security video time detection method.
[0093] This application also provides an electronic device, such as Figure 9 As shown, the device includes a processor 901 and a memory 902, which are connected via a bus. The memory stores program instructions. The processor calls the program instructions in the memory to execute any of the above-described embodiments of the bank branch security video time detection method.
[0094] The processor mentioned in this article can be the terminal's CPU, an integrated MCU within the terminal, or a combination of a CPU and an MCU. Furthermore, the processor contains a kernel that retrieves the corresponding program from memory; one or more kernels can be configured.
[0095] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.
[0096] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.
[0097] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0098] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
[0099] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for detecting the time of security videos in bank branches, characterized in that, The method includes: Retrieve videos with time information; The video is detected using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where a second transition occurs; The current time of the specified video frame image is identified from the initial time region of the specified video frame image using a pre-built TDN model; wherein the TDN model is a variant based on a convolutional neural network.
2. The method according to claim 1, characterized in that, The video is detected using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image, including: The video is traversed using a 2D-PDN model based on a convolutional neural network to select multiple stable pixels from all pixels; wherein the grayscale value sequence of the stable pixels is periodically stable. The 2D-PDN model based on the convolutional neural network identifies multiple second pixels from all pixels based on each of the stable pixels, and determines a specified video frame image and its initial time region based on each of the second pixels.
3. The method according to claim 2, characterized in that, The 2D-PDN model based on a convolutional neural network identifies multiple second pixels from all pixels based on each of the stable pixels, and determines a specified video frame image and its initial time region based on each of the second pixels, including: The 2D-PDN model based on the convolutional neural network is used to determine the neighboring pixels of each stable pixel from all pixels, and multiple second pixels are determined from each neighboring pixel based on the periodic feature map of each neighboring pixel. The 2D-PDN model based on convolutional neural networks determines the second-transition frame image and the time region on the second-transition frame from the video based on each second pixel, and determines the previous video frame image of the second-transition frame as the designated video frame image; wherein, the initial time region of the designated video frame image is at the same position as the time region of the second-transition frame image.
4. The method according to claim 1, characterized in that, Identifying the current time of the specified video frame image from its initial time region using a pre-built TDN model includes: The initial time region of the specified video frame image is adjusted, and the adjusted initial time region is input into a pre-constructed TDN model so that the TDN model can identify the current time of the specified video frame image from the adjusted initial time region.
5. The method according to claim 1, characterized in that, The method further includes: Synchronize the current time of the specified video frame image with the video time of the specified video frame image.
6. A security video time detection system for bank branches, characterized in that, The system includes: The video acquisition module is used to acquire videos with time information. The time region detection module is used to detect the video using a 2D-PDN model based on a convolutional neural network to identify the initial time region of a specified video frame image; wherein, the specified video frame image is the next video frame image after the second-transition frame image in the video where the second transition occurs; The time recognition module is used to identify the current time of the specified video frame image from the initial time region of the specified video frame image using a pre-built TDN model; wherein the TDN model is a variant based on a convolutional neural network.
7. The system according to claim 6, characterized in that, A time region detection module, which uses a 2D-PDN model based on a convolutional neural network to detect the video and identify the initial time region of a specified video frame image, is specifically used for: The video is traversed using a 2D-PDN model based on a convolutional neural network to select multiple stable pixels from all pixels; wherein the grayscale value sequence of the stable pixels is periodically stable. The 2D-PDN model based on the convolutional neural network identifies multiple second pixels from all pixels based on each of the stable pixels, and determines a specified video frame image and its initial time region based on each of the second pixels.
8. The system according to claim 7, characterized in that, The 2D-PDN model based on a convolutional neural network identifies multiple second pixels from all pixels based on each stable pixel, and a time region detection module determines a specified video frame image and its initial time region based on each second pixel. Specifically, this module is used for: The 2D-PDN model based on the convolutional neural network is used to determine the neighboring pixels of each stable pixel from all pixels, and multiple second pixels are determined from each neighboring pixel based on the periodic feature map of each neighboring pixel. The 2D-PDN model based on convolutional neural networks determines the second-transition frame image and the time region on the second-transition frame from the video based on each second pixel, and determines the previous video frame image of the second-transition frame as the designated video frame image; wherein, the initial time region of the designated video frame image is at the same position as the time region of the second-transition frame image.
9. An electronic device, characterized in that, include: A processor and a memory are connected via a bus; wherein the processor is used to call and execute a program stored in the memory; The memory is used to store a program for implementing the bank branch security video time detection method as described in any one of claims 1-5.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for performing the bank branch security video time detection method as described in any one of claims 1-5.