A surveillance video information stream storage method based on blockchain

Abstract

The present invention belongs to the field of computer blockchain security technology, and in particular, is a method for storing surveillance video information streams based on blockchain. In view of the fact that video information stored in the prior art cannot be stored in real time, it is easy to cause fraud, which makes the real-time storage of video information exist in name only. The following scheme is proposed, which includes the following steps: S1: making a data set; S2: constructing a loss function; S3: training a neural network; S4: transmitting a video stream recognized by the model; S5: passing blockchain verification. The present invention uses a method for real-time storage of information based on irreversible verification information of blockchain, especially in the process of blockchain verification, the real-time data is judged by the trained neural network whether it is related to the previous information, which further strengthens the upload monitoring of real-time video and prevents the uploaded video information from being fraudulent data.

Description

Technical Field

[0001] The present invention relates to the field of computer blockchain security technology, and in particular to a monitoring video information stream storage method based on blockchain. Background Art

[0002] Blockchain is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. In a narrow sense, blockchain is a chain data structure that combines data blocks in a sequential manner in chronological order, and a distributed ledger that is tamper-proof and unforgeable by cryptography. In a broad sense, blockchain technology is a new distributed infrastructure and computing method that uses block chain data structure to verify and store data, uses distributed node consensus algorithm to generate and update data, uses cryptography to ensure the security of data transmission and access, and uses smart contracts composed of automated script code to program and operate data. At present, many areas need to conduct video surveillance, especially the current dashcams and remote places to set up security cameras, which can well protect people's safety, so it is necessary to store this information, but the current storage method is through hard disk or upload to the cloud. In this way, it is impossible to know whether the information has been tampered with. The blockchain technology currently used is that after each distributed node generates a new block, it waits for the distributed nodes of the entire network to vote on the generated new block. Generally, after the new block is recognized, the next new block will be generated. The voting mechanism here is also an information comparison mechanism, and a similar solution is to directly verify through information defined by people. In this way, even if the stored information is not real-time data, it will be falsified because the verification information defined by humans is matched with the verification information of the blockchain.

[0003] As for the video information that needs to be stored at present, since it cannot be stored in real time, it is easy to be falsified, thus making the real-time storage of video information exist in name only. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that video information stored in the prior art cannot be stored in real time and is prone to fraud, thereby making the real-time storage of video information exist in name only. A method for storing surveillance video information stream based on blockchain is proposed.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for storing surveillance video information streams based on blockchain, comprising the following steps:

[0007] S1: Create dataset;

[0008] S2: Construction of loss function;

[0009] S3: neural network training;

[0010] S4: Transmission video stream for model recognition;

[0011] S5: Blockchain verification passed.

[0012] Preferably, in S1, the spatial position information and time information between consecutive video frames are two-dimensionally jointly trained, that is, the spatial position information is trained alone, and the spatial position information and time information are trained together. This can well remember the spatial position changes between video frames, as well as the changes in spatial position following the changes in time information, obtain the correlation between video frames, and identify the incoherence between video frames.

[0013] Preferably, authority authentication is required before making a data set in S1, which is divided into administrator authority and operator authority. Administrator authority can upload, browse and observe, and set browsing authority, while operator authority is limited to browsing authority.

[0014] Preferably, the construction of the loss function adopted in S2 can make blockchain technology as an irreversible upload, so that the information before uploading is very important. The loss function can optimize the training data very well. The loss function of the detection task is very different from the loss function of the classification task. In the loss function of the detection, it is necessary not only to calculate the difference in category confidence and coordinate difference, but also to use various tricks, such as hard negative mining.

[0015] Preferably, the loss function in S2 includes not only the joint training results of spatial position information and time information, but also geographic location information (positioned by GPS), uploaded traffic speed information (using 4G or 5G for traffic statistics) and upload time interval information.

[0016] Preferably, the joint training results of spatial location information and time information, geographic location information, uploaded traffic speed information and upload time interval information are combined in S3 to achieve optimal training of video frames. Subsequently, any uploaded video frame will be compared with the comparison module in the existing neural network model to determine whether it is a transmission recognized by the model. If so, it will be compared with the verification data of the blockchain and enter step S4. If not, it will be directly judged as false data and processed by the processing module.

[0017] Preferably, the comparison module includes a digital information comparison unit, an image comparison unit, a big data comparison unit and a video comparison unit.

[0018] Preferably, the processing module includes a data storage unit, an image storage unit, a video storage unit, a deletion unit, and a warning unit. The data storage unit can store false data information, the image storage unit can store false images, and the false images include synthetic images and unhealthy and dangerous images. The video storage unit can store false videos. The deletion unit can delete false information, and the deleted information is the stored data, images and videos. The warning unit can mark and remind false data, images and videos, and multiple verifications are required before uploading.

[0019] Preferably, in S4, the transmission video stream recognized by the model is verified using blockchain information, and the blockchain verification method also uses spatial location information, geographic location information, uploaded traffic speed information and uploaded time interval information verification method for verification.

[0020] Preferably, if the blockchain verification in S5 is passed, irreversible upload is achieved and cannot be modified, and the data can be well retained. When uploading video information for the second time, the information video to be uploaded can be directly intercepted through the interception module, and then the false data, images and videos stored in the processing module are retrieved for comparison. There is no need to directly judge it as false data or make warning marks through the processing and analysis of S1 and S2. The interception module includes an image interception unit, a data interception unit and a video interception unit. The image interception unit can intercept images, the data interception unit can intercept data, and the video interception unit can intercept videos. The division of labor is clear and efficiency is improved.

[0021] Compared with the prior art, the advantages of the present invention are:

[0022] (1) First, the video,information is processed by using a neural network training method to,strongly remove false data from the video frame information;

[0023] (2) Second, blockchain verification is used to ensure the authenticity of irreversible uploads;

[0024] (3) Third, during the training and operation of the neural network, the same information as the blockchain [spatial location information, geographic location information, uploaded traffic speed information, and uploaded time interval information] is verified for secondary information verification, and no secondary errors in blockchain uploads will be caused by errors in the training of the neural network.

[0025] The present invention adopts a method of storing information in real time based on irreversible verification information of blockchain. In particular, during the verification process of blockchain, a trained neural network is used to judge whether the real-time data is related to the previous information, thereby further strengthening the upload monitoring of real-time video and preventing the uploaded video information from being false data. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flowchart of a blockchain-based surveillance video information stream storage method proposed by the present invention;

[0027] Figure 2 A block diagram of a data set for a blockchain-based surveillance video information stream storage method proposed by the present invention;

[0028] Figure 3 This is a neural network training block diagram of a blockchain-based surveillance video information stream storage method proposed by the present invention;

[0029] Figure 4 This is a block diagram of a comparison module of a surveillance video information stream storage method based on blockchain proposed by the present invention;

[0030] Figure 5 This is a permission authentication block diagram of a blockchain-based surveillance video information stream storage method proposed by the present invention;

[0031] Figure 6 This is a processing module block diagram of a blockchain-based surveillance video information stream storage method proposed by the present invention;

[0032] Figure 7 This is an interception block diagram of a blockchain-based surveillance video information stream storage method proposed by the present invention;

[0033] Figure 8 This is a data point verification diagram of the neural network training process of the blockchain-based surveillance video information stream storage method proposed in the present invention. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0035] Embodiment 1

[0036] Reference Figure 1-8 , a surveillance video information stream storage method based on blockchain, comprising the following steps:

[0037] S1: Create dataset;

[0038] S2: Construction of loss function;

[0039] S3: neural network training;

[0040] S4: Transmission video stream for model recognition;

[0041] S5: Blockchain verification passed.

[0042] In this embodiment, S1 performs two-dimensional joint training on the spatial position information and time information between consecutive video frames, that is, the spatial position information is trained alone, and the spatial position information and time information are trained together. This can well remember the spatial position changes between video frames and the changes in the spatial position following the changes in time information, obtain the correlation between video frames, and identify the incoherence between video frames.

[0043] In this embodiment, permission authentication is required before making a data set in S1, which is divided into administrator permission and operator permission. Administrator permission allows uploading, browsing and observing, and setting browsing permission, while operator permission is limited to browsing permission.

[0044] In this embodiment, the construction of the loss function adopted in S2 can make blockchain technology as an irreversible upload, so that the information before uploading is very important. The loss function can optimize the training data very well. The loss function of the detection task is very different from the loss function of the classification task. In the loss function of the detection, it is necessary not only to calculate the difference in category confidence and coordinate difference, but also to use various tricks, such as hard negative mining.

[0045] In this embodiment, the loss function in S2 includes not only the joint training results of spatial location information and time information, but also geographic location information (positioned by GPS), uploaded traffic speed information (using 4G or 5G for traffic statistics) and upload time interval information.

[0046] In this embodiment, the joint training results of spatial location information and time information, geographic location information, uploaded traffic speed information and upload time interval information are combined in S3 to achieve optimal training of video frames. Subsequently, any uploaded video frame will be compared with the comparison module in the existing neural network model to determine whether it is a transmission recognized by the model. If so, it will be compared with the verification data of the blockchain and enter step S4. If not, it will be directly judged as false data and processed by the processing module.

[0047] In this embodiment, the comparison module includes a digital information comparison unit, an image comparison unit, a big data comparison unit and a video comparison unit.

[0048] In this embodiment, the processing module includes a data storage unit, an image storage unit, a video storage unit, a deletion unit, and a warning unit. The data storage unit can store false data information, the image storage unit can store false images, and the false images include synthetic images and unhealthy and dangerous images. The video storage unit can store false videos. The deletion unit can delete false information, and the deleted information is the stored data, images and videos. The warning unit can mark and remind false data, images and videos, and multiple verifications are required before uploading.

[0049] In this embodiment, S4 uses blockchain information verification for the transmission video stream recognized by the model. The blockchain verification method also uses spatial location information, geographic location information, uploaded traffic speed information and uploaded time interval information verification method for verification.

[0050] In this embodiment, if the blockchain verification in S5 is passed, irreversible upload is achieved and cannot be modified, so the data can be well preserved. When uploading video information for the second time, the information video to be uploaded can be directly intercepted through the interception module, and then the false data, images and videos stored in the processing module are retrieved for comparison. There is no need to directly judge it as false data or make warning marks through the processing and analysis of S1 and S2. The interception module includes an image interception unit, a data interception unit and a video interception unit. The image interception unit can intercept images, the data interception unit can intercept data, and the video interception unit can intercept videos. The division of labor is clear and efficiency is improved.

[0051] Embodiment 2

[0052] Reference Figure 1-8 , a surveillance video information stream storage method based on blockchain, comprising the following steps:

[0053] S1: Create dataset;

[0054] S2: Construction of loss function;

[0055] S3: neural network training;

[0056] S4: Transmission video stream for model recognition;

[0057] S5: Blockchain verification passed.

[0058] In this embodiment, S1 performs two-dimensional joint training on the spatial position information and time information between consecutive video frames, that is, the spatial position information is trained alone and the spatial position information and time information are trained together. This can well remember the spatial position changes between video frames and the changes in the spatial position following the changes in time information, obtain the correlation between video frames, identify the incoherence between video frames, visualize each frame of video, and then digitize the image, attach a corresponding label to each image, and use the Image module in PIL to digitize the image information.

[0059] In this embodiment, permission authentication is required before making a data set in S1, which is divided into administrator permission and operator permission. Administrator permission allows uploading, browsing and observing, and setting browsing permission, while operator permission is limited to browsing permission.

[0060] In this embodiment, the construction of the loss function adopted in S2 can make blockchain technology as an irreversible upload, so that the information before uploading is very important. The loss function can optimize the training data very well. The loss function of the detection task is very different from the loss function of the classification task. In the loss function of the detection, it is necessary not only to calculate the difference in category confidence and coordinate differences, but also to use various tricks. The loss function is the fluctuation of quality characteristics (that is, the deviation of product performance from the design target value) is the cause of quality loss and quality problems. At present, researchers have established a quality loss function to describe the relationship between quality loss and quality fluctuation. Quality loss QL (QualityLoss) is a function of quality characteristic y. Different products and different quality characteristics correspond to different quality loss curves. When the product performance is exactly the target value m, the quality loss is the smallest, and the relative value can be defined as zero. The farther the product performance deviates from the target value, the greater the quality loss. The graph of the quality loss function L (y) is a curve with a minimum value of zero at y = m. Assuming that L(y) has a second-order derivative at y=m, L(y) can be expanded at y=m into a Taylor series. Considering L(y)=0, L￠(m)=0, and ignoring higher-order infinitesimals, L(y) can be simplified to where k=L￠￠(m) / 2! is a constant that does not depend on y. Therefore, the image of the quality loss function is approximately equal to a parabola near y=m. j(y) is the performance probability distribution density function of a batch of products, with a mean of μ and a standard deviation of σ. The mathematical expectation of the quality loss of this batch of products is When the random variable y obeys the normal distribution N(μ, σ2), it can be obtained from formula (1-8). It can be seen that the mathematical expectation L of the quality loss is related to the product performance variance σ2, the square of the average fluctuation (μ-m)2 and the loss coefficient k. σ2 and (μ-m)2 determine the shape and position of the curve j(y), while k determines the shape of the quality loss function L(y). There are two goals of robust design. One is to minimize [s2+(mm)2], that is, the curve j(y) is very steep and the mean is close to m. The other goal is to minimize k, that is, the curve L(y) is very flat, thereby minimizing the quality loss of the product.

[0061] In this embodiment, the loss function in S2 includes not only the joint training results of spatial location information and time information, but also geographic location information (positioned by GPS), uploaded traffic speed information (using 4G or 5G for traffic statistics) and upload time interval information. Since this application uses blockchain technology as an irreversible upload, information comparison before uploading is very important, and the loss function can optimize the training data, record the upload time, judge the rationality of the data, and the collection and integration time of real information is different from that of false information.

[0062] In this embodiment, the joint training results of spatial location information and time information, geographic location information, uploaded traffic speed information and upload time interval information are combined in S3 to achieve optimal training of video frames. Subsequently, any uploaded video frame will be compared with the comparison module in the existing neural network model to determine whether it is a transmission recognized by the model. If so, it will be compared with the verification data of the blockchain and enter step S4. If not, it will be directly judged as false data and processed by the processing module.

[0063] In this embodiment, the comparison module includes a digital information comparison unit, an image comparison unit, a big data comparison unit and a video comparison unit, and an introduction is made to the digital comparison method, including: Step 1: Setting a configuration table for data comparison, the configuration table includes data comparison control process information, address information of data comparison content and data difference threshold information; Step 2: According to the data comparison control process information in the configuration table, the data to be audited is first taken out from the database, and sorted according to the same field, and the field includes an index field and an attribute field; Step: 3: Compare the data; Step 4: Determine whether the data difference exceeds the data difference threshold set in the configuration table, if so, record the data comparison result, if not, continue to compare; Step: 5: After the comparison is completed, determine whether a second comparison is required based on the attributes of the data. This type of comparison can be used for images, big data and videos.

[0064] In this embodiment, the processing module includes a data storage unit, an image storage unit, a video storage unit, a deletion unit, and a warning unit. The data storage unit can store false data information, the image storage unit can store false images, and the false images include synthetic images and unhealthy and dangerous images. The video storage unit can store false videos. The deletion unit can delete false information, and the deleted information is the stored data, images and videos. The warning unit can mark and remind false data, images and videos, and multiple verifications are required before uploading.

[0065] In this embodiment, S4 uses blockchain information verification for the transmission video stream recognized by the model. The blockchain verification method also uses spatial location information, geographic location information, uploaded traffic speed information, and upload time interval information verification method for verification.

[0066] In this embodiment, if the blockchain verification in S5 is passed, irreversible upload is achieved and cannot be modified, so the data can be well preserved. When uploading video information for the second time, the information video to be uploaded can be directly intercepted through the interception module, and then the false data, images and videos stored in the processing module are retrieved for comparison. There is no need to directly judge it as false data or make warning marks through the processing and analysis of S1 and S2. The interception module includes an image interception unit, a data interception unit and a video interception unit. The image interception unit can intercept images, the data interception unit can intercept data, and the video interception unit can intercept videos. The division of labor is clear and efficiency is improved.

[0067] In this embodiment, the application cleverly utilizes the re-verification of "spatial location information, geographic location information, uploaded traffic speed information and upload time interval information". Compared with neural network training, in order to save time, the blockchain verification information of this application no longer performs time verification, and the other verification methods can better capture the information that the model misjudges, and compared with the various values ​​set by the blockchain [the verification information of the blockchain cannot be changed artificially, the neural network is trained and can be changed], it can meet the current upload requirements and will not let the training data affect the upload accuracy.

[0068] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for storing surveillance video information stream based on blockchain, characterized in that: The following steps are involved: S1: Create dataset; S2: Construction of loss function; S3: neural network training; S4: Transmission video stream for model recognition; S5: Blockchain verification passed; In the S1, the spatial position information and time information between consecutive video frames are trained in two dimensions, that is, the spatial position information is trained alone, and the spatial position information and time information are trained together. This can well remember the spatial position changes between video frames, as well as the changes in spatial position following the changes in time information, obtain the correlation between video frames, and identify the incoherence between video frames; before making a data set in the S1, permission authentication is required, which is divided into administrator permission and operator permission. Administrator permission can upload, browse and observe, and set browsing permissions, while operator permission is limited to browsing permissions; the construction of the loss function in the S2 can use blockchain technology as an irreversible upload, making the information before uploading very important. The loss function can well optimize the training data, the loss function of the detection task, and the loss function of the classification task The loss function of the detection is very different. In the loss function of detection, it is necessary not only to calculate the difference in category confidence and coordinate differences, but also to use various tricks; the loss function in S2 includes not only the joint training results of spatial location information and time information, but also geographic location information, uploaded traffic speed information and uploaded time interval information; in S3, the joint training results of spatial location information and time information, geographic location information, uploaded traffic speed information and uploaded time interval information are combined to achieve the best training for video frames. Next, any uploaded video frame will be compared with the comparison module in the existing neural network model to determine whether it is a transmission recognized by the model. If it is, it will be compared with the verification data of the blockchain and enter step S4. If not, it is directly judged as false data and processed by the processing module. In S4, the transmission video stream recognized by the model is verified by blockchain information. The blockchain verification method also uses spatial location information, geographic location information, uploaded traffic speed information and upload time interval information verification method for verification; if the blockchain verification in S5 is passed, irreversible upload is achieved and cannot be modified, and the data can be well preserved. When uploading the video information for the second time, the information video to be uploaded can be directly intercepted through the interception module, and then the false data, images and videos stored in the processing module are retrieved for comparison. There is no need to directly judge it as false data or make warning marks through the processing and analysis of S1 and S2. The interception module includes an image interception unit, a data interception unit and a video interception unit. The image interception unit can intercept images, the data interception unit can intercept data, and the video interception unit can intercept videos. The division of labor is clear and efficiency is improved.

2. According to a blockchain-based surveillance video information stream storage method according to claim 1, it is characterized in that: The comparison module includes a digital information comparison unit, an image comparison unit, a big data comparison unit and a video comparison unit.

3. According to a blockchain-based surveillance video information stream storage method according to claim 1, it is characterized in that: The processing module includes a data storage unit, an image storage unit, a video storage unit, a deletion unit, and a warning unit. The data storage unit can store false data information, the image storage unit can store false images, and the false images include synthetic images and unhealthy and dangerous images. The video storage unit can store false videos. The deletion unit can delete false information, and the deleted information is the stored data, images and videos. The warning unit can mark and remind false data, images and videos, and multiple verifications are required before uploading.

Images