CN107015193B - A binocular CCD visual mine moving target positioning method and system - Google Patents

Abstract

The invention discloses a binocular CCD visual mine moving target positioning method and system. The system for realizing the method includes above-hole equipment and down-hole equipment. The above-hole equipment includes a base station controller, a positioning server, an Ethernet switch and a monitoring terminal. The down-hole equipment includes Mine intrinsically safe base stations, radio frequency identification tags. The method steps for implementing the system are: (1) using the RSSI algorithm to estimate the position of the moving target; (2) performing matching feature extraction training on the radio frequency identification tags of the moving target; (3) performing ORB algorithm feature matching on the sampled image (4) Use the binocular CCD visual sensor to carry out stereo calibration to the left and right images of the moving target; (5) Obtain the internal and external parameter matrix of the binocular CCD visual sensor, and calibrate the world coordinates of the moving target; (6) The world coordinate information of the moving target Perform position correction to obtain accurate position information of moving targets in the mine. The invention solves the problem of precise positioning of the moving target in the mine NLOS environment, and improves the reliability and robustness of the system.

Description

A binocular CCD visual mine moving target positioning method and system

technical field

The invention relates to wireless communication technology and computer vision technology, in particular to a binocular CCD vision mine moving target positioning method and system.

Background technique

Coal is an important basic energy in our country, and the energy structure dominated by coal will not change for a long time. With the development of the national economy and the increasing demand for coal, the accompanying coal mine safety production accidents are also increasing . Therefore, the requirements for mine safety are getting higher and higher, which also puts forward higher requirements for the accuracy and reliability of the positioning of moving targets such as mine operators and operating equipment.

At present, some positioning technologies based on RFID, WIFI, and Zigbee, as well as positioning algorithms such as triangle centroid positioning, TOA, and AOA, have been applied in underground personnel positioning. Existing problems: On the one hand, due to the non-direct line-of-sight propagation and multipath interference of electromagnetic waves in the underground NLOS environment, the real-time performance and accuracy of positioning technologies such as WIFI and Zigbee are easily affected, and due to the large transmission loss of wireless electromagnetic waves in mine roadways and serious signal attenuation, There are large errors in positioning algorithms such as triangle centroid positioning, TOA, and AOA, and it is impossible to achieve accurate positioning of mine moving targets. On the other hand, RFID positioning technology can only identify the entry and exit of moving targets underground, and cannot realize the two-dimensional positioning of moving targets. It is also difficult to accurately locate the problem of inaccurate inspection or identification.

CCD visual positioning technology is widely used in industrial non-contact measurement. It has the characteristics of high positioning accuracy, strong anti-interference ability, and long-distance acquisition of target images. Moreover, with the continuous promotion of CCD visual sensor and its measurement technology in industrial application results , and its application in underground coal mine mobile target positioning has also received more and more attention.

Contents of the invention

The technical problem to be solved by the present invention is to propose a binocular CCD visual mine moving target positioning method and system to realize real-time and precise positioning of the mine moving target in view of the above existing problems.

The technical solution of the present invention is: to propose a binocular CCD visual mine moving target positioning method and system, the system for realizing the method includes the above-ground equipment and the down-hole device, and the above-ground equipment includes a base station controller, an Ethernet switch, a positioning server and a monitoring terminal , Underground devices include mine intrinsically safe base stations and radio frequency identification tags. Uphole equipment communicates with the downhole device via an optical link.

The mine intrinsically safe base station has wireless interfaces such as LTE, WIFI, and UWB, and is used for radio frequency identification of underground moving targets;

The mine intrinsically safe base station has a built-in binocular CCD vision sensor, which is used to collect the left and right stereo image information of the moving target in the mine, and perform ORB feature matching and stereo calibration on it;

The intrinsically safe base station of the mine sends the position information calibrated by the binocular CCD visual sensor to the positioning server on the mine through the LTE wireless network and the optical link;

The radio frequency identification tag is installed or fixed on the underground mobile target as an identification mark, and is used for RSSI radio frequency identification, feature extraction and matching of the mobile target;

The binocular CCD visual mine moving target positioning method and system, its implementation steps include:

Step 1. The mine intrinsically safe base station estimates the position of the moving target by using the RSSI algorithm by receiving the radio frequency identification tag signal installed or fixed on the moving target.

Step 2. Define the judgment threshold Thr of the binocular CCD visual recognition as the farthest sampling distance of the image information of the binocular CCD visual sensor.

Step 3. When the mine intrinsically safe base station detects that the distance between the moving target and the base station is less than or equal to the decision threshold, that is, d _Rssi ≤ Thr, the binocular CCD vision sensor performs image information sampling and distance calculation on the moving target, otherwise, return to step 1 Re-estimate the position of the moving target.

Step 4: The binocular CCD vision sensor uses the ORB algorithm to perform feature extraction training on the radio frequency identification tag, and performs matching according to the extracted features and the characteristics of the moving target identification signs collected by the binocular CCD vision sensor.

Step 5. When the feature matching of the moving target is successful, the binocular CCD vision sensor performs left and right stereo calibration on the moving target image, obtains the internal and external parameter matrix and the mean value of the parallax of the binocular CCD vision sensor, and calculates the moving target by fusion of the binocular CCD vision sensor world coordinate information, otherwise return to step 1 to re-estimate the position of the moving target.

Step 6: Correct the world coordinate information of the moving target according to the location of the mine intrinsically safe base station, and obtain the final position coordinate information of the moving target.

The mine intrinsically safe base station uses the RSSI algorithm to monitor the received signal strength of the underground mobile target and locate the underground mobile target. According to the propagation loss theoretical model Calculate the sampling and analysis distance Thr of the image information of the moving target by the binocular CCD vision sensor, where P _r (d _Rssi ) is the signal strength when the signal transmission distance in the downhole environment is d _Rssi , and P _r (d ₀ ) is the ideal space state The signal strength when the signal transmission distance is d ₀ , k is the signal strength attenuation coefficient, i=1, . . . , N _Loss , which is the attenuation coefficient of the downhole signal strength of the measured N _Loss times.

The binocular CCD visual mine moving target positioning method and system, the step of realizing the ORB feature matching of the underground moving target further includes:

Step 1. Use the ORB algorithm to detect the feature points, and use the FAST operator to detect the feature points of the downhole moving target.

Step 2. Add direction information to the detected feature points of the downhole moving target to form oFAST.

Step 3: Add direction information to the detected corner points of the downhole moving target by using the gray scale centroid method. ORB adopts the gray-scale centroid method to add direction information to the detected corner points. Rosin defines neighborhood moments: m _pq = ∑ _{x, y} x ^p y ^q I(x, y), and the centroid is: The angle between the feature point and the centroid is defined as the direction of the FAST feature point: θ=atan2(m ₀₁ , m ₁₀ ).

Step 4, use the BRIEF descriptor to describe the feature points of the detected downhole moving target.

Step 5. Add rotation invariance to the detected feature point descriptor of the downhole moving target to form SteerBRIEF, which not only utilizes the simple and fast calculation advantages of BRIEF, but also solves the fact that BRIEF itself does not have rotation invariance. Define the S×S size image, and the algorithm extracts the feature point descriptor where p(x) is the gray value of the image neighborhood P at x=(u,v) ^T after smoothing. For n position pairs of ( _xi , y _i ), Steer BRIEF is at ( _xi , y _i ), for any feature set of n position pairs Use the rotation matrix R _θ to rotate the matching point, and get the feature set S _θ = R _θ S with direction.

Step 6. Calculate the descriptor Hamming distance Ham of two matching point pairs, and perform feature matching judgment. The rotated binary criterion descriptor g _n (p, θ): = f _n (p)|( _xi , y _i )∈S _θ , only need to calculate the Hamming distance Ham of the two feature point descriptors when matching, Make a feature matching decision.

Step 7. Search for the set n least correlated pixel block pairs from all possible pixel block pairs by using a greedy algorithm, retrieve and match the features of the moving target, and determine the feature matching result.

The binocular CCD vision mine moving target positioning method and system, the step of carrying out stereo calibration to the underground moving target further includes:

After the binocular CCD vision sensor successfully matches the underground moving target through the ORB algorithm feature, the binocular CCD vision sensor is used to locate the collected moving target, and the image pixel coordinate system Among them, (X _W , Y _W , Z _W ) are the world coordinates, u, v are the coordinates of the image pixel coordinate system, the x, y axes of the image physical coordinate system are parallel to the u, v axes of the pixel coordinate system, d _x , d _y Respectively, the pixel spacing in the horizontal and vertical directions, Z _C is the optical axis of the binocular CCD vision sensor, and f is the focal length;

Step 1. The binocular CCD vision sensor performs left and right stereo calibration on the collected image of the moving target, and obtains the internal and external parameter matrix of the binocular CCD vision sensor. That is, the external parameter matrix of the binocular CCD vision sensor is estimated by calibration of the camera coordinate system and the world coordinate system Where R is a 3×3 rotation matrix, 0 ^T is a 1×3 translation matrix, and the internal reference matrix of the binocular CCD vision sensor is solved by the internal parameters of the binocular CCD vision sensor and the left and right stereo calibration of the moving target image Where f _x , f _y are effective focal lengths.

Step 2. Obtain the mean value of parallax of the binocular CCD visual sensor, calculate the parallax d _i according to the pair of matching points obtained by stereo matching, i=1, 2, ..., N _Ver , and calculate the mean value of the parallax d _i

Step 3. Find the world coordinates of the moving target through the image pixel coordinate system equation, and solve the coordinates (X _W , Y _W , Z _W ) of the moving target in the world coordinate system according to the internal and external parameter matrix and the binocular CCD vision sensor parallax d _Ver , Among them, Z _W is the vertical distance from the spatial positioning moving target to the binocular CCD vision sensor, which is the depth information.

Step 4. Correct the world coordinate system coordinates of the moving target through the coordinates of the base station. The mine intrinsically safe base station with a built-in binocular CCD vision sensor has the coordinates (0, 0, 0) in the world coordinate system, and locates the coordinates in the world coordinate system of the moving target. (X _W , Y _W , Z _W ), and combined with the actual position of the mine roadway where the base station is located, the world coordinate information of the moving target is corrected to obtain the final position information of the moving target in the mine, and the precise positioning of the moving target is realized.

The beneficial effects of the present invention are:

Based on the RSSI and ORB algorithm, the invention provides a binocular CCD vision mine moving target positioning method and system through the fusion of RSSI radio frequency identification and binocular CCD vision calibration of underground moving targets. The invention solves the problems of missed detection, inaccurate identification and low positioning accuracy in the existing positioning technology, especially overcomes the problem that the moving target cannot be identified and accurately positioned due to target occlusion in the underground NLOS environment, and improves the positioning of the underground moving target Real-time performance, accuracy and system robustness. The system has high positioning accuracy and strong anti-interference ability, and is suitable for real-time tracking and precise positioning of moving targets in various complex environments and limited spaces.

Description of drawings

Figure 1 is a schematic diagram of the system composition of a binocular CCD vision mine moving target positioning method and system.

Fig. 2 is a positioning flowchart of a binocular CCD vision mine moving target positioning method and system.

Fig. 3 is a kind of binocular CCD visual mine moving target location method and the ORB algorithm flow chart adopted by the system.

FIG. 4 is a flow chart of a binocular CCD visual mine moving target positioning method and a binocular CCD positioning algorithm adopted by the system.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the specific implementation manners of the present invention will be described in detail below in conjunction with the accompanying drawings.

Figure 1 shows the system composition of binocular CCD mine moving target positioning method and system. With reference to shown in Figure 1, the system that realizes above-mentioned mine positioning method mainly comprises: radio frequency identification tag (101), built-in binocular CCD intrinsically safe positioning base station (102), Ethernet switchboard (103), base station controller (104) , a positioning server (105), a monitoring terminal (106).

Among them, the radio frequency signal emitted by the radio frequency identification tag (101) can be recognized by the intrinsically safe positioning base station for rough positioning, and the identification and extraction training of feature matching is performed on the tag, which is used for feature matching of binocular CCD visual sensor ORB algorithm. Binocular CCD intrinsically safe positioning base station (102), the intrinsically safe positioning base station is based on the theoretical model of propagation loss, uses the RSSI algorithm to identify the radio frequency identification tag on the moving target and obtains the radio frequency information of the moving target in the mine, built in the intrinsically safe positioning After being triggered, the binocular CCD vision sensor of the base station collects the image of the moving target and performs feature matching and stereo calibration to achieve precise positioning of the moving target. The Ethernet switch (103) is used as a convergence node of the positioning system network to realize the interconnection between the positioning system and the ground communication network. The base station controller (104) is used for underground wireless network management, wireless resource management, management, control and monitoring of mine intrinsically safe base stations, and handover control of underground mobile targets. The positioning server (105) is used to process and store the real-time location information of the moving target in the mine, provide mobile positioning service for the moving target in the mine, realize the positioning and monitoring function of the moving target in the mine, and provide the historical information query and retrieval service of the monitoring terminal . The monitoring terminal (106) is used for receiving images and real-time positioning information from the positioning server, and visually displaying the images and real-time positioning information.

Fig. 2 is a positioning flow chart of the binocular CCD mine moving target positioning method and system. With reference to shown in Fig. 2, the positioning process steps and function description of the binocular CCD mine moving target positioning method are as follows: (1) signal attenuation coefficient measurement and calibration (201), the mine intrinsically safe base station with built-in CCD vision sensor adopts logarithmic- Normal Propagation Loss Theoretical Model Carry out signal attenuation coefficient measurement and calibration. Since the attenuation coefficient of the mine signal is significantly higher than that of the mine environment, the actual measurement N _Loss times in the mine is used to calculate the average value As the mine signal attenuation coefficient; (2) radio frequency identification tag feature recognition extraction training (202), the identification and extraction training of matching features are carried out for the radio frequency tag identification installed in the moving target of the mine, and the matching features are provided for the ORB algorithm matching of the target (3) RSSI algorithm roughly locates and obtains the distance (203) between the mobile target and the base station, utilizes the RSSI algorithm of the base station to identify the radio frequency identification tag on the mobile target, and adopts the logarithmic-normal propagation loss theoretical model to roughly estimate the mine mobile target (4) Judging whether the distance between the moving target and the base station is less than or equal to the binocular CCD sampling distance Thr (203), by judging whether the distance between the moving target and the base station is less than or equal to the binocular CCD sampling distance Thr, triggering the binocular CCD collects moving target images, otherwise return to take RSSI to re-estimate the position of moving targets; (5) trigger binocular CCD to moving target image acquisition (205), when base station collects moving target distance base station distance d _Rssi is less than or equal to threshold binocular The CCD vision sensor can perform image information sampling and analysis distance Thr, which can trigger the mine intrinsically safe base station with built-in binocular CCD vision sensor, and carry out image information sampling of moving targets through the binocular CCD vision sensor; (6) ORB algorithm for moving target characteristics Whether the matching is successful (206), when the mine intrinsically safe base station monitors that the distance between the mobile target and the base station is less than or equal to the decision threshold, that is, when d _Rssi ≤ Thr, the binocular CCD vision sensor performs image information sampling and distance calculation on the mobile target, Otherwise, return and adopt RSSI to re-estimate the position of the moving target; (7) move target coordinate calibration (207), obtain the internal and external parameter matrix and parallax mean value d _Ver required for coordinate calibration according to the coordinate formula of the image pixel coordinate system, and calibrate the world coordinates of the moving target (8) Base station calibration and positioning is completed (208), the mine intrinsically safe base station with built-in binocular CCD vision sensor coordinates (0,0,0) in the world coordinate system, coordinates (0,0,0) in the world coordinate system of the positioning moving target ( X _W , Y _W , Z _W ), and combined with the actual position of the mine roadway where the base station is located, correct the coordinate position of the world coordinate information of the mine moving target, obtain the final position information of the mine moving target, and realize the precise positioning of the moving target.

Figure 3 is a flow chart of the binocular CCD mine moving target positioning method and the ORB algorithm adopted by the system. As shown in Figure 3, the ORB algorithm process steps that binocular CCD mine moving target localization method takes comprises: (1) input moving target image information (301), the moving target image that the binocular CCD visual sensor gathers is used as ORB algorithm The input of feature matching; (2) use the oFAST operator to perform feature point detection (302), take the FAST operator to add rotation invariance to form the oFAST operator, use the oFAST operator to perform feature point detection, and use the gray scale centroid method as The detected feature points increase the matching local invariance; (3) use Steer BRIEF to describe the feature points (303), the Steer BRIEF algorithm utilizes the simple and fast advantages of the calculation of BRIEF to carry out feature descriptions to the feature points detected by the oFAST algorithm, And it solves the shortcoming that BRIEF itself does not have rotation invariance; (4) Construct a feature set and a binary criterion descriptor (304), define an S×S size image, and select any n position feature pairs (x _i , y _i ) Constitute the feature set And in the S×S size image, the construction algorithm extracts the feature point descriptor Steer BRIEF at ( _xi , y ₎ , for any feature set of n position pairs Using the rotation matrix R _θ to rotate the matching point, the feature set S _θ = R _θ S with direction is obtained, and the rotated binary criterion descriptor g _n (p, θ): = f _n (p)|( _xi , y _i )∈S _θ ; (5) Set the threshold Haming distance of feature matching (305), and use the same feature image matching training to obtain the threshold Haming distance of the same feature image matching and calculate the average value to obtain the threshold Haming distance of the best feature matching Distance, the threshold Haming distance mean value of this feature matching is set as the judgment standard of whether the feature is matched; (6) Greedy retrieval meets the feature matching pixel block (306), and the greedy algorithm is used to detect the pixel block less than or equal to the setting Haming distance threshold, Retrieve the result of feature matching; (7) Output the feature matching retrieval result (307), retrieve the result of feature matching according to the greedy algorithm, and judge the feature matching result.

Figure 4 shows the binocular CCD mine moving target positioning method and the binocular CCD positioning algorithm flow adopted by the system. As shown in Figure 4, the binocular CCD positioning algorithm process adopted by the mine positioning method mainly includes: (1) the binocular CCD collects left and right images (401) of the moving target, and collects the moving target under the mine by using the binocular CCD visual sensor. left and right images. (2) The binocular CCD carries out stereo calibration (402) to the left and right images of the moving target, and the left and right stereo calibration is carried out to the collected moving target images by the internal parameters of the binocular CCD visual sensor and the binocular CCD visual sensor, to obtain the camera center distance, Focal length f, parallax d _i , internal and external parameter matrix, set world coordinate system coordinates (X _W , Y _W , Z _W ), camera coordinate system coordinates Among them, (X _C , Y _C , Z _C ) are the coordinates of the camera coordinate system, R is a 3×3 rotation matrix, and 0 ^T is a 1×3 translation matrix. Left and right stereo calibration, that is, to calibrate the camera coordinate system and the world coordinate system to estimate the extrinsic parameter matrix Image physical coordinate system Among them, Z _C is the optical axis, x, y are the coordinates of the image physical coordinate system, f is the focal length, and the image pixel coordinate system u, v are the coordinates of the image pixel coordinate system, the image physical coordinate system x, y axis is parallel to the pixel coordinate system u, v axis, d _x , d _y are the pixel spacing in the horizontal and vertical directions respectively, using binocular CCD vision The internal parameters of the sensor and the binocular CCD vision sensor perform left and right stereo calibration on the collected moving target images to solve the internal parameter matrix Wherein f _x , f _y are effective focal lengths; (3) obtain the parallax of binocular CCD (403), according to the matching point pair that binocular CCD visual sensor stereo matching obtains, seek parallax d _i , i=1, 2, ..., N _Ver , and find the mean value of the disparity d _i (4) The image pixel coordinate system equation seeks the world coordinates (404) of the moving target, and obtains the world coordinates (X _W , Y _W , Z _W ) of the moving target according to the image pixel coordinate formula, wherein Z _W is exactly the location of the moving target in space. The vertical distance of the binocular CCD vision sensor, also known as the depth information; (5) correct the world coordinate system coordinates (405) of the moving target through the base station coordinates, and the mine intrinsically safe base station with the built-in binocular CCD vision sensor is in the world coordinate system Coordinates (0, 0, 0), locate the coordinates (X _W , Y _W , Z _W ) in the world coordinate system of the moving target, and perform coordinate position correction on the world coordinate information of the moving target in combination with the actual position of the mine roadway where the base station is located. Obtain the final position information of the moving target in the mine, and realize the precise positioning of the moving target.

Apparently, those skilled in the art should understand that the positioning method and system functions involved in the present invention and the above-mentioned embodiments, in addition to being applied to the coal mine underground environment as a mine mobile target positioning, are also applicable to non-metallic mines after proper integration or improvement. Monitoring, tracking and positioning of moving targets in non-coal mines such as mines and metals, as well as precise positioning of underground intelligent working face equipment. In this way, the present invention does not limit communication technology fields such as non-coal mines, intelligent working surface mobile monitoring and equipment precise positioning except coal mine moving target positioning.

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 (6)

1. A kind of binocular CCD visual mine mobile target location method, the system realizing this method comprises above-hole equipment and down-hole device, it is characterized in that, The above-ground equipment of the system includes a base station controller, a positioning server, an Ethernet switch and a monitoring terminal; The underground device of the system includes a mine intrinsically safe base station and a radio frequency identification tag; the system's uphole equipment communicates with the downhole device via an optical link; The mine intrinsically safe base station of the system has LTE, WIFI, and UWB wireless interfaces, which are used for radio frequency identification of underground moving targets; The mine intrinsically safe base station of the system has a built-in binocular CCD vision sensor, which is used to collect the left and right stereo image information of the moving target in the mine, and perform ORB feature matching and stereo calibration on it; The mine intrinsically safe base station of the system sends the position information calibrated by the binocular CCD visual sensor to the positioning server on the mine through the LTE wireless network and the optical link; The radio frequency identification tag of the system is installed or fixed on the underground mobile target as an identification mark, and is used for RSSI radio frequency identification, feature extraction and training of the mobile target; it is also characterized in that, The binocular CCD visual mine mobile target positioning method, the steps include: (1) The mine intrinsically safe base station uses RSSI algorithm to estimate the position of the moving target by receiving the radio frequency identification tag signal installed or fixed on the moving target; (2) the judgment threshold Thr of defining described binocular CCD visual recognition is the farthest distance of the image information sampling of binocular CCD visual sensor; (3) When the mine intrinsically safe base station monitors that the distance between the moving target and the base station is less than or equal to the decision threshold, that is, when d _Rssi≤Thr , the binocular CCD vision sensor performs image information sampling and distance calculation on the moving target, Otherwise, return to step (1) to re-estimate the position of the moving target; (4) described binocular CCD visual sensor adopts ORB algorithm to carry out feature extraction training to radio frequency identification label, and according to the moving object identification mark feature that extraction feature and binocular CCD visual sensor gather; (5) When the feature matching of the moving target is successful, the binocular CCD visual sensor performs left and right stereo calibration on the moving target image, obtains the internal and external parameter matrix and the parallax mean value of the binocular CCD visual sensor, and adopts binocular CCD visual sensor fusion to calculate The world coordinate information of the moving target, otherwise return to step (1) to re-estimate the position of the moving target; (6) Correct the world coordinate information of the moving target according to the location of the mine intrinsically safe base station, and obtain the final position coordinate information of the moving target. 2. binocular CCD vision mine moving target positioning method according to claim 1, is characterized in that, The mine intrinsically safe base station uses the RSSI algorithm to monitor the received signal strength of the underground mobile target and locate the underground mobile target. According to the propagation loss theoretical model Calculate the sampling and analysis distance Thr of the image information of the moving target by the binocular CCD vision sensor, where P _r (d _Rssi ) is the signal strength when the signal transmission distance in the downhole environment is d _Rssi , and P _r (d ₀ ) is the ideal space state The signal strength when the signal transmission distance is d ₀ , k is the signal strength attenuation coefficient, is the attenuation coefficient of downhole signal strength measured N _Loss times. 3. binocular CCD vision mine shaft moving target positioning method according to claim 1, is characterized in that, the step of carrying out ORB feature matching to underground moving target further comprises: (1) Use the ORB algorithm to detect feature points, and use the FAST operator to detect feature points of downhole moving targets; (2) Add direction information to the detected downhole moving target feature points to form oFAST; (3) Using the gray scale centroid method to add direction information to the corner point of the detected downhole moving target; (4) Use the BRIEF descriptor to describe the feature points of the detected downhole moving target; (5) Add rotation invariance to the detected downhole moving target feature point descriptor to form Steer BRIEF; (6) Calculate the descriptor Hamming distance Ham of two matching point pairs, and perform feature matching judgment; (7) Use the greedy algorithm to search for the set n least correlated pixel block pairs from all possible pixel block pairs, retrieve and match the features of the moving target, and judge the feature matching results. 4. binocular CCD vision mine shaft moving target positioning method according to claim 1, is characterized in that, the step that underground moving target is carried out three-dimensional calibration further comprises: (1) described binocular CCD vision sensor carries out left and right stereo calibration to the mobile target image that gathers, obtains the internal and external parameter matrix of binocular CCD vision sensor; (2) Obtain the parallax mean value of the binocular CCD visual sensor, find the parallax d _i , i=1, 2,..., N _Ver , and find the mean value of the parallax d _i according to the matching point pair obtained by the stereo matching (3) The image pixel coordinate system equation is used to calculate the world coordinates of the moving target, and the coordinates (X _W , Y _W , Z _W ) of the moving target are solved according to the internal and external parameter matrix and the average parallax value d _Ver of the binocular CCD vision sensor, Among them, Z _W is the vertical distance from the spatial positioning moving target to the binocular CCD vision sensor, which is the depth information; (4) Correct the world coordinate system coordinates of the moving target through the coordinates of the base station, the mine intrinsically safe base station with built-in binocular CCD vision sensor has the coordinates (0,0,0) in the world coordinate system, and locates the world coordinate system of the moving target Center coordinates (X _W , Y _W , Z _W ), combined with the actual position of the mine roadway where the base station is located, corrects the coordinate position of the world coordinate information of the moving target, obtains the final position information of the mine moving target, and realizes the precise positioning of the moving target . 5. binocular CCD vision mine moving target positioning method according to claim 1, is characterized in that, Base station controller, used for underground wireless network management, wireless resource management, management, control and monitoring of mine intrinsically safe base stations, and handover control of underground mobile targets; The positioning server is used to process and store real-time location information of mine moving targets, provide mobile positioning services for underground moving targets, and realize the positioning and monitoring functions of underground moving targets; An Ethernet switch, used for the convergence node of the positioning system network, realizes the interconnection between the positioning system and the ground communication network; The monitoring terminal is used for receiving images and real-time positioning information from the positioning server, and visually displaying the images and real-time positioning information. 6. The binocular CCD vision mine moving target positioning method according to claim 1, characterized in that the mine intrinsically safe base station and the radio frequency identification tag are mine intrinsically safe explosion-proof devices.