Emergency rescue scene image transmission method, processing method and emergency communication system

Abstract

The invention provides an emergency rescue scene image transmission method, an emergency rescue scene image processing method and an emergency communication system. An original emergency rescue scene image is received; pre-compressing the original emergency rescue scene image to obtain a preliminary compressed image and a progressive compressed image; performing deep compression on the preliminary compressed image to obtain a compressed image bit stream; the compressed image bit stream and the progressive compressed image are sequentially sent to a cloud control platform, the compressed image bit stream and the progressive compressed image are used for the cloud control platform to reconstruct an original emergency rescue scene image, and the number of subpackages during image transmission is reduced through image pre-compression and deep compression. The defect that the single communication bandwidth of the Beidou short message is low is overcome; and progressive transmission is realized based on pre-compression, so that the image definition is improved while the real-time display requirement is met.

Description

Technical Field

[0001] The present invention relates to the field of emergency rescue technology, and in particular to an emergency rescue scene image transmission method, a processing method and an emergency communication system. Background Art

[0002] In emergency communications, fast and clear transmission of image information is essential for making accurate decisions, conducting on-site assessments, and coordinating rescue efforts. However, Beidou short message communication has limitations on single communication capacity and communication frequency in long-distance data transmission. Therefore, when using Beidou short messages for image transmission, images need to be highly compressed to improve the transmission efficiency of their code streams. Traditional image compression methods use image compression algorithms to compress images, and then transmit them in packets to the cloud platform, where images are obtained through image recovery algorithms. However, this image compression method has problems such as low compression rate, the need to split into a large number of sub-packets for time-sharing transmission, and the image obtained through the image recovery algorithm is not clear enough and has poor real-time performance. It cannot meet the application needs of emergency rescue scenarios. Summary of the invention

[0003] The present invention provides an emergency rescue scene image transmission method, a processing method and an emergency communication system, which are used to solve the defects of traditional image compression methods, such as low compression rate, unclear images after restoration and poor real-time performance.

[0004] The present invention provides an emergency rescue scene image transmission method, which is applied to a local terminal and includes: Receiving original emergency rescue scene images; Pre-compressing the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; Performing deep compression on the preliminary compressed image to obtain a compressed image bit stream; The compressed image bit stream and the progressively compressed image are sent to a cloud control platform in sequence, and the compressed image bit stream and the progressively compressed image are used by the cloud control platform to reconstruct the original emergency rescue scene image.

[0005] According to the emergency rescue scene image transmission method provided by the present invention, the original emergency rescue scene image is pre-compressed to obtain a preliminary compressed image and a progressively compressed image, including: Taking a first average pooling window to compress the original emergency rescue scene image to obtain the preliminary compressed image; Taking a second average pooling window to compress the original emergency rescue scene image to obtain a first-layer progressive compression image; Taking a third average pooling window to perform secondary compression on the first-layer progressively compressed image to obtain a second-layer progressively compressed image; The first average pooling window is larger than the second average pooling window and the third average pooling window.

[0006] According to the emergency rescue scene image transmission method provided by the present invention, taking the second average pooling window to compress the original emergency rescue scene image to obtain a first layer of progressive compression image includes: Divide the original emergency rescue scene image into a plurality of small-block images, calculate a first average value of pixels in each of the small-block images, and use an image represented by the first average value and an image represented by any two vertex pixels in each of the small-block images as the first-layer progressive compression image; The method of taking the third average pooling window to perform secondary compression on the first-layer progressively compressed image to obtain the second-layer progressively compressed image includes: taking the third average pooling window to perform secondary compression on a plurality of the small-block images, dividing each of the small-block images into a plurality of sub-block images, calculating a second average value of pixels in each of the sub-block images, and taking an image represented by the second average value and an image represented by any two vertex pixels in each of the sub-block images as the second-layer progressively compressed image.

[0007] According to the emergency rescue scene image transmission method provided by the present invention, the progressively compressed images are sequentially sent to the cloud control platform, including: The image represented by the second average value, the image represented by any two vertex pixels in each of the sub-block images, the image represented by the first average value, and the image represented by any two vertex pixels in each of the small-block images are sent to the cloud control platform in this order.

[0008] According to the emergency rescue scene image transmission method provided by the present invention, the deep compression of the preliminary compressed image to obtain a compressed image bit stream includes: Performing normalization processing on the preliminary compressed image; Extracting feature data of the normalized image, quantizing the feature data, reducing the dimension of the feature data, and obtaining a feature potential representation vector; Discretizing the feature potential representation vector, inputting the discretized feature potential representation vector into a generative image compression network based on entropy coding compression, and controlling the compressed image output by the generative image compression network based on entropy coding compression based on a conditional generative adversarial network to obtain a compressed image bit stream; The entropy coding-based generative image compression network is obtained by jointly training using a neural network compression loss function and a discriminator loss function of the conditional generative adversarial network.

[0009] According to the emergency rescue scene image transmission method provided by the present invention, the neural network compression loss function is constructed by the perceptual quality optimization item corresponding to the discriminator of the conditional generative adversarial network, the bit rate of the potential representation feature vector, the compressed image distortion rate, the control trade-off hyperparameter and the joint optimization hyperparameter; The discriminator loss function of the conditional generative adversarial network is constructed by the perceptual quality optimization item corresponding to the discriminator of the conditional generative adversarial network.

[0010] According to the emergency rescue scene image transmission method provided by the present invention, the discriminator loss function of the conditional generative adversarial network includes a target bit rate hyperparameter, and the target bit rate hyperparameter is used to adjust the average bit rate of the generative image compression network based on entropy coding compression.

[0011] The present invention also provides an emergency rescue scene image transmission method, which is applied to a cloud control platform, comprising: receiving a compressed image bit stream and a progressively compressed image in sequence; Restoring the blurred original emergency rescue scene image according to the compressed image bit stream; The blurred original emergency rescue scene image and the progressively compressed image are combined to reconstruct a clear original emergency rescue scene image.

[0012] According to the emergency rescue scene image transmission method provided by the present invention, the method of restoring the blurred original emergency rescue scene image according to the compressed image bit stream includes: Performing entropy decoding on the compressed image bit stream to obtain a quantized discrete feature potential representation vector; Dequantizing the quantized discrete feature potential representation vector to restore it into feature data represented by floating point numbers; Reconstructing a decompressed image from the feature data represented by the floating point numbers; The decompressed image is subjected to inverse normalization processing to obtain a blurred original emergency rescue scene image.

[0013] According to the emergency rescue scene image transmission method provided by the present invention, the combining of the blurred original emergency rescue scene image and the progressively compressed image to reconstruct a clear original emergency rescue scene image comprises: After the blurred original emergency rescue scene image is restored, the sequentially received progressively compressed images are superimposed on the blurred original emergency rescue scene image to reconstruct a clear original emergency rescue scene image.

[0014] The present invention also provides an emergency communication system, comprising: a local terminal; The local terminal includes an image acquisition device, a local processing device and a sending device, and the image acquisition device collects original emergency rescue scene images; The local processing device is used to pre-compress the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; and perform deep compression on the preliminary compressed image to obtain a compressed image bit stream; The sending device is used to send the compressed image bit stream and the progressively compressed image to the cloud control platform in sequence.

[0015] The present invention also provides an emergency communication system, comprising: a cloud control platform; The cloud control platform includes a receiving device, a cloud processing device and a cloud platform; the receiving device is used to receive a compressed image bit stream and a progressively compressed image; the cloud processing device is used to restore a blurred original emergency rescue scene image based on the compressed image bit stream; the cloud platform is used to combine the blurred original emergency rescue scene image and the progressively compressed image to reconstruct a clear original emergency rescue scene image.

[0016] The emergency rescue scene image transmission method, processing method and emergency communication system provided by the present invention receive the original emergency rescue scene image; pre-compress the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; deeply compress the preliminary compressed image to obtain a compressed image bit stream; send the compressed image bit stream and the progressively compressed image to a cloud control platform in sequence, and the compressed image bit stream and the progressively compressed image are used by the cloud control platform to reconstruct the original emergency rescue scene image, and the number of subpackets when transmitting the image is reduced through image pre-compression and deep compression, thereby compensating for the low single communication bandwidth of the Beidou short message; and progressive transmission is achieved based on pre-compression, thereby meeting real-time display requirements while improving image clarity. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 This is one of the flow charts of the emergency rescue scene image transmission method provided by an embodiment of the present invention; Figure 2 This is a second flow chart of the emergency rescue scene image transmission method provided by an embodiment of the present invention; Figure 3 is a schematic diagram of the structure of an emergency rescue system provided by an embodiment of the present invention; Figure 4This is the third flow chart of the emergency rescue scene image transmission method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0020] Figure 1 A flowchart of an emergency rescue scene image transmission method provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the emergency rescue scene image transmission method provided by the embodiment of the present invention is applied to a local terminal, and includes: Step 101: receiving an original emergency rescue scene image; Step 102: pre-compress the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; Step 103: deeply compress the preliminary compressed image to obtain a compressed image bit stream; Step 104: Send the compressed image bit stream and the progressively compressed image to the cloud control platform in sequence. The compressed image bit stream and the progressively compressed image are used by the cloud control platform to reconstruct the original emergency rescue scene image.

[0021] In the traditional emergency rescue scene image transmission method, an image compression algorithm is used to compress the image, which is then transmitted in packets to the cloud platform, where the image is obtained through an image recovery algorithm. However, this image compression method has problems such as insufficient compression rate, the need to split the image into a large number of sub-packets for time-sharing transmission, and the image obtained after the image recovery algorithm is not clear enough and has poor real-time performance. It cannot meet the application needs of emergency rescue scenarios.

[0022] The emergency rescue scene image transmission method provided by the embodiment of the present invention receives the original emergency rescue scene image; pre-compresses the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; deeply compresses the preliminary compressed image to obtain a compressed image bit stream; and sends the compressed image bit stream and the progressively compressed image to a cloud control platform in sequence, wherein the compressed image bit stream and the progressively compressed image are used by the cloud control platform to reconstruct the original emergency rescue scene image, and the number of subpackets during image transmission is reduced through image pre-compression and deep compression, thereby compensating for the low single communication bandwidth of Beidou short messages; and progressive transmission is achieved based on pre-compression, thereby meeting real-time display requirements while improving image clarity.

[0023] Based on any of the above embodiments, the pre-compressing the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image includes: Step 201: Take a first average pooling window to compress the original emergency rescue scene image to obtain the preliminary compressed image; Step 202: Taking a second average pooling window to compress the original emergency rescue scene image to obtain a first-layer progressive compression image; Step 203: Take a third average pooling window to perform secondary compression on the first-layer progressively compressed image to obtain a second-layer progressively compressed image; The first average pooling window is larger than the second average pooling window and the third average pooling window.

[0024] In an embodiment of the present invention, the second average pooling window is used to compress the original emergency rescue scene image to obtain a first-layer progressive compression image, including: dividing the original emergency rescue scene image into a plurality of small-block images, calculating a first average value of pixels in each small-block image, and taking an image represented by the first average value and an image represented by any two vertex pixels in each small-block image as a first-layer progressive compression image; In an embodiment of the present invention, taking the third average pooling window to perform secondary compression on the multiple small-block images to obtain a first-layer progressively compressed image includes: taking the third average pooling window to perform secondary compression on the multiple small-block images, dividing each small-block image into multiple sub-block images, calculating a second average value of pixels in each sub-block image, and using an image represented by the second average value and an image represented by any two vertex pixels in each sub-block image as a second-layer progressively compressed image.

[0025] In an embodiment of the present invention, sending the progressively compressed images to the cloud control platform in sequence includes: The image represented by the second average value, the image represented by any two vertex pixels in each sub-block image, the image represented by the first average value, and the image represented by any two vertex pixels in each small block image are sent to the cloud control platform in this order.

[0026] For example, an 8*8 average pooling window is taken to compress the original image, and the image at this time is sent to the High-Fidelity Image Compression Technology (HiFiC) algorithm; a 2*2 average pooling window is taken to compress the original image, and the average of the first two data in each window, the first and the third data are stored, and recorded as A1, A2, A3; then a 2*2 average pooling window is taken to perform secondary compression on the image after pooling compression in step b, and the average of the first two data in each window, the first and the third data are stored, and recorded as B1, B2, B3, and B1, B2, B3, A1, A2, A3 are sent in sequence.

[0027] Due to the use of progressive image compression technology, the images are transmitted in a preset order. At the beginning of the transmission, the cloud control platform can display a low-resolution version of the image. As more data is transmitted, the details and clarity of the image are gradually improved, and finally a high-quality image is displayed.

[0028] Based on any of the above embodiments, the step of deeply compressing the preliminary compressed image to obtain a compressed image bit stream includes: Step 301, normalizing the preliminary compressed image; Step 302: extracting feature data of the normalized image, quantizing the feature data, reducing the dimension of the feature data, and obtaining a feature potential representation vector; Step 303: discretize the feature potential representation vector, input the discretized feature potential representation vector into a generative image compression network based on entropy coding compression, and use a conditional generative adversarial network to control the compressed image output by the generative image compression network based on entropy coding compression to obtain a compressed image bit stream; The entropy coding-based generative image compression network is obtained by joint training using a neural network compression loss function and a discriminator loss function of the conditional generative adversarial network.

[0029] In the embodiment of the present invention, the learnable generative image compression network based on entropy coding is based on Shannon's rate-distortion theory and is modeled by an autoencoder consisting of an encoder E and a decoder G. The input image x is encoded by E to obtain a quantized latent representation y=E(x). During decoding, G is used to obtain a lossy reconstructed image At the same time, we introduce a probability model P for y and combine it with bits r(y)=−log(P(y)) to store y losslessly (ignoring the tiny bit overhead caused by the entropy encoder). Then, we parameterize E, G, and P as a convolutional neural network, and we can jointly optimize the training by minimizing the rate-distortion tradeoff, where λ controls the tradeoff:

[0030] is the basic loss function of the compression system, is the bit rate of the latent representation y, i.e., the number of bits required to store the latent variable after compression, Distortion caused by compressed images.

[0031] Compressing images can cause distortion Including objective loss (mean square error MSE) and perceptual loss (Learning Perceptual Patch Similarity LPIPS). The former is an indicator to measure the difference between image pixel values, and the latter reflects human perception of image differences. By setting the weighting parameters and , the relative importance of these two in the overall loss function can be adjusted:

[0032] In an embodiment of the present invention, the neural network compression loss function is constructed based on the perceptual quality optimization term of the discriminator of the conditional generative adversarial network, the bit rate of the potential representation feature vector, the distortion caused by the compressed image, the control trade-off hyperparameters, and the joint optimization hyperparameters; Using Hyperparameters To jointly optimize the loss function:

[0033] is the loss function of the expanded neural network compression system, It is a perceptual quality optimization term introduced by the GAN discriminator to make the generated compressed image look more realistic. This loss function combines the compression rate, distortion and perceptual quality. The optimization goal is to maintain high image quality at the lowest possible bit rate.

[0034] In an embodiment of the present invention, a conditional generative adversarial network is constructed to control the generated images. The conditional generative adversarial network can solve the problem that the images generated by the original generative adversarial network are unpredictable, and the control of the generated images is achieved by introducing conditional information. In the conditional generative adversarial network, the generator G uses additional information (standard distribution s) as a condition to guide the generation of image data, and the discriminator combines the label s with the input sample to score and judge the authenticity of the data. Through the continuous confrontation between the two, the data generated by the generator is eventually made as similar as possible to the real data distribution under given conditions, and the discriminator is more accurate. Given a real sample and generate samples After fixing s, the formula for optimizing the “unsaturated” loss is as follows:

[0035] , are the loss functions of the generator and the discriminator, respectively. represents the probability that the discriminator judges the generated data as real data, represents the probability that the real data is considered real by the discriminator, represents the probability that the discriminator judges the generated data as “fake” data. This part encourages the discriminator to recognize the compressed images produced by the generator. is fake, i.e. the discriminator should distinguish between the reconstructed image and the real image as correctly as possible. is to encourage the discriminator to recognize real images is real, that is, improving the discriminator's ability to distinguish between real images and generated images. The discriminator's goal is to accurately determine whether the image is real or generated, while the generator tries to "cheat" the discriminator and make the generated image as close to the real image as possible. The game between the two makes the generator continue to improve and generate more realistic images.

[0036] The learnable generative image compression network is jointly trained using the neural network compression loss function and the discriminator loss function of the conditional generative adversarial network to balance the compression quality and image clarity of the generated images.

[0037] By combining the formula and The whole system can find a balance between compression rate, distortion and perceptual quality, and improve the perceptual quality of compressed images through the generative adversarial network mechanism.

[0038] The discriminator loss function of the conditional generative adversarial network includes a target bit rate hyperparameter, and the target bit rate hyperparameter is used to adjust the average bit rate of the generated image compression network based on entropy coding compression. In an embodiment of the present invention, when the generated image compression network is jointly trained, it also includes: Introducing target bitrate hyperparameters , and the neural network compression loss function Replace with an adaptive item , the adaptive term Contains two hyperparameters ,like ,but , used to punish excessively high bit rates, otherwise , used to relax the bit rate limit, setting , so that the average bit rate of the entropy coding-based image compression network is within the target bit rate hyperparameter nearby.

[0039] Figure 2A flowchart of an emergency rescue scene image processing method provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, an embodiment of the present invention provides an emergency rescue scene image processing method, which is applied to a cloud control platform, including: Step 401: receiving a compressed image bit stream and a progressively compressed image in sequence; Step 402: restore the blurred original emergency rescue scene image according to the compressed image bit stream; Step 403: Reconstruct a clear original emergency rescue scene image by combining the blurred original emergency rescue scene image and the progressively compressed image.

[0040] The embodiment of the present invention provides an emergency rescue scene image processing method that realizes progressive transmission based on pre-compression. A low-definition image can be reconstructed when the first image is received. As the number of received subpackets increases, the image clarity continues to improve, thereby meeting real-time requirements while improving image clarity.

[0041] In an embodiment of the present invention, restoring the blurred original emergency rescue scene image according to the compressed image bit stream includes: Step 501: entropy decoding the compressed image bit stream to obtain a quantized discrete feature potential representation vector; Step 502: Dequantize the quantized discrete feature potential representation vector to restore it into feature data represented by floating point numbers; Step 503: reconstructing a decompressed image from the feature data represented by the floating point number; Step 504: perform inverse normalization processing on the decompressed image to obtain a blurred original emergency rescue scene image.

[0042] In an embodiment of the present invention, the step of combining the blurred original emergency rescue scene image and the progressively compressed image to reconstruct a clear original emergency rescue scene image includes: After the blurred original emergency rescue scene image is restored, the sequentially received progressively compressed images are superimposed on the blurred original emergency rescue scene image to reconstruct a clear original emergency rescue scene image.

[0043] Based on any of the above embodiments, an embodiment of the present invention also provides an emergency communication system, comprising: a local terminal; the local terminal comprises an image acquisition device, a local processing device and a sending device, the image acquisition device collects the original emergency rescue scene image; the local processing device is used to pre-compress the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; the preliminary compressed image is deeply compressed to obtain a compressed image bit stream; the sending device is used to send the compressed image bit stream and the progressively compressed image to the cloud control platform in sequence.

[0044] In the embodiment of the present invention, the image acquisition device, the local processing device and the sending device may be integrated inside the local terminal, or may be separate hardware devices.

[0045] Based on any of the above embodiments, an embodiment of the present invention also provides an emergency communication system, including: a cloud control platform; the cloud control platform includes a receiving device, a cloud processing device and a cloud platform; the receiving device is used to receive a compressed image bit stream and a progressively compressed image; the cloud processing device is used to restore the blurred original emergency rescue scene image according to the compressed image bit stream; the cloud platform is used to combine the blurred original emergency rescue scene image and the progressively compressed image to reconstruct a clear original emergency rescue scene image.

[0046] In the embodiment of the present invention, the image acquisition device is, for example, a handheld terminal, the local processing device is, for example, a local computing platform, the sending device is, for example, a Beidou terminal, the receiving device is, for example, a Beidou user terminal, and the cloud processing device is, for example, a cloud computing platform. Figure 3 As shown, the handheld terminal is used to collect the original emergency rescue scene image; the local computing power platform is used to pre-compress the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; the preliminary compressed image is deeply compressed to obtain a compressed image bit stream; the Beidou terminal is used to send the compressed image bit stream and the progressively compressed image to the cloud control platform in sequence; The Beidou user terminal, cloud computing platform and cloud platform can be integrated on a server, the Beidou user terminal is used to receive the compressed image bit stream and progressively compressed image sent by the computing platform; the cloud computing platform is used to restore the blurred original emergency rescue scene image according to the compressed image bit stream; the cloud platform is used to combine the blurred original emergency rescue scene image and the progressively compressed image to reconstruct a clear original emergency rescue scene image.

[0047] In an embodiment of the present invention, each device is equipped with a Beidou communication APP, and the user's handheld devices such as handheld terminals, tablets, mobile phones, etc. can realize functions such as image acquisition and user interaction, and the collected pictures are transmitted to the computing power platform via Bluetooth. The local computing power platform performs block averaging on the received image, divides the image into multiple small blocks (regions), and then uses the average value of the pixels in each block to represent the entire block, while reducing the number of pixels. The average information in the region is retained, and the HiFiC technology is used to deeply compress the pre-processed images. The compressed data is transmitted to the Beidou terminal via WIFI. The Beidou terminal determines whether to unpack the data, transmits the data to the Beidou user machine via a short message, and decompresses and assembles the data in the cloud. Due to the use of progressive image transmission, the cloud control platform can display a low-resolution version of the image at the beginning of the transmission. As more data is transmitted, the details and clarity of the image are gradually improved, and finally a high-quality image is displayed.

[0048] Based on any of the foregoing embodiments, the sending the compressed image bit stream and the progressively compressed image to the cloud control platform in sequence includes: Determine whether the data bytes of the compressed image bit stream are greater than a preset byte threshold, and if so, unpack the compressed image bit stream according to the preset byte threshold number, and send the data packet formed by the unpacking to the Beidou user terminal; Receive the response information sent by the Beidou user terminal, determine whether the packet is lost according to the response information, and if so, obtain the serial number of the lost data packet in the response information to resend the corresponding data packet according to the serial number of the lost data packet.

[0049] like Figure 4 As shown, the emergency rescue scene image transmission method provided by the embodiment of the present invention specifically includes: (1) The user connects the handheld device to the computing platform via Bluetooth, opens the Beidou communication APP on the handheld device, performs Beidou terminal monitoring configuration settings, and sets the message frequency, safety report frequency, and monitoring service center number; (2) Set the compression rate, click the capture button to capture the image, select the crop ratio (M*M), and transfer the image to the computing platform; (3) Computing platform compression preprocessing, the specific steps are as follows: a. Take an 8*8 average pooling window to compress the original image, and send the image to the HiFiC algorithm; b. Take a 2*2 average pooling window to compress the original image, and store the average of the first 2 data in each window, the first and the third data, and record them as A1, A2, A3; c. Take a 2*2 average pooling window to perform secondary compression on the image after pooling compression in step b, and store the average of the first 2 data in each window, the first and the third data, and record them as B1, B2, B3, and send B1, B2, B3, A1, A2, A3 in sequence; (4) The computing platform uses the HiFiC algorithm to achieve deep compression. The algorithm balances compression quality and image clarity by adjusting the weights of each loss function. The specific steps are as follows: a. Normalize the image; b. The encoder uses CNN to extract the main features of the image and convert the high-dimensional image data into a low-dimensional potential representation; c. Discretize the feature vector; d. Further compress it into a bit stream through entropy coding technology; (5) The computing platform establishes a connection with the Beidou terminal via WIFI, and uploads the compressed data to the Beidou terminal via WIFI; (6) The Beidou terminal determines whether to unpack the data. If the number of data bytes is less than N bytes (N is determined by the Beidou card communication level), there is no need to unpack the data and the data is sent directly to the Beidou user terminal. Otherwise, the data is unpacked according to the N bytes of a single frame message, and a flag bit and a header are added. The multi-card polling communication coordinates the data to send it to the Beidou user terminal. The Beidou user terminal performs an XOR and check on the received data. If the check is wrong, it stops receiving the information.

[0050] (7) After the data is sent, the Beidou terminal sends a query command to the Beidou user terminal and waits for a response message; the Beidou user terminal sends a response packet. If no response message is received within the specified time (3 minutes), the query command is sent to the Beidou user terminal again, waiting for a response message, and the number of times is increased by 1. If the number of times exceeds three times, the sending is recorded as a failure and the next communication is waited for; if a response message is received, the response message is queried, and the number and sequence number of the lost data packets are determined according to the content of the response packet message. If the data is complete, the communication is recorded as successful and other content is allowed to be sent. Otherwise, the sequence number of the missing data packet is sent to the sender, and the sender resends the missing data packet to complete the packet replenishment process; (8) The Beidou user terminal parses each packet and sends the complete data packet to the cloud computing platform once it is received; (9) In the cloud computing platform, a. perform entropy decoding on the compressed bit stream to restore the quantized discrete feature vector; b. perform dequantization to restore the latent features represented by floating point numbers; c. use the generator part of the generative adversarial network (GAN) to reconstruct the original image from the feature vector; d. perform post-processing and denormalization; (10) Image superposition gradient output on the cloud platform.

[0051] The emergency rescue scene image transmission method provided by the embodiment of the present invention is that the Beidou terminal unpacks and processes the Beidou short message transmission, the cloud receives the data, the background decompresses and displays it on the front end, and the efficient transmission of emergency scene images is achieved by exchanging computing power for bandwidth, providing real-time on-site multimedia data for rescue decisions, solving the emergency transmission problem of images without public networks and other communication networks, and making up for the low single communication bandwidth of Beidou short messages. In the image transmission process of emergency scenes, the image is averaged and pre-processed, and then HiFiC technology is applied for deep compression, which realizes fast and clear transmission of images. It has high compression rate, supports progressive image transmission, and the images obtained after the image recovery algorithm are clear enough to meet the needs of practical applications.

[0052] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0053] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. 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 make equivalent replacements for some of the technical features therein. However, 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 embodiments of the present invention.

Claims

1. A method for transmitting emergency rescue scene images, characterized in that: Applied to local terminals, including: Receiving original emergency rescue scene images; Pre-compressing the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; Performing deep compression on the preliminary compressed image to obtain a compressed image bit stream; The compressed image bit stream and the progressively compressed image are sent to a cloud control platform in sequence, and the compressed image bit stream and the progressively compressed image are used by the cloud control platform to reconstruct the original emergency rescue scene image.

2. The method for transmitting emergency rescue scene images according to claim 1, characterized in that: The pre-compressing the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image includes: Taking a first average pooling window to compress the original emergency rescue scene image to obtain the preliminary compressed image; Taking a second average pooling window to compress the original emergency rescue scene image to obtain a first-layer progressive compression image; Taking a third average pooling window to perform secondary compression on the first-layer progressively compressed image to obtain a second-layer progressively compressed image; The first average pooling window is larger than the second average pooling window and the third average pooling window.

3. The method for transmitting emergency rescue scene images according to claim 2, characterized in that: The taking of the second average pooling window to compress the original emergency rescue scene image to obtain a first-layer progressive compression image includes: Divide the original emergency rescue scene image into a plurality of small-block images, calculate a first average value of pixels in each of the small-block images, and use an image represented by the first average value and an image represented by any two vertex pixels in each of the small-block images as the first-layer progressive compression image; The method of taking the third average pooling window to perform secondary compression on the first-layer progressively compressed image to obtain the second-layer progressively compressed image includes: taking the third average pooling window to perform secondary compression on a plurality of the small-block images, dividing each of the small-block images into a plurality of sub-block images, calculating a second average value of pixels in each of the sub-block images, and taking an image represented by the second average value and an image represented by any two vertex pixels in each of the sub-block images as the second-layer progressively compressed image.

4. The method for transmitting emergency rescue scene images according to claim 3, characterized in that: The progressively compressed images are sequentially sent to the cloud control platform, including: The image represented by the second average value, the image represented by any two vertex pixels in each of the sub-block images, the image represented by the first average value, and the image represented by any two vertex pixels in each of the small-block images are sent to the cloud control platform in this order.

5. The method for transmitting emergency rescue scene images according to claim 1 or 2, characterized in that: The step of deeply compressing the preliminary compressed image to obtain a compressed image bit stream comprises: Performing normalization processing on the preliminary compressed image; Extracting feature data of the normalized image, quantizing the feature data, reducing the dimension of the feature data, and obtaining a feature potential representation vector; Discretizing the feature potential representation vector, inputting the discretized feature potential representation vector into a generative image compression network based on entropy coding compression, and controlling the compressed image output by the generative image compression network based on entropy coding compression based on a conditional generative adversarial network to obtain a compressed image bit stream; The entropy coding-based generative image compression network is obtained by jointly training using a neural network compression loss function and a discriminator loss function of the conditional generative adversarial network.

6. The method for transmitting emergency rescue scene images according to claim 5, characterized in that: The neural network compression loss function is constructed by the perceptual quality optimization term corresponding to the discriminator of the conditional generative adversarial network, the bit rate of the potential representation feature vector, the compressed image distortion rate, the control trade-off hyperparameter, and the joint optimization hyperparameter; The discriminator loss function of the conditional generative adversarial network is constructed by the perceptual quality optimization item corresponding to the discriminator of the conditional generative adversarial network.

7. The method for transmitting emergency rescue scene images according to claim 5, characterized in that: The discriminator loss function of the conditional generative adversarial network includes a target bit rate hyperparameter, which is used to adjust the average bit rate of the generative image compression network based on entropy coding compression.

8. A method for processing emergency rescue scene images, characterized in that: Applied to cloud control platform, including: receiving a compressed image bit stream and a progressively compressed image in sequence; Restoring the blurred original emergency rescue scene image according to the compressed image bit stream; The blurred original emergency rescue scene image and the progressively compressed image are combined to reconstruct a clear original emergency rescue scene image.

9. The emergency rescue scene image processing method according to claim 8, characterized in that: The method of restoring the blurred original emergency rescue scene image according to the compressed image bit stream includes: Performing entropy decoding on the compressed image bit stream to obtain a quantized discrete feature potential representation vector; Dequantizing the quantized discrete feature potential representation vector to restore it into feature data represented by floating point numbers; Reconstructing a decompressed image from the feature data represented by the floating point numbers; The decompressed image is subjected to inverse normalization processing to obtain a blurred original emergency rescue scene image.

10. The emergency rescue scene image processing method according to claim 8, characterized in that: The step of combining the blurred original emergency rescue scene image and the progressively compressed image to reconstruct a clear original emergency rescue scene image comprises: After the blurred original emergency rescue scene image is restored, the sequentially received progressively compressed images are superimposed on the blurred original emergency rescue scene image to reconstruct a clear original emergency rescue scene image.

11. An emergency communication system, characterized in that: include: Local terminal; The local terminal includes an image acquisition device, a local processing device and a sending device, and the image acquisition device collects original emergency rescue scene images; The local processing device is used to pre-compress the original emergency rescue scene image to obtain a preliminary compressed image and a progressively compressed image; Performing deep compression on the preliminary compressed image to obtain a compressed image bit stream; The sending device is used to send the compressed image bit stream and the progressively compressed image to the cloud control platform in sequence.

12. An emergency communication system, characterized in that: include: Cloud control platform; The cloud control platform includes a receiving device, a cloud processing device and a cloud platform; The receiving device is used for receiving a compressed image bit stream and a progressively compressed image; The cloud processing device is used to restore the blurred original emergency rescue scene image according to the compressed image bit stream; the cloud platform is used to reconstruct a clear original emergency rescue scene image by combining the blurred original emergency rescue scene image and the progressively compressed image.

Images