Insurance claim settlement video frame filtering method and device, equipment and medium

Abstract

The invention relates to the field of big data processing technology and financial science and technology, and discloses an insurance claim settlement video frame filtering method, device, equipment and medium, a to-be-filtered target video in insurance claim settlement is acquired, each frame of image in the target video is segmented into N sub-images, and N is greater than or equal to 2; performing graying processing on each sub-image to obtain a pixel difference value between each sub-image in each frame of image and the corresponding sub-image of the previous frame of image, and filtering each sub-image in each frame of image according to the pixel difference value and a preset dynamic threshold value; and recombining all the filtered sub-images to generate a new video frame image. According to the scheme of the invention, the problem that the frame filtering efficiency is low when the insurance claim settlement video frame is filtered by the existing frame filtering method is solved.

Description

Technical Field

[0001] The present invention relates to the fields of big data processing technology and financial technology, and in particular to a method, device, equipment and medium for filtering insurance claim video frames. Background Art

[0002] Insurance claims video frame filtering is the process of screening and processing the captured video at the frame level during the insurance claims process, especially in the scenario of auto insurance claims, or for personal insurance claims for high-speed sports (such as racing, skiing, extreme sports, etc.), in order to improve the efficiency of damage assessment and claims.

[0003] Currently, existing video analysis uses computer vision technology to interpret the images of video frames and extract key information such as characters and postures. However, a video contains a lot of frames, and direct video analysis will consume a lot of computing resources and network bandwidth. For this reason, researchers are committed to designing frame filtering solutions to filter out video frames that do not need to be analyzed.

[0004] Traditional frame filtering solutions often use frame difference technology to perform frame filtering. The specific approach is to calculate the pixel difference between each frame and each point of the previous frame. If the pixel difference does not exceed a certain threshold, it is considered that the pixel contents of the two frames are very similar. If the sub-image contents of frames with very similar pixel contents are also very similar, there is no need to perform analysis, and the result of the previous frame can be used directly.

[0005] However, this traditional frame filtering method is not suitable for insurance claim video frame filtering tasks, because the vehicles or people in the insurance claim video are moving at high speed, resulting in large image pixel differences between each frame, and thus it is impossible to filter out frames with small pixel differences by comparing pixels as described above. Summary of the invention

[0006] The present invention provides a method, device, equipment and medium for filtering insurance claim video frames, so as to solve the problem of low frame filtering efficiency when the existing frame filtering method filters the insurance claim video frames.

[0007] In a first aspect, a method for filtering insurance claim video frames is provided, comprising: Obtain a target video to be filtered in an insurance claim, and divide each frame image in the target video into N sub-images, where N≥2; Grayscale processing is performed on each of the sub-images to obtain a pixel difference value between each sub-image in each frame of image and a corresponding sub-image in a previous frame of image, and each sub-image in each frame of image is filtered according to the pixel difference value and a preset dynamic threshold; All filtered sub-images are recombined to generate a new video frame image.

[0008] In a second aspect, a filtering device for insurance claim video frames is provided, comprising: An acquisition module is used to acquire a target video to be filtered in insurance claims, and divide each frame image in the target video into N sub-images, where N≥2; A filtering module, used to grayscale each of the sub-images to obtain a pixel difference value between each sub-image in each frame of image and a corresponding sub-image in a previous frame of image, and filter each sub-image in each frame of image according to the pixel difference value and a preset dynamic threshold; The result module is used to recombine all filtered sub-images to generate a new video frame image.

[0009] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for filtering insurance claim video frames when executing the computer program.

[0010] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned insurance claim video frame filtering method are implemented.

[0011] In the scheme implemented by the above-mentioned insurance claim video frame filtering method, device, computer equipment and storage medium, each frame image in the target video can be divided into several blocks of sub-images to obtain the pixel difference between each sub-image in each frame image and the corresponding sub-image of the previous frame image, and the sub-image of each frame image is filtered according to the pixel difference of each sub-image and the dynamic threshold; all the sub-images remaining after filtering are recombined according to the set rules to generate a new image, thereby obtaining a filtered new video.

[0012] Because the analysis target in the insurance claim video has a high-speed nature, it is difficult for traditional frame filtering methods to perform frame filtering by analyzing the pixel differences between frames. The scheme of the present invention makes full use of the characteristics that the analysis target in the insurance claim video is in high-speed motion while the surrounding objects often do not move much. By dividing the video frame into blocks and performing pixel difference calculations respectively, the difference value of each sub-image is obtained, and some sub-images are filtered out by comparing the difference value with the dynamic threshold, thereby effectively filtering out most of the sub-images that do not need to be analyzed, greatly speeding up the analysis speed of the insurance claim video, saving the computing resources of the equipment and the bandwidth resources of the network, thereby solving the problem of low frame filtering efficiency when the existing frame filtering method filters the insurance claim video frame. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0014] Figure 1 It is a schematic diagram of an application environment of a method for filtering insurance claim video frames in one embodiment of the present invention; Figure 2 It is a flow chart of a method for filtering insurance claim video frames in one embodiment of the present invention; Figure 3 is a schematic diagram of frame image segmentation of a filtering method for insurance claim video frames in one embodiment of the present invention; Figure 4 is a schematic diagram of the network structure of a DQN algorithm of a method for filtering insurance claim video frames in one embodiment of the present invention; Figure 5 It is a structural schematic diagram of a filtering device for insurance claim video frames in one embodiment of the present invention; Figure 6 is a schematic diagram of a structure of a computer device in one embodiment of the present invention; Figure 7 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] The present invention provides a method for filtering insurance claim video frames, which can be applied in the following aspects: Figure 1In an application environment, the application environment includes a client and a server, wherein the client can communicate with the server through a network. Specifically, the server obtains the target video to be filtered in the insurance claim, divides each frame image in the target video into N sub-images, N≥2; grayscales each sub-image to obtain the pixel difference value between each sub-image in each frame image and the corresponding sub-image of the previous frame image, and filters each sub-image in each frame image according to the pixel difference value and the preset dynamic threshold; recombines all filtered sub-images to generate a new video frame image, and sends the new video frame image to the terminal client. Among them, the client is also called the user end, which refers to the program corresponding to the server and providing local services to customers. The client can be installed on but not limited to various personal computers, laptops, smart phones, tablets and portable wearable devices. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0017] See also Figure 2 As shown, Figure 2 A flowchart of a method for filtering insurance claim video frames provided by an embodiment of the present invention includes the following steps: S01: Obtain a target video to be filtered in insurance claims, and divide each frame image in the target video into N sub-images, where N≥2.

[0018] Among them, the target video to be filtered in insurance claims is the video that needs frame filtering. For example, the insurance company needs to analyze the vehicle driving video or the sports video of high-speed sports in accident insurance when making insurance claims. Most accident insurances include accidental medical insurance, which means that if the hand is injured due to an accident (such as sprains and strains in badminton), medical expenses can be compensated in theory.

[0019] Specifically, in this embodiment, each frame in the insurance claim video is divided into N sub-images. For each frame At, the N sub-images are respectively recorded as At1, ..., AtN, where N≥2.

[0020] S02: grayscale each of the sub-images to obtain a pixel difference value between each sub-image in each frame image and a corresponding sub-image in a previous frame image, and filter each sub-image in each frame image according to the pixel difference value and a preset dynamic threshold.

[0021] Among them, in order to better compare the pixel differences, these images are grayed, that is, the original three-channel color images are converted into one-channel grayscale images. Then the pixel differences of Ati and A(t-1)i are compared to obtain the pixel difference value between each sub-image in each frame and the corresponding sub-image of the previous adjacent image. Image graying is a basic operation in image processing. It converts a color image into a grayscale image, that is, each pixel has only one brightness value, instead of the three components of red, green, and blue in the color image. There are several general ways to grayscale an image: (1) Quantization. The brightness of the three components in the color image is used as the grayscale values ​​of the three grayscale images. A grayscale image can be selected according to application needs.

[0022] (2) Maximum value method: The maximum value of the three components of brightness in the color image is used as the grayscale value of the grayscale image.

[0023] (3) Average value method: The three brightness components in the color image are averaged to obtain a gray value.

[0024] (4) Weighted average method. According to the importance and other indicators, the three components are weighted averaged with different weights. Since the human eye is most sensitive to green and least sensitive to blue, a more reasonable grayscale image can be obtained by weighted averaging the three components of RGB according to the following formula.

[0025] Specifically, the grayscale processing of the image in this embodiment uses the average method to calculate the grayscale value. First, the color image corresponding to each sub-image is converted into a grayscale image using the average method, that is, each pixel of each sub-image has only one brightness value after the conversion, and the grayscale value of each sub-image is calculated based on the brightness value. The pixel difference value is calculated based on the grayscale value of each sub-image in each frame image and the grayscale value of the sub-image corresponding to the previous frame image, and the sub-image of each frame image is filtered based on the pixel difference value of each sub-image and the preset dynamic threshold.

[0026] S03: Recombining all filtered sub-images to generate a new video frame image.

[0027] Specifically, all the sub-images remaining after filtering are first grouped according to the set rules, and then recombined to generate new video frame images. The video frame images are then synthesized into a new video. After analyzing the new video, key information such as characters and postures are extracted.

[0028] This application takes the scenario of insurance claim video frame analysis for accidental medical insurance purchased for badminton sports as an example. Figure 3 As shown, by observing the badminton match video, it is found that the doubles match of badminton has a characteristic, that is, often a person is constantly moving, and the other person is on standby in the same place because there is no ball right, so it can be found that the partial area of ​​the image is that the pixel hardly changes, and the partial area is that the pixel changes dynamically. Therefore, in the present embodiment, the badminton match video is used as the target video, and is applied among the above-mentioned S01, S02 and S03 steps. Specifically, in the present embodiment, each frame of the badminton match video is divided into 4 sub-images, and for each frame At, these four sub-images are respectively recorded as At1, At2, At3, At4. In order to compare pixel differences better, these pictures are grayed, and the color picture that was originally three channels is changed into the grayscale image of one channel. Then, the pixel differences of Ati and A(t-1)i are compared to obtain the pixel difference value between each sub-image in each frame image and the corresponding sub-image of the previous adjacent frame image, and the sub-image of each frame image is filtered according to the pixel difference value of each sub-image and the preset dynamic threshold; all the sub-images remaining after filtering are first grouped according to the set rules, and then recombined to generate a new video frame image, and then the video frame images are synthesized into a new video.

[0029] Due to the high-speed nature of badminton sports, it is difficult for traditional frame filtering methods to perform frame filtering by analyzing pixel differences between frames. The badminton doubles match video frame filtering technology based on block frame difference proposed by the present invention makes full use of the fact that one of the two athletes in a doubles match often does not move much. By dividing the video frames into blocks and performing pixel difference calculations respectively, and then integrating them into a new video, most of the images that do not need to be analyzed are effectively filtered, which greatly speeds up the analysis speed of the badminton doubles match video and saves the computing resources of the equipment and the bandwidth resources of the network.

[0030] In one embodiment, obtaining the pixel difference value between each sub-image in each frame of image and the corresponding sub-image in the previous frame of image includes: The pixel difference between each sub-image in each frame image and each pixel point in the sub-image corresponding to the previous adjacent frame image is determined.

[0031] The average value of the pixel difference of the current sub-image is determined based on the absolute value of the pixel difference between each pixel point, and the pixel difference value of the corresponding sub-image is obtained.

[0032] Specifically, the pixel difference in grayscale value between each pixel in the i-th sub-image Ati in the current frame image At and the i-th sub-image A(t-1)i corresponding to the previous adjacent frame image A(t-1) is calculated, where i=1,2,3,4.

[0033] Take the absolute value of the pixel difference between the grayscale values ​​of each pixel to determine the average value of the pixel difference of the current frame image At ,average value That is the pixel difference value of the current frame image At. The calculation formula is as follows: in, and They represent the pixel values ​​of the corresponding pixels in the two sub-images respectively, and m and n represent the width and height of the sub-images.

[0034] The calculation of the pixel difference value as another implementation of this embodiment can also be performed by calculating the mean square error (MSE). The mean square error is a method for calculating the similarity of images. It calculates the difference between the pixel levels of two images and expresses the similarity as the mean of the square of the difference between the pixel values. The calculation formula of MSE is as follows: in, and They represent the pixel values ​​of the corresponding pixels in the two sub-images respectively, and m and n represent the width and height of the sub-images.

[0035] In one embodiment, filtering the sub-image of the current frame according to the pixel difference value of each sub-image and a preset dynamic threshold value includes: The current sub-image is obtained, and if it is determined that the pixel difference value of the current sub-image is less than the dynamic threshold, the current sub-image is deleted.

[0036] If it is determined that the pixel difference value of the current sub-image is not less than the dynamic threshold, the current sub-image is retained.

[0037] Specifically, compare the average values ​​calculated above , if the average value of the current sub-image If the value is smaller than the preset dynamic threshold A, the current sub-image is deleted and filtered out, and the current sub-image of the frame image of the video is no longer interpreted.

[0038] If the average value of the current sub-image When it is not less than the preset dynamic threshold A, the current sub-image is saved in the current saving order, so as to facilitate the subsequent interpretation of the current sub-image of the frame image of the video and extract key information such as the person and posture therein.

[0039] In one embodiment, the step of setting the preset dynamic threshold comprises: An initial dynamic threshold is preset, and the current sub-image is deleted or retained according to the determination result between the pixel difference value of the current sub-image and the initial dynamic threshold.

[0040] According to a preset time interval, the initial dynamic threshold is updated using a reinforcement learning algorithm to obtain a first dynamic threshold, and the current sub-image is deleted or retained according to a determination result between a pixel difference value of the current sub-image and the first dynamic threshold.

[0041] By analogy, according to the preset time interval, the reinforcement learning algorithm is used to update the last M-1th dynamic threshold to obtain the Mth dynamic threshold. According to the pixel difference value of the current sub-image and the judgment result of the Mth dynamic threshold, the current sub-image is deleted or retained.

[0042] Specifically, in this embodiment, the DQN algorithm in the reinforcement learning algorithm is used to update the above dynamic threshold.

[0043] Among them, DQN, or Deep Q-Network, is an algorithm that combines deep learning and reinforcement learning, and is mainly used to solve optimization problems in the decision-making process. The core idea of ​​DQN is to use a deep neural network to estimate the Q value corresponding to each action, that is, given a state and an action, the network can output the expected return that can be obtained by taking this action in this state. During the training process, DQN updates the network weights by constantly trying different actions and observing the results, thereby gradually learning the optimal strategy.

[0044] like Figure 4 As shown in the figure, the key technologies of the DQN algorithm include Experience Replay and Target Network. Experience Replay is a method of storing past experience (state, action, reward, and next state) and randomly extracting a portion of experience to update the network during training. This method helps to break the correlation between samples and improve the stability and efficiency of training. The target network is the network used to calculate the target Q value in the DQN algorithm. Its weights are copied from the current network at regular intervals to stabilize the training process and avoid oscillations during training.

[0045] In this embodiment, the reward in the DQN algorithm is set to a weighted sum of the video analysis accuracy and the video analysis delay after filtering according to the current dynamic threshold.

[0046] in, is the reward value in the DQN algorithm, is the analytical accuracy, is the analysis delay, and is a weight value parameter, which can be adjusted and set as needed. In this embodiment, =1, =-1.

[0047] Video analysis accuracy and analysis delay are two important indicators in video analysis technology, which measure the accuracy and real-time performance of video analysis respectively.

[0048] Video analysis accuracy refers to the accuracy of the video analysis system in identifying, classifying, detecting and other operations on video content. It is usually affected by many factors. Among them, the clarity, resolution, frame rate, etc. of the video itself will also affect the analysis accuracy. High-quality video can provide more detailed information, which helps the algorithm to analyze more accurately. Frame rate (Framerate) refers to the frequency (rate) at which bitmap images in units of frames appear continuously on the display. In this embodiment, filtering the sub-image will cause a decrease in frame rate. If the frame rate is too small, the accuracy of the final video analysis may be low. If the frame rate is too high, it will affect the video analysis delay, which may result in a longer analysis delay, affecting the output of the video analysis results, thereby affecting the judgment results of the game. In addition, it will also consume a lot of computing resources and network bandwidth.

[0049] Analysis delay refers to the time required for the video analysis system to complete the analysis and output the results after receiving the video data. It measures the real-time performance of the video analysis system. The length of the analysis delay is affected by many factors, among which the number of frames in the video is one of the factors affecting the analysis delay. The larger the number of frames in the video, the longer the analysis delay may be. The smaller the number of frames in the video, the shorter the analysis delay will be.

[0050] In practical applications, video analysis accuracy and analysis latency often need to be weighed. On the one hand, the pursuit of high accuracy may require a relatively high frame rate, more complex algorithms, and higher hardware performance, thereby increasing the analysis latency; on the other hand, reducing the analysis latency may require reducing the frame rate, simplifying the algorithm, or reducing the hardware requirements, which may affect the analysis accuracy. Therefore, in specific application scenarios, it is necessary to comprehensively consider these two indicators based on actual needs and requirements, and take corresponding measures to optimize them.

[0051] In this embodiment, the state space in the DQN algorithm is set to the value range of the dynamic threshold A, where the value range of A is 0 to 255.

[0052] In this embodiment, the step value is set to 5, and the action in the DQN algorithm is set to adjust the dynamic threshold A according to the set step value 5. One adjustment means adding 5 to the current dynamic threshold A, or subtracting 5 from the current dynamic threshold A.

[0053] The specific implementation steps of the DQN algorithm are as follows: (1) Initialize the Q network Q(s,a;θ) and the target network Q(s,a; ). The two networks have the same structure but independent parameters. The parameters of the target network The corresponding parameters θ are copied from the Q network to the target network periodically. A deep neural network is used as the Q value network and its weight parameters are initialized. This network will be used to estimate the Q value of each state-action pair.

[0054] (2) Initialize the experience replay pool D to store the agent's transfer , in order to reduce the correlation between samples and improve sample utilization.

[0055] (3) Obtain the initial state information s_1 by randomly selecting actions in the environment.

[0056] (4) For each step t, perform the following operations: 1) According to the ε-greedy strategy, an action a is randomly selected with a probability of ε, and an action a is randomly selected with a probability of 1-ε Q(s,a;θ).

[0057] 2) Perform action a in the environment and observe the reward r and the new state s_{t+1}.

[0058] 3) Transfer Deposit into experience replay pool D.

[0059] 4) Randomly select a batch of samples from the experience replay pool D.

[0060] 5) For each sample drawn, calculate the corresponding target Q value. The target value is usually calculated by adding the reward r obtained after executing action a to the discounted value of the maximum Q value in the next state s'. The specific calculation formula is as follows: Where r is the reward, is the next state, γ is the discount factor, is the optimal action for the next state. Using the target network to calculate the target Q value can improve the stability of the algorithm.

[0061] 6) Use SGD (stochastic gradient descent) or other optimization algorithms to update the weight parameters of the neural network to minimize the loss function (usually the mean square error between the target value and the Q network output Q value), and perform gradient descent on the network parameters to minimize the loss function: in, is the target Q value, N is the number of samples extracted, and the Q network corresponding to sample j .

[0062] (5) After a certain number of steps, the parameters of the target network are updated using the parameters θ of the Q network. (6) Repeat step (4) until the agent exhibits stable learning performance or reaches a predetermined number of iterations.

[0063] (7) The training process ends when the agent’s performance meets certain criteria or reaches the iteration limit.

[0064] The Q value in the DQN algorithm is a complex expected value that takes into account immediate rewards as well as possible future rewards. It is not a simple cumulative value of the reward value, but approximates the true value through continuous learning and updating. This process is automatic and does not require manual setting or calculation of Q values. Instead, the algorithm learns and updates the Q value function by continuously interacting with the environment and collecting experience.

[0065] Specifically, in this embodiment, an initial dynamic threshold A0 is preset at the beginning, and the current sub-image is deleted or retained according to the judgment result of the pixel difference value of the current sub-image and the initial dynamic threshold. According to the preset time interval, the initial dynamic threshold A0 is updated using the trained target network to obtain the first dynamic threshold A1, and the current sub-image is deleted or retained according to the judgment result of the pixel difference value of the current sub-image and the first dynamic threshold A1. Similarly, according to the preset time interval, the last M-1th dynamic threshold A M-1 Use the trained target network for update processing to obtain the Mth dynamic threshold A M , according to the pixel difference value of the current sub-image and the Mth dynamic threshold A M The current sub-image is deleted or retained according to the judgment result. In this embodiment, the dynamic threshold output by the trained target network is continuously used to perform the above sub-image filtering operation until the video frame filtering is completed.

[0066] In one embodiment, the step of recombining all filtered sub-images comprises: The remaining sub-images after filtering are grouped according to the order of the corresponding frame images in the target video and the total number of sub-images into which each frame image is divided.

[0067] The sub-images in each group are reassembled according to the grouping result.

[0068] Specifically, in order to ensure that the resolution of the image that is then sent to the neural network for processing remains unchanged, this embodiment will re-stitch the remaining sub-images after filtering into complete images in groups of 4 sub-images according to the order of the corresponding frame images in the original badminton game video to form a new video.

[0069] In one embodiment, the filtering method further comprises: According to the mapping relationship between each sub-image in the new video frame image and the corresponding frame image in the original target video, the new video frame image is restored to the image in the target video.

[0070] Specifically, because some sub-images were discarded before, the new video frame image reassembled here may have content from different frames of the original video. To this end, the scheme of the present invention specifically establishes and maintains a hash table to record the relationship between the Ati image of the original video and the Bkj of the new video frame, where k is the kth frame of the image in the new video, and j represents the jth region of the image in the kth frame. When the new video frame is processed, the new video frame image is restored to the image in the original badminton game video according to the mapping relationship between each sub-image in the new video frame image and the corresponding frame image in the original badminton game video, that is, the mapping relationship stored in the hash table.

[0071] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0072] In one embodiment, a filtering device for insurance claim video frames is provided, and the filtering device for insurance claim video frames corresponds one-to-one to a filtering device method for insurance claim video frames in the above embodiment. Figure 5 As shown, the filtering device for the insurance claim video frame includes an acquisition module 101, a filtering module 102 and a result module 103. The detailed description of each functional module is as follows: The acquisition module 101 is used to acquire a target video to be filtered in insurance claims, and divide each frame image in the target video into N sub-images, where N≥2.

[0073] The filtering module 102 is used to grayscale each of the sub-images to obtain a pixel difference value between each sub-image in each frame image and a corresponding sub-image in a previous frame image, and filter each sub-image in each frame image according to the pixel difference value and a preset dynamic threshold.

[0074] The result module 103 is used to reassemble all the filtered sub-images to generate a new video frame image.

[0075] In one embodiment, the determination module 102 is specifically configured to: The pixel difference between each sub-image in each frame image and each pixel point in the sub-image corresponding to the previous adjacent frame image is determined.

[0076] The average value of the pixel difference of the current sub-image is determined based on the absolute value of the pixel difference between each pixel point, and the pixel difference value of the corresponding sub-image is obtained.

[0077] In one embodiment, the determination module 102 is further configured to: The current sub-image is obtained, and if it is determined that the pixel difference value of the current sub-image is less than the dynamic threshold, the current sub-image is deleted.

[0078] If it is determined that the pixel difference value of the current sub-image is not less than the dynamic threshold, the current sub-image is retained.

[0079] In one embodiment, the sorting module 103 is specifically configured to: The remaining sub-images after filtering are grouped according to the order of the corresponding frame images in the target video and the total number of sub-images into which each frame image is divided.

[0080] The sub-images in each group are reassembled according to the grouping result.

[0081] In one embodiment, the recall module 103 is specifically configured to: According to the mapping relationship between each sub-image in the new video frame image and the corresponding frame image in the original target video, the new video frame image is restored to the image in the target video.

[0082] In one embodiment, the recall module 102 is specifically configured to: Preset an initial dynamic threshold, and delete or retain the current sub-image according to the determination result between the pixel difference value of the current sub-image and the initial dynamic threshold; According to a preset time interval, the initial dynamic threshold is updated by using a reinforcement learning algorithm to obtain a first dynamic threshold, and the current sub-image is deleted or retained according to a determination result between a pixel difference value of the current sub-image and the first dynamic threshold; By analogy, according to the preset time interval, the reinforcement learning algorithm is used to update the last M-1th dynamic threshold to obtain the Mth dynamic threshold. According to the pixel difference value of the current sub-image and the judgment result of the Mth dynamic threshold, the current sub-image is deleted or retained.

[0083] The present invention provides a filtering device for insurance claim video frames, which effectively filters out most of the images that do not need to be analyzed, greatly speeds up the analysis speed of badminton doubles match videos, and saves computing resources of equipment and bandwidth resources of the network.

[0084] For the specific definition of the filtering device for insurance claim video frames, please refer to the definition of the filtering method for insurance claim video frames in the above text, which will not be repeated here. Each module in the above-mentioned filtering device for insurance claim video frames can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0085] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a service-side method for filtering insurance claim video frames.

[0086] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, the client-side functions or steps of a filtering method for insurance claim video frames based on artificial intelligence are realized In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: A target video to be filtered in an insurance claim is obtained, and each frame image in the target video is divided into N sub-images, where N≥2.

[0087] Each of the sub-images is grayed to obtain a pixel difference value between each sub-image in each frame of image and a corresponding sub-image in a previous frame of image, and each sub-image in each frame of image is filtered according to the pixel difference value and a preset dynamic threshold.

[0088] All filtered sub-images are recombined to generate a new video frame image.

[0089] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: A target video to be filtered in an insurance claim is obtained, and each frame image in the target video is divided into N sub-images, where N≥2.

[0090] Each of the sub-images is grayed to obtain a pixel difference value between each sub-image in each frame of image and a corresponding sub-image in a previous frame of image, and each sub-image in each frame of image is filtered according to the pixel difference value and a preset dynamic threshold.

[0091] All filtered sub-images are recombined to generate a new video frame image.

[0092] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0093] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0094] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0095] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should be included in the protection scope of the present invention. The non-company software tools or components that appear in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A method for filtering insurance claim video frames, characterized in that: include: Obtain a target video to be filtered in an insurance claim, and divide each frame image in the target video into N sub-images, where N≥2; Grayscale processing is performed on each of the sub-images to obtain a pixel difference value between each sub-image in each frame of image and a corresponding sub-image in a previous frame of image, and each sub-image in each frame of image is filtered according to the pixel difference value and a preset dynamic threshold; All filtered sub-images are recombined to generate a new video frame image.

2. The method for filtering insurance claim video frames according to claim 1, characterized in that: The step of obtaining the pixel difference value between each sub-image in each frame of image and the corresponding sub-image in the previous frame of image includes: Determine the pixel difference between each sub-image in each frame image and each pixel in the sub-image corresponding to the previous adjacent frame image; The average value of the pixel difference of the current sub-image is determined based on the absolute value of the pixel difference between each pixel point, and the pixel difference value of the corresponding sub-image is obtained.

3. The method for filtering insurance claim video frames according to claim 1 or 2, characterized in that: The filtering of the sub-image of the current frame according to the pixel difference value of each sub-image and a preset dynamic threshold value includes: Acquire a current sub-image, and if it is determined that the pixel difference value of the current sub-image is less than the dynamic threshold, delete the current sub-image; If it is determined that the pixel difference value of the current sub-image is not less than the dynamic threshold, the current sub-image is retained.

4. The method for filtering insurance claim video frames according to claim 1, characterized in that: The step of recombining all filtered sub-images comprises: Grouping the remaining sub-images after filtering according to the order of the corresponding frame images in the target video and the total number of sub-images into which each frame image is divided; The sub-images in each group are reassembled according to the grouping result.

5. The method for filtering insurance claim video frames according to claim 1, characterized in that: The filtering method further comprises: According to the mapping relationship between each sub-image in the new video frame image and the corresponding frame image in the original target video, the new video frame image is restored to the image in the target video.

6. The method for filtering insurance claim video frames according to claim 3, characterized in that: The step of setting the preset dynamic threshold comprises: Preset an initial dynamic threshold, and delete or retain the current sub-image according to the determination result between the pixel difference value of the current sub-image and the initial dynamic threshold; According to a preset time interval, the initial dynamic threshold is updated by using a reinforcement learning algorithm to obtain a first dynamic threshold, and the current sub-image is deleted or retained according to a determination result between a pixel difference value of the current sub-image and the first dynamic threshold; By analogy, according to the preset time interval, the reinforcement learning algorithm is used to update the last M-1th dynamic threshold to obtain the Mth dynamic threshold. According to the pixel difference value of the current sub-image and the judgment result of the Mth dynamic threshold, the current sub-image is deleted or retained.

7. A filtering device for insurance claim video frames, characterized in that: include: An acquisition module is used to acquire a target video to be filtered in insurance claims, and divide each frame image in the target video into N sub-images, where N≥2; A filtering module, used to grayscale each of the sub-images to obtain a pixel difference value between each sub-image in each frame of image and a corresponding sub-image in a previous frame of image, and filter each sub-image in each frame of image according to the pixel difference value and a preset dynamic threshold; The result module is used to recombine all filtered sub-images to generate a new video frame image.

8. The device for filtering insurance claim video frames according to claim 7, characterized in that: The filtering module comprises: A first filtering submodule, used to determine the pixel difference between each sub-image in each frame image and each pixel point in the sub-image corresponding to the previous adjacent frame image; The second filtering submodule is used to determine the average value of the pixel difference of the current sub-image based on the absolute value of the pixel difference between each pixel point, and obtain the pixel difference value of the corresponding sub-image.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for filtering the insurance claim video frame according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for filtering insurance claim video frames as described in any one of claims 1 to 7 are implemented.

Images