CN106899810A - A kind of mine video image fusion method and device - Google Patents

Abstract

The invention discloses a mine video image fusion method, which is used to realize the sampling, fusion and reconstruction of mine images. The method first compresses and samples mine video images in blocks, then performs source entropy fusion on the compressed and sampled signals of the image blocks, and finally reconstructs the fused signals and restores them to the original image after splicing. The invention further discloses a mine video image fusion device, which includes a mine image collection device, a mine image fusion node and a mine image receiving device. The device adopts a wireless transmission mode, has strong image compression processing and transmission capabilities, less redundant calculations for video image processing, requires less storage space, and has low bandwidth requirements for downhole transmission channels. The excellent compression processing and transmission capabilities ensure the real-time transmission of video signals under mine WSN and Zigbee wireless network environments.

Description

A mine video image fusion method and device

technical field

The invention relates to a wireless communication and image fusion technology, in particular to a mine video image fusion method and device.

Background technique

The existing mine image information fusion technology needs to consider all the pixels of the image, that is, it needs to decompress and restore the compressed and encoded image information and then fuse it. The use of this technology device has complex design, high transmission bandwidth requirements, and storage space requirements. Big and other issues. In particular, the video image in the mine environment is greatly affected by noise interference, and there are low resolution and blurred images in the mine video. The communication environment has limited bandwidth resources, and it is difficult to solve the problems of video signal transmission delay and image jitter during video image compression processing by using traditional image fusion technology. Therefore, it is urgent to research and invent a new mine video image fusion method and device to solve the existing problems.

In 2006, Donoho et al. proposed the Compressed Sensing (CS) theory. By analyzing the sparse matrix of the signal itself, it tried to break through the bandwidth bottleneck in traditional information theory, which greatly improved the signal processing and compression limits. The theory shows that if the signal is compressible or sparse in some transform domain, then the transformed high-dimensional signal can be projected onto a low-dimensional space with a random observation matrix uncorrelated with the transform basis, and then By solving a sparse optimization problem to achieve accurate reconstruction of the signal, the signal can be reconstructed with a sampling number much lower than that required by the sampling theorem. Moreover, existing studies have shown that using information entropy as a weighting coefficient to fuse information in the perceptual domain can completely reconstruct the original image, which can reduce the requirements for transmission performance, the complexity of hardware design, and the requirements for storage space.

Therefore, based on the above-mentioned CS theory, the research on mine video image fusion technology and method based on compressive sensing theory will provide a new idea for the compression processing of mine wireless video images under mine wireless network environment.

Contents of the invention

The invention mainly solves the problems existing in the prior art, and invents a mine video image fusion method and device based on block compression sensing. Apply the block compressed sensing and source entropy fusion algorithm to the compressed sampling and image fusion of mine video images, and use the block compressed sensing algorithm to reduce the sampling rate of the signal and the complexity of the random observation matrix, thereby reducing the video image processing Redundant calculations, saving storage space; the use of information source entropy fusion algorithm can achieve information fusion before image reconstruction, thereby reducing the requirements for transmission channel bandwidth.

The technical solution adopted in the present invention is: a mine video image fusion method and device, which adopts a source entropy fusion algorithm based on block compressed sensing to realize the sampling, fusion and reconstruction of mine video image signals. The mine video image fusion method comprises the steps of:

Step 1, the sampling of the signal comprises the following sub-steps:

1.1) Divide the input mine video image signal into S image blocks of size K×K pixels, where S, K∈N;

1.2) Construct the observation matrix corresponding to the sth (s=1, 2, ... S) image block as Φ _K , where, m _s <K ² , and Φ _K is a Hadamard matrix;

1.3) Obtaining the different observation vectors of the sth image block are respectively x ₁ =Φ _K Θ ₁ and x ₂ =Φ _K Θ ₂ , where And Θ ₁ , Θ ₂ are the column vectors of the sth block of the source image respectively;

Step 2, fusion of signals, including the following sub-steps:

2.1) Calculate the source entropy entropy(x ₁ ) and entropy(x ₂ ), the joint entropy entropy(x ₁ , x ₂ ), and the mutual information MI(x ₁ , x ₂ ) of the observation vectors x ₁ and x ₂ ,

2.2) Calculate the weight coefficients of the observation vectors x ₁ and x ₂ ,

2.3) Fusing different observation vectors into the final observation value x _b =t ₁ x ₁ +t ₂ x ₂ , where t ₁ and t ₂ are normalized to satisfy t ₁ +t ₂ =1;

Step 3, signal reconstruction, according to the perception matrix corresponding to the image block signal, using the orthogonal matching pursuit algorithm to reconstruct the reconstruction value corresponding to the image block signal;

Step 4, signal integration, according to the method of dividing the source image blocks, the reconstruction values corresponding to each block are spliced and reorganized to obtain a reconstructed image.

The present invention further discloses a mine video image fusion device based on compressed sensing, which is applied to the mine video image fusion method, and the device includes: a mine image collection device, a mine image fusion node and a mine image receiving device; wherein, the The mine image collection equipment is used for mine image block collection and image compression encoding; the mine image fusion node is used for image fusion in the perceptual domain; the mine image receiving device is used for mine image reconstruction decoding, video processing and image processing Display; the mine image acquisition equipment, mine image fusion node and mine image receiving device are connected through a wireless communication interface, and are used to fuse the collected mine image signal and transmit it to the mine image receiving device; the wireless communication interface adopts Zigbee, WiFi Or/and WCDMA, WiMAX or/and LTE air interface.

The mine image acquisition device includes: a video acquisition unit and an image encoding unit; the video acquisition unit is used for block acquisition of continuous images, including an image block module and an image acquisition module;

The image block module is used for mine image block processing.

The image collection module is used to collect continuous block images and transmit the collected block image information to an image coding unit for compression coding.

The image coding unit is used for compressing and coding the collected block image information, including a matrix module, a storage module and a multiplication module.

The matrix module is used to generate a compressed observation matrix.

The storage module is used to store the compressed observation matrix generated by the matrix module.

The multiplication module is used to multiply the block image signal collected by the video acquisition unit with the compressed observation matrix stored in the storage module, so as to obtain the block image code after compression encoding.

The mine image fusion node includes a source entropy calculation unit and a weighted fusion processing unit; the source entropy calculation unit is used for the calculation of related source entropy, including a source entropy module, a joint entropy module, a mutual information module and storage module;

The source entropy module, the joint entropy module and the mutual information module are respectively used to calculate the source entropy, the joint entropy and the mutual information of the block-collected signals.

The storage module is used to store the source entropy, joint entropy and mutual information of the block-collected signals.

The weighted fusion processing unit is used for weighted fusion processing of sampled signals, including a weight calculation module and a fusion module.

The weight calculation module is used to calculate the weight coefficient of the image information collected by blocks.

The fusion module is used for weighted fusion processing of image information collected in blocks.

The mine image receiving device includes an image reconstruction unit and a video synthesis unit; the image reconstruction unit is used to decode and restore the image after block compression and fusion, including a matrix module, a storage module, a multiplication module, a correction module and Block synthesis module.

The matrix module is used to generate a sparse expression matrix;

The storage module is used to store the sparse expression matrix generated by the matrix module;

The multiplication module is used to calculate the orthogonal base matrix of the image code and the matrix sparse expression matrix after block compression coding;

The correction module is used to calculate the noise residual of the image after the orthogonal basis matrix and block compression encoding, and iteratively calculate and correct to obtain the restored image after sparse decoding;

The block synthesis module is used to stitch and reorganize the block image information into the original image;

The video synthesis unit is used for merging the sparsely coded restored images into a continuous video.

The mine video image fusion device includes communication equipment such as a wireless camera, a vehicle-mounted mobile station, and a hand-held mobile station.

The mine image acquisition equipment, mine image fusion node and mine image receiving equipment are intrinsically safe explosion-proof devices.

The beneficial effects of the present invention are:

The present invention adopts block compression sampling, and the required random observation matrix is simple, which effectively reduces the redundant calculation and storage requirements of image encoding and compression processing; Data fusion, thereby reducing the requirements of mine video images on the transmission channel.

Description of drawings

Fig. 1 is a schematic diagram of a mine video image fusion method;

Fig. 2 is a kind of mine video image fusion processing schematic diagram based on block compressed sensing;

Fig. 3 is a schematic diagram of a mine video image fusion device;

detailed description

Specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

Referring to Figure 1, it is a schematic diagram of a mine video image fusion method. The mine video image signal sampling and fusion process is as follows: first, the original mine video image is divided into blocks, and then the block mine images are randomly observed, and then the observed The signal is the block-compressed signal of the mine image for source entropy fusion, and finally the fused signal is reconstructed and spliced to obtain the original mine image. Apply the block compressed sensing and source entropy fusion algorithm to the compressed sampling and image fusion of mine video images, and use the block compressed sensing algorithm to reduce the sampling rate of the signal and the complexity of the random observation matrix, thereby reducing the video image processing Redundant calculations save storage space; the information source entropy fusion algorithm is used to fuse information before image reconstruction, aiming to reduce transmission channel bandwidth requirements.

Referring to FIG. 2 , it is a schematic diagram of mine video image fusion processing based on block compression sensing. The CS observation of mine video images is a linear process. In order to ensure accurate reconstruction, the necessary and sufficient condition for the existence of solutions to the linear equations is that the observation matrix and the sparse transformation matrix satisfy the finite equidistant condition. Considering that the Hadamard matrix is irrelevant to the matrix composed of most fixed orthogonal bases, this characteristic determines that it is selected as the observation matrix of the block image signal, and the sparse base matrix corresponding to the block image signal is set to Ψ, so it is necessary to construct the first The observation matrix corresponding to s (s=1, 2, ... S) image blocks is Φ _K .

The mine video image fusion process mainly includes the steps of signal sampling, signal fusion, signal reconstruction and signal integration. The specific implementation steps are as follows:

Step 1, the sampling of the signal comprises the following sub-steps:

1.1) Divide the input mine video image signal into S image blocks of size K×K pixels, where S, K∈N;

1.2) Construct the observation matrix corresponding to the sth (s=1, 2, ... S) image block as Φ _K , where, m _s <K ² , and Φ _K is a Hadamard matrix;

1.3) Obtaining the different observation vectors of the sth image block are respectively x ₁ =Φ _K Θ ₁ and x ₂ =Φ _K Θ ₂ , where And Θ ₁ , Θ ₂ are the column vectors of the sth block of the source image respectively;

Step 2, fusion of signals, including the following sub-steps:

2.1) Calculate the source entropy entropy(x ₁ ) and entropy(x ₂ ), the joint entropy entropy(x ₁ , x ₂ ), and the mutual information MI(x ₁ , x ₂ ) of the observation vectors x ₁ and x ₂ ,

2.2) Calculate the weight coefficients of the observation vectors x ₁ and x ₂ ,

2.3) Fusing different observation vectors into the final observation value x _b =t ₁ x ₁ +t ₂ x ₂ , where t ₁ and t ₂ are normalized, that is, t ₁ +t ₂ =1;

In the calculation of the above formula, among them, the weight coefficients t ₁ of the observation vectors x ₁ and x ₂ , the first item in t ₂ respectively describe the proportion of the source entropy entropy(x ₁ ) and entropy(x ₂ ) in the joint entropy, Contains the common part of two items of information to be fused, which needs to be corrected by subtracting the second item; the weight coefficient t ₁ of the observation vector x ₁ and x ₂ , the MI mutual information in the second item of t ₂ reflects the two items The information contained in the information to be fused is finally normalized by t ₁ and t ₂ , that is, t ₁ +t ₂ =1;

2.3) Fusing different observation vectors into the final observation value x _b =t ₁ x ₁ +t ₂ x ₂ .

Step 3. Reconstruction of the signal. According to the perception matrix corresponding to the image block signal, an orthogonal matching reconstruction algorithm is used to reconstruct the reconstruction value corresponding to the block signal. The known parameters of the algorithm are: fusion measurement value x _b , observation matrix (Hadamard Matrix) Φ _K , sparse base matrix Ψ, sparsity is k, noise residual e, subject to N(0, σ ² ), the specific signal reconstruction process is as follows:

3.1) Initialization parameters: estimated signal Sensing matrix A = Φ _K Ψ, index set of selected column vectors Iteration times t=1, noise residual e ₀ =x _b , residual threshold Θ _threshold =||e _t -e _t-1 || ₂ ;

3.2) Calculate λ _t =arg _j max|<e _t-1 , a _j >|, that is, the column vector index λ with the largest correlation coefficient between the column vector a _j of the perception matrix A and the noise residual e;

3.3) Update the index set Γ _t ＝ Γ _t-1 ∪{λ _t }, and record the selected column vector at the same time

3.4) Seek sparse signal estimation And make

3.5) Update residuals t=t+1;

3.6) Judging whether the iteration stop condition is satisfied. When t>k, or if there is ||e _t -e _t-1 || ₂ <ξ, when ||e _t || ² <Θ _threshold , the iteration ends, otherwise return to step 3.2) to continue iterating and solve the optimal solution until the condition is met.

Step 4, signal integration, the embodiment in Figure 2 illustrates that when the reconstruction of the s image block signals is completed, according to the method for dividing the original image block, the reconstruction values corresponding to each block are spliced and reorganized to obtain the reconstructed composition image.

With reference to Fig. 3, it is a kind of mine video image fusion device schematic diagram, and this device mainly comprises: mine image acquisition equipment (10), mine image fusion node (20) and mine image receiving equipment (30); Wherein, mine image acquisition equipment ( 10), used for image block acquisition, image compression encoding; the mine image fusion node (20) is used for image fusion in the perceptual domain; the mine image receiving device (30) is used for image reconstruction decoding, video processing and image display.

A kind of mine video image fusion device shown in Fig. 3, mine image acquisition equipment (10) comprises: video acquisition unit (101), image coding unit (102); Wherein, video acquisition unit (101) is used for the analysis of continuous image Block acquisition, including image block module (101A) and image acquisition module (101B); image block module (101A) is used for mine image block processing; image acquisition module (101B) is used for collecting continuous block images, and The collected block image information is transmitted to an image encoding unit (102) for compression encoding. Image encoding unit (102), used for compressing and encoding the collected block image information, including matrix module (102A), storage module (102B) and multiplication module (102C); matrix module (102A) is used to generate compressed observation matrix; the storage module (102B) is used to store the compressed observation matrix generated by the matrix module; the multiplication module (102C) is used to compare the compressed observation matrix stored by the video acquisition unit (101) with the block image collected by the storage module (102B) Multiply to obtain the block image code after compression coding.

A kind of mine video image fusion device shown in Fig. 3, mine image fusion node (20) comprises information source entropy calculation unit (201) and weighted fusion processing unit (202); Wherein, information source entropy calculation unit (201) comprises information Source entropy module (201A), joint entropy module (201B), mutual information module (201C) and storage module (201D); source entropy module (201A), joint entropy module (201B) and mutual information module (201C) They are respectively used to calculate the source entropy, joint entropy and mutual information of the block-collected signals; the storage module (201D) is used to store the source entropy, joint entropy and mutual information of the block-collected signals. The weighted fusion processing unit (202) includes a weight calculation module (202A) and a fusion module (202B); the weight calculation module (202A) calculates the weight coefficient through the information entropy of the block image; the fusion module (202B) uses the block image information according to the weight The calculation module calculates the obtained weight coefficients to perform weighted fusion processing.

A mine video image fusion device shown in Fig. 3, a mine image receiving device (30) includes an image reconstruction unit (301) and a video synthesis unit (302); wherein, the image reconstruction unit (301) is used for dividing The image after the block compression fusion is decoded and restored, including a matrix module (301A), a storage module (301B), a multiplication module (301C), a correction module (301D) and a block synthesis module (301E); wherein, the matrix module (301A) uses To generate a sparse expression matrix; the storage module (301B) is used to store the sparse expression matrix generated by the matrix module; the multiplication module (301C) is used to calculate the orthogonal base matrix of the image coding and matrix sparse expression matrix after block compression coding; correction The module (301D) is used to calculate the noise residual of the image after the orthogonal base matrix and block compression encoding, and iteratively calculates and corrects to obtain the restored image after sparse decoding; the block synthesis module (301E) is used for splicing block image information , Reshape to the original image. The video synthesis unit (302) is used for merging the restored images after sparse coding into continuous video.

The present invention applies the algorithm of block compressed sensing and information source entropy fusion to the compressed sampling and image fusion of mine video images, adopts the algorithm of block compressed sensing, reduces the sampling rate of the signal and the complexity of the random observation matrix, thereby reducing the The redundant calculation of image processing saves storage space; the information source entropy fusion algorithm is used to fuse information before image reconstruction, aiming to reduce the transmission channel bandwidth requirements. Combined with the mine video image fusion device provided by the method of the present invention, it can meet the specific use environment and safety requirements of the coal mine underground.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments. It cannot be determined that the specific embodiments of the present invention are limited thereto. Under the premise of the idea, some simple substitutions and changes can also be made, which should be regarded as belonging to the scope of protection involved in the claims submitted by the present invention.

Claims (4)

1. A mine video image fusion method, used to realize sampling, fusion and reconstruction of mine video image signal, is characterized in that, comprises the following steps: Step 1, the sampling of the signal comprises the following sub-steps: 1.1) Divide the input mine video image signal into S image blocks of size K×K pixels, where S, K∈N; 1.2) Construct the observation matrix corresponding to the sth (s=1, 2, ... S) image block as Φ _K , where, m _s <K ² , and Φ _K is a Hadamard matrix; 1.3) Obtaining the different observation vectors of the sth image block are respectively x ₁ =Φ _K Θ ₁ and x ₂ =Φ _K Θ ₂ , where And Θ ₁ , Θ ₂ are the column vectors of the sth block of the source image respectively; Step 2, fusion of signals, including the following sub-steps: 2.1) Calculate the source entropy entropy(x ₁ ) and entropy(x ₂ ), the joint entropy entropy(x ₁ , x ₂ ), and the mutual information MI(x ₁ , x ₂ ) of the observation vectors x ₁ and x ₂ , $\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; g</span>}\frac{\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ;</span>$ 2.2) Calculate the weight coefficients of the observation vectors x ₁ and x ₂ , ${\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&CenterDot;</span>\frac{\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}$ ${\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&CenterDot;</span>\frac{\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}$ 2.3) Fusing different observation vectors into the final observation value x _b =t ₁ x ₁ +t ₂ x ₂ , where t ₁ and t ₂ are normalized to satisfy t ₁ +t ₂ =1; Step 3, signal reconstruction, according to the perception matrix corresponding to the image block signal, using the orthogonal matching pursuit algorithm to reconstruct the reconstruction value corresponding to the image block signal; Step 4, signal integration, according to the method of dividing the source image blocks, the reconstruction values corresponding to each block are spliced and reorganized to obtain a reconstructed image. 2. A mine video image fusion device based on compressed sensing, applied to the mine video image fusion method, is characterized in that, comprising: mine image acquisition equipment (10), mine image fusion node (20) and mine image receiving equipment (30); wherein, the mine image acquisition device (10) is used for mine image block collection and image compression encoding; the mine image fusion node (20) is used for image fusion on the perception domain; the mine image The receiving device (30) is used for image reconstruction decoding, video processing and image display; the mine image acquisition device (10), the mine image fusion node (20) and the mine image receiving device (30) are connected through a wireless communication interface. After merging the collected mine image signal, it is transmitted to the mine image receiving device (30); it is also characterized in that, The wireless communication interface adopts Zigbee, WiFi or/and WCDMA, WiMAX or/and LTE air interface; and The mine image acquisition device (10) includes: a video acquisition unit (101), an image encoding unit (102); the video acquisition unit (101) is used for block acquisition of continuous images, including an image block module (101A) And image acquisition module (101B); Described image block module (101A) is used for mine image block processing; Described image acquisition module (101B) is used for collecting continuous block image and the block image information that gathers Transmit to image encoding unit (102) and carry out compression encoding; Described image encoding unit (102), is used for compressing and encoding the block image information collected, including matrix module (102A), storage module (102B) and multiplication module (102C); the matrix module (102A) is used to generate a compressed observation matrix; the storage module (102B) is used to store the compressed observation matrix generated by the matrix module; the multiplication module (102C) is used to use the video acquisition unit ( 101) Multiply the collected block image with the compressed observation matrix stored in the storage module (102B) to obtain the block image code after compression encoding; The mine image fusion node (20) includes a source entropy calculation unit (201), a weighted fusion processing unit (202); the source entropy calculation unit (201) includes a source entropy module (201A), a joint entropy module ( 201B), a mutual information module (201C) and a storage module (201D); the source entropy module (201A), the joint entropy module (201B) and the mutual information module (201C) are respectively used to calculate each block acquisition signal source entropy, joint entropy and mutual information; the storage module (201D) is used to store the source entropy, joint entropy and mutual information of the block acquisition signal; the weighted fusion processing unit (202) includes weight calculation module (202A) and fusion module (202B); the weight calculation module (202A) calculates the weight coefficient by the information entropy of the block image; the fusion module (202B) calculates the weight coefficient of the block image information according to the weight calculation module Perform weighted fusion processing; The mine image receiving device (30) includes an image reconstruction unit (301) and a video synthesis unit (302); the image reconstruction unit (301) is used to decode and restore the image after block compression and fusion, including A matrix module (301A), a storage module (301B), a multiplication module (301C), a correction module (301D) and a block synthesis module (301E); wherein, the matrix module (301A) is used to generate a sparse expression matrix; the storage The module (301B) is used to store the sparse expression matrix generated by the matrix module; the multiplication module (301C) is used to calculate the orthogonal base matrix of the image coding and matrix sparse expression matrix after block compression coding; the correction module (301D ) is used to calculate the noise residual of the image after the orthogonal basis matrix and block compression encoding, and iterative calculation and correction to obtain the restored image after sparse decoding; the block synthesis module (301E) is used to stitch the block image information , reforming into an original image; the video synthesis unit (302) is used for merging the restored image after sparse coding into a continuous video. 3. The mine video image fusion device according to claim 2 includes communication equipment such as a wireless camera, a vehicle-mounted mobile station, and a handheld mobile station. 4. mine image acquisition equipment, mine image fusion node and mine image receiving equipment according to claim 2, characterized in that said mine image acquisition equipment, mine image fusion node and mine image receiving equipment are intrinsically safe explosion-proof devices .