Hydrogen near-infrared absorption remote monitoring optimized light path and signal processing system

Abstract

A hydrogen near-infrared absorption remote monitoring optimized light path and signal processing system adopts improved near-infrared absorption signals and established scale space pyramids, can perform grid division on the pyramid of each layer of signals so as to increase space scale information, and can eliminate signal errors by using improved optimized light path angular point extraction. The method can overcome the defects that when a traditional near-infrared absorption signal is used for carrying out feature point detection and extraction on a to-be-processed signal, the distribution uniformity degree of extracted feature points is poor, and a feature point aggregation phenomenon occurs in a certain area in the signal, can prevent monitoring errors, and improves the detection accuracy of the to-be-processed signal. The device has the advantages of being non-contact, capable of achieving remote detection and high in reliability and safety.

Description

Technical Field

[0001] The present invention relates to a leakage prevention system, and more specifically to a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system that adopts improved near-infrared absorption signals and establishes a scale space pyramid, can grid the pyramid of each signal layer to increase spatial scale information, and uses improved optimized optical path corner point extraction to eliminate signal errors. Background Art

[0002] High-pressure hydrogen leak prevention requires high safety and reliability. When using traditional near-infrared absorption signals to detect and extract feature points from high-pressure hydrogen leak prevention signals, the extracted feature points are poorly distributed and often cluster in certain areas of the signal. This can lead to monitoring errors, which, when severe, can pose a safety hazard. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system that uses improved near-infrared absorption signals and establishes a scale space pyramid, can grid the pyramid of each layer of signals to increase spatial scale information, uses improved optimized optical path corner point extraction, and can eliminate signal errors.

[0004] The technical solution adopted by the present invention to solve its technical problems is: a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, including improving the near-infrared absorption signal and establishing a scale space pyramid. Establishing the scale space pyramid means gridding the pyramid of each layer of the signal to increase spatial scale information and add grid row and column restrictions. Improving the near-infrared absorption signal means using improved optimized optical path corner point extraction during feature point detection, setting regions, improving feature point extraction accuracy, and considering corner point neighborhood gradients when calculating the Harris corner point response. When eliminating signal error overlapping feature points, a non-maximum suppression method is used to suppress the low-threshold feature point output of the Harris corner point response. The improved near-infrared absorption signal should be set as follows: Region of interest (ROI) is a method frequently used in the field of signal processing. It can select a region from the signal and make this region the focus. Using this method in the improved optimized optical path signal can effectively reduce system processing time and improve feature point extraction accuracy. The following settings should be made in the improved near-infrared absorption signal: in the improved optimized optical path corner point extraction, a signal scale pyramid is constructed, all parameters used in this part are initialized, and each layer of the Gaussian pyramid is gridded, grid row restrictions and grid column restrictions are added, and in the improved optimized optical path corner point extraction flow chart: i is the number of rows, j is the number of columns, rows and cols are the number of rows and columns of each layer of the pyramid respectively. In the process of optimized optical path corner point extraction, the signal boundary needs to be processed, and an effective signal area is added. The initial threshold is used to extract the feature points in the effective signal area, and the method of lowering the threshold first and then detecting is adopted for the grid part where the feature points that are difficult to extract are located. At the same time, a minimum threshold is set for the case where the feature points are empty, to achieve improved optimized optical path corner point extraction, accumulate all extracted feature points, and restore the positions of the feature points extracted in the effective signal area to the divided grid area. In the improved near-infrared absorption signal, the elimination of overlapping feature points should be set as follows: when the feature points are output, the non-maximum suppression method is used to improve the quality of the feature point output and reduce the aggregation and overlapping phenomenon of feature points in certain areas. The essence of this signal is that when there are multiple candidate result values, the data is filtered according to the set target conditions to select the most suitable result. The top layer of the scale space pyramid established by the signal is the first layer, and the feature points ai, ai∈A, A={a1, a2,…, ai}, ai=(xi, yi, εi) that appear on each layer of the pyramid are selected as existing feature points. For the candidate feature points bj (not appearing on each layer), bj∈B, B={b1, b2,…, bj}, bj={xj, yj, εj}, overlapping feature points are eliminated.The equipment for improving near-infrared absorption signals includes a leak prevention module, a host-computer communication module, a chassis drive module, and a wireless communication module. The system operates on a Linux PC with an i3 processor and 4.0GB of memory running Ubuntu 14.04. When the program is running, a high-pressure hydrogen leak detected by a Kinect 1.0 depth-prevention probe is detected at a linear velocity of 0.02m / s and an angular velocity of 0.01m / s, achieving leak prevention. The feature points of the improved near-infrared absorption signal are extracted as follows: The feature point extraction process is performed on the system's high-pressure hydrogen leak signal data using the improved near-infrared absorption signal. The feature points extracted from the improved near-infrared absorption signal are evenly distributed, with a high density of feature points in certain areas.

[0005] The beneficial effects of the present invention are as follows: a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system of the present invention uses improved near-infrared absorption signals and establishes a scale space pyramid, which can grid the pyramid of each layer of the signal to increase spatial scale information, and uses improved optimized optical path corner point extraction to eliminate signal errors. The present invention can overcome the shortcomings of using traditional near-infrared absorption signals to detect and extract feature points of the processed signal, such as poor uniformity of the distribution of the extracted feature points and the occurrence of feature point aggregation in a certain area of ​​the signal. The present invention can prevent monitoring errors and prevent safety hazards caused by errors in the monitoring signal. The present invention has the advantages of non-contact and remote detection, as well as high reliability and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The present invention will be further described below with reference to the accompanying drawings and examples.

[0007] Figure 1 This is a flow chart of improving the near-infrared absorption signal according to the first embodiment of the present invention.

[0008] Figure 2 This is a flowchart of the improved and optimized light path corner point extraction according to the first embodiment of the present invention. DETAILED DESCRIPTION

[0009] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, including improving the near-infrared absorption signal and establishing a scale space pyramid. Establishing the scale space pyramid refers to gridding the pyramid of each layer of the signal to increase spatial scale information and add grid row restrictions and column restrictions. Improving the near-infrared absorption signal refers to using improved optimized optical path corner point extraction, setting areas, improving feature point extraction accuracy, and considering the corner point neighborhood gradient when calculating the Harris corner point response. When eliminating signal error overlapping feature points, a non-maximum suppression method is used to suppress the low threshold feature point output of the Harris corner point response.

[0010] In implementing the present invention, a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system should be set as follows in the improved near-infrared absorption signal: Region of interest (ROI) is a method frequently used in the field of signal processing. It can select an area from the signal and make this area the focus. In the improved optimized optical path signal, using this method can effectively reduce the system processing time and improve the accuracy of feature point extraction.

[0011] A hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system of the present invention is implemented. The following settings should be made in the improved near-infrared absorption signal: in the improved optimized optical path corner point extraction, a signal scale pyramid is constructed, all parameters used in this part are initialized, and each layer of the Gaussian pyramid is gridded, and grid row restrictions and grid column restrictions are added. In the improved optimized optical path corner point extraction flow chart: i is the number of rows, j is the number of columns, rows and cols are the number of rows and columns of each layer of the pyramid, respectively. In the process of optimized optical path corner point extraction, the signal needs to be boundary processed, an effective signal area is added, and the initial threshold is used to extract feature points in the effective signal area. The method of first lowering the threshold and then detecting is adopted for the grid part where the feature points that are more difficult to extract are located. At the same time, a minimum threshold is set for the case where the feature points are empty, to achieve improved optimized optical path corner point extraction, accumulate all extracted feature points, and restore the positions of the feature points extracted in the effective signal area to the divided grid area.

[0012] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, improves the near-infrared absorption signal, and eliminates overlapping feature points as follows: when the feature point is output, a non-maximum suppression method is used to improve the quality of the feature point output and reduce the aggregation and overlapping phenomenon of feature points in certain areas. The essence of the signal is that when multiple candidate result values ​​appear, the data is filtered according to the set target conditions to select the most suitable result. The top layer of the scale space pyramid established by the signal is the first layer, and the feature points ai, ai∈A, A={a1, a2, ..., ai}, ai=(xi, yi, εi) that appear on each layer of the pyramid are selected as existing feature points. For candidate feature points bj (not appearing on each layer), bj∈B, B={b1, b2, ..., bj}, bj={xj, yj, εj}, the process of eliminating overlapping feature points is as follows:

[0013] Step 1: When outputting feature points, calculate the distance between a grid area on the pyramid layer where the existing candidate feature point bj in the signal is located and the coordinates of the existing feature point ai, and set the critical threshold of the distance to 1;

[0014] Step 2: Calculate the distance between the coordinates of the candidate feature point bj = {xj, yj, εj} and the feature point A = {a1, a2, ..., ai}, and select the candidate feature point bj′ = {xj, yj, εj} with a distance ≤ 1;

[0015] Step 3: Calculate the Harris corner response threshold εj′ of the candidate feature points bj′={xj, yj, εj} whose distance value does not exceed l in Step 2, and arrange these candidate feature points in descending order according to the intensity of the response threshold of the candidate feature point b={xj, yj, εj}, that is, the value of εj′;

[0016] Step 4: According to the principle of non-maximum suppression signal, the candidate feature point bj′={xj, yj, εj} where the Harris corner response threshold εj′ is the maximum is retained, and the low threshold adjacent candidate feature points are eliminated to obtain the adjacent feature points at the best position of the feature point, so as to achieve the purpose of reducing the output of overlapping feature points.

[0017] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, improves the near-infrared absorption signal should be based on the regional statistical variance feature point distribution evaluation, the specific evaluation is as follows: the evaluation of the uniformity of the signal feature point distribution is mainly used to estimate whether the number of feature points is the same in the signal of different regions. The same or similar number of feature points indicates that the uniformity of the feature point distribution in the signal is good. The large difference in the number of feature points indicates that the uniformity of the feature point distribution in the signal is poor. In view of the above evaluation method, a feature point distribution uniformity evaluation method based on the regional statistical variance value is established. First, the signal of the feature point to be detected is divided into five regions according to vertical, horizontal, center and surrounding, upper left to lower right, and lower left to upper right, and the above is obtained. , bottom, left, right, lower left, upper right, upper left, lower right, center and surrounding directions of 10 signal areas, and then the number of feature points in each area of ​​the signal to be detected is counted according to the above division directions to obtain the data values ​​of the feature points in 10 directions, and then the average value of the number of feature points in the 10 divided areas of the signal to be detected is calculated, and the variance value of the number of feature points of the signal to be detected in the 10 divided areas is calculated. It can be inferred from the size relationship of the obtained variance value of the number of feature points that the larger the variance value of the number of feature points, the more obvious the floating of the data, which indirectly reflects that the uniformity of the distribution of feature points in the signal is poor, and the smaller the variance value of the number of feature points, the smaller the floating of the data, which indirectly reflects that the uniformity of the distribution of feature points in the signal is good.

[0018] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system. The equipment for improving the near-infrared absorption signal includes: a leakage prevention module, a host and host computer communication module, a chassis drive module, and a wireless communication module. The system operating environment is a Linux PC configured with an i3 processor and 4.0GB of memory and running Ubuntu 14.04. When the program is running, high-pressure hydrogen leaks equipped with a Kinect 1.0 depth leakage prevention probe are detected at a linear velocity of 0.02m / s and an angular velocity of 0.01m / s, thereby preventing the leakage of high-pressure hydrogen.

[0019] A hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system of the present invention is implemented to improve the near-infrared absorption signal. The characteristic points of the signal are extracted as follows: the improved near-infrared absorption signal is used to perform a characteristic point extraction link on the high-pressure hydrogen leakage signal data of the system. The characteristic points extracted by the improved near-infrared absorption signal are evenly distributed, and the characteristic points are relatively dense in a certain area.

[0020] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, which uses an improved near-infrared absorption signal. When using a traditional near-infrared absorption signal to detect and extract feature points from the signal to be processed in a high-pressure hydrogen leak prevention signal, the extracted feature points are poorly distributed and feature points may cluster within a certain area of ​​the signal. To address this issue, an improved near-infrared absorption signal is proposed, which establishes a scale space pyramid and grids each layer of the signal pyramid to increase spatial scale information and impose grid row and column restrictions. When detecting feature points, an improved optimized optical path corner point extraction is used, and regions are set to improve feature point extraction accuracy. The corner point neighborhood gradient is considered when calculating the Harris corner response. When eliminating feature points overlapping with signal errors, a non-maximum suppression method is used to suppress the output of low-threshold feature points from the Harris corner response.

[0021] When implementing the present invention's optimized optical path and signal processing system for remote monitoring of hydrogen near-infrared absorption, the following should be noted in the flowchart for improving near-infrared absorption signals: Region of interest (ROI) is a frequently used method in signal processing. It selects a region within a signal and makes it the focal point. Using this method in the improved optimized optical path signal effectively reduces system processing time and improves the accuracy of feature point extraction.

[0022] In implementing the present invention, a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, the following should be noted in the improved optimized optical path corner point extraction flow chart: in the improved optimized optical path corner point extraction, a signal scale pyramid is constructed, all parameters used in this part are initialized, and each layer of the Gaussian pyramid is gridded. Grid row restrictions and grid column restrictions are added. In the improved optimized optical path corner point extraction flow chart: i is the number of rows; j is the number of columns; rows and cols are the number of rows and columns for each layer of the pyramid, respectively. In the process of optimizing the optical path corner point extraction, the signal needs to be processed at the boundary, and an effective signal area is added; the initial threshold is used to extract the feature points in the effective signal area, and the method of first lowering the threshold and then detecting is adopted for the grid part where the feature points that are more difficult to extract are located. At the same time, a minimum threshold is set for the case where the feature points are empty, to achieve improved optimized optical path corner point extraction; all extracted feature points are accumulated, and the positions of the feature points that have been extracted in the effective signal area are restored to the divided grid area.

[0023] In implementing the present invention, a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, the following should be noted when eliminating overlapping feature points: when outputting feature points, a non-maximum suppression method is used to improve the quality of feature point output and reduce the phenomenon of clustering and overlapping of feature points in certain areas. The essence of this signal is that when multiple candidate result values ​​appear, the data is filtered according to the set target conditions to select the most suitable result. The top layer of the scale space pyramid established with the signal is the first layer, and the feature points ai, ai∈A, A={a1, a2,…, ai}, ai=(xi, yi, εi) that appear on each layer of the pyramid are selected as existing feature points. For candidate feature points bj (not appearing on each layer), bj∈B, B={b1, b2,…, bj}, bj={xj, yj, εj}. Eliminate overlapping feature points.

[0024] In implementing the present invention's optimized optical path and signal processing system for remote monitoring of hydrogen near-infrared absorption, the regional statistical variance-based feature point distribution evaluation is performed as follows: The evaluation of signal feature point distribution uniformity primarily estimates whether the number of feature points is consistent across signals in different regions. If the number of feature points is the same or similar, the signal's feature point distribution is relatively uniform; if the number of feature points differs significantly, the signal's feature point distribution is relatively poorly uniform. Based on this evaluation method, a feature point distribution uniformity evaluation method based on regional statistical variance values ​​is established. First, the signal of the feature point to be detected is divided into five regions according to the five directions of vertical, horizontal, center and surrounding, upper left to lower right, and lower left to upper right, and the directions of 10 signal regions are obtained, namely, upper, lower, left, right, lower left, upper right, upper left, lower right, center and surrounding. Then, the number of feature points in each region of the signal to be detected is counted according to the above division directions, and the data values ​​of the feature points in the 10 directions are obtained. Then, the average value of the number of feature points in the 10 divided regions of the signal to be detected is calculated; and the variance value of the number of feature points in the 10 divided regions of the signal to be detected is calculated. From the relationship between the obtained variance values ​​of the number of feature points, it can be inferred that the larger the variance value of the number of feature points, the more obvious the fluctuation of the data, which indirectly reflects the poor uniformity of the distribution of feature points in the signal; the smaller the variance value of the number of feature points, the smaller the fluctuation of the data, which indirectly reflects the good uniformity of the distribution of feature points in the signal.

[0025] The experimental results and analysis of the present invention's optimized optical path and signal processing system for remote monitoring of hydrogen near-infrared absorption are as follows. The system includes a leak prevention module, a host-server communication module, a chassis drive module, and a wireless communication module. The system operates on a Linux PC with an i3 processor and 4.0GB of RAM running Ubuntu 14.04. When the program is running, a Kinect 1.0 depth detection probe detects high-pressure hydrogen leaks at a linear velocity of 0.02 m / s and an angular velocity of 0.01 m / s, effectively preventing leaks.

[0026] A hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system of the present invention is implemented, and the signal feature point extraction experiment is as follows: the improved near-infrared absorption signal is used to perform the feature point extraction link on the high-pressure hydrogen leakage signal data of the ROS system. The feature points extracted by the improved near-infrared absorption signal are evenly distributed, and the feature points are relatively dense in a certain area.

[0027] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, and the signal feature point distribution experiment is as follows: when extracting feature points of the signal, the feature point distribution uniformity obtained by using the improved near-infrared absorption signal is better than that of the traditional near-infrared absorption signal. A signal feature point distribution uniformity judgment method based on regional statistical variance is adopted to calculate the variance values ​​of the feature points distributed in 10 regions in the signal, thereby evaluating the distribution of the signal feature points. The results are shown in Table 1.

[0028]

[0029] Table 1

[0030] Table 1 is the comparative data of the uniformity of distribution of signal feature points in different signals:

[0031] It can be seen from Table 1 that with the increase of the expected number of feature points, the variance value of the output number of feature points in the signal and the reduction rate of the variance value are also changing; when the expected number of feature points is the same, the variance value of the improved near-infrared absorption signal is smaller than that of the traditional near-infrared absorption signal, indicating that the signal feature points extracted using the improved near-infrared absorption signal are more uniformly distributed.

[0032] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, and the signal feature point matching experiment is as follows: the feature point extraction results of the traditional near-infrared absorption signal and the improved near-infrared absorption signal are used as the signal matching points, and a matching experiment based on the feature points is performed. First, a rough match is obtained by using brute force matching combined with the Hamming distance method, and then a RANSAC signal is used to eliminate the false matches in the rough match to obtain a fine match. When performing feature point matching, the improved near-infrared absorption signal uses a non-maximum suppression method to process the threshold of overlapping feature points when outputting feature points, so that it only outputs feature points with a maximum threshold in a certain range. In addition, the distribution of feature points obtained using the improved near-infrared absorption signal is relatively uniform. Therefore, when using the improved near-infrared absorption signal for feature point matching, the line segments between the constraint matching points are relatively sparsely distributed, and there will be no special aggregation.

[0033] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system. The signal real-time analysis experiment is as follows: Because high-pressure hydrogen has relatively strict requirements on the length of time consumed in the signal execution process when preventing leakage, this paper uses traditional near-infrared absorption signals and improved near-infrared absorption signals to compare the time consumed in the feature point extraction and matching links in the signal execution step. The experimental results are shown in Table 2. As can be seen from Table 2, the larger the expected number of feature points set, the longer the time consumed for signal execution feature point extraction and matching. When the number of expected feature points is set to the same, from the perspective of signal execution time, the time consumed for feature point extraction and matching using the improved near-infrared absorption signal for the same signal is less than that of the traditional near-infrared absorption signal, indicating that the improved near-infrared absorption signal is less time-consuming and has higher real-time performance.

[0034] Table 2 is the real-time analysis of the near-infrared absorption signal.

[0035]

[0036] Table 2

[0037] The present invention implements a hydrogen near-infrared absorption remote monitoring and optimization optical path and signal processing system, including the design and development of hardware and software, wherein the software R&D tasks include the design and development of the underlying driver, the physical layer and the access control layer (MAC), the network layer and the network management system. The present invention mainly includes: the design and development of the underlying driver, the design and development of the MDI module, MDI (MCUDSPInterafce) is the PCI bus interface between the physical layer and the MAC layer, and is the bridge between the MCU and the DSP. The design and development of the physical layer is mainly responsible for the design and implementation of the OFDMA physical layer, and is also responsible for the implementation of the interface between the physical layer and the MAC layer. The design and development of the network management system is mainly responsible for the design and development of Agnet in the WIMxa network management system, and is also responsible for the design and development of the communication interface between Agnet and Mnagaer. System testing work, during the system integration process, is mainly responsible for the integration of the physical layer and the MAC layer, designing the integration plan and performing real-time testing. During the system testing process, it is responsible for the formulation of the system testing plan and completes part of the system testing work.

[0038] The present invention implements a hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system. This system utilizes improved near-infrared absorption signals and establishes a scale-space pyramid. This system can grid each signal pyramid layer to increase spatial scale information. It also uses improved optimized optical path corner point extraction to eliminate signal errors. The present invention overcomes the shortcomings of using traditional near-infrared absorption signals for feature point detection and extraction of processed signals, resulting in poorly uniform distribution of extracted feature points and the occurrence of feature point clustering within certain areas of the signal. The present invention prevents monitoring errors and potential safety hazards caused by errors in monitoring signals. The present invention offers the advantages of non-contact, remote detection, and high reliability and safety.

[0039] The implementation of the present invention has a very good effect.

Claims

1. A hydrogen near-infrared absorption remote monitoring optimized optical path and signal processing system, including improving near-infrared absorption signals and establishing a scale space pyramid, characterized in that: The establishment of the scale space pyramid refers to gridding the pyramid of each layer of the signal to increase the spatial scale information and add grid row restrictions and column restrictions; improving the near-infrared absorption signal refers to using improved and optimized optical path corner point extraction, setting areas, improving the feature point extraction accuracy, and considering the corner point neighborhood gradient when calculating the Harris corner point response part during feature point detection. When eliminating signal error overlapping feature points, a non-maximum suppression method is used to suppress the low threshold feature point output of the Harris corner point response.

2. A hydrogen near-infrared absorption remote monitoring optimization optical path and signal processing system according to claim 1, characterized in that The improved near-infrared absorption signal should be set as follows: Region of interest (ROI) is a method frequently used in the field of signal processing. It can select a region from the signal and make this region the focus. In the improved optimized optical path signal, using this method can effectively reduce the system processing time and improve the accuracy of feature point extraction.

3. The hydrogen near-infrared absorption remote monitoring optimization optical path and signal processing system according to claim 1 is characterized in that The improved near-infrared absorption signal should be set as follows: in the improved optimized optical path corner point extraction, a signal scale pyramid is constructed, all parameters used in this part are initialized, and each layer of the Gaussian pyramid is gridded, and grid row restrictions and grid column restrictions are added. In the improved optimized optical path corner point extraction flow chart: i is the number of rows, j is the number of columns, rows and cols are the number of rows and columns of each layer of the pyramid respectively. In the process of optimized optical path corner point extraction, the signal needs to be boundary processed, an effective signal area is added, and the initial threshold is used to extract the feature points in the effective signal area. For the grid part where the feature points that are difficult to extract are located, the method of first lowering the threshold and then detecting is adopted. At the same time, a minimum threshold is set for the case where the feature points are empty, to achieve improved optimized optical path corner point extraction, accumulate all extracted feature points, and restore the positions of the feature points extracted in the effective signal area to the divided grid area.

4. The hydrogen near-infrared absorption remote monitoring optimization optical path and signal processing system according to claim 1 is characterized in that In the improved near-infrared absorption signal, the elimination of overlapping feature points should be set as follows: when the feature points are output, the non-maximum suppression method is used to improve the quality of the feature point output and reduce the aggregation and overlapping phenomenon of feature points in certain areas. The essence of the signal is that when multiple candidate result values ​​appear, the data is filtered according to the set target conditions to select the most suitable result. The top layer of the scale space pyramid established by the signal is the first layer, and the feature points ai, ai∈A, A={a1, a2, ..., ai}, ai=(xi, yi, εi) that appear on each layer of the pyramid are selected as existing feature points. For candidate feature points bj (not appearing on each layer), bj∈B, B={b1, b2, ..., bj}, bj={xj, yj, εj}, the process of eliminating overlapping feature points is: Step 1: When outputting feature points, calculate the distance between a grid area on the pyramid layer where the existing candidate feature point bj in the signal is located and the coordinates of the existing feature point ai, and set the critical threshold of the distance to l; Step 2: Calculate the distance between the coordinates of the candidate feature point bj = {xj, yj, εj} and the feature point A = {a1, a2, ..., ai}, and select the candidate feature point bj′ = {xj, yj, εj} with a distance ≤ l; Step 3: Calculate the Harris corner response threshold εj′ of the candidate feature points bj′={xj, yj, εj} whose distance values ​​are less than 1 in Step 2, and arrange these candidate feature points in descending order according to the intensity of the response threshold of the candidate feature point b={xj, yj, εj}, that is, the value of εj′; Step 4: According to the principle of non-maximum suppression signal, the candidate feature point bj′={xj, yj, εj} where the Harris corner response threshold εj′ is the maximum is retained, and the low threshold adjacent candidate feature points are eliminated to obtain the adjacent feature points at the best position of the feature point, so as to achieve the purpose of reducing the output of overlapping feature points.

5. The hydrogen near-infrared absorption remote monitoring optimization optical path and signal processing system according to claim 1 is characterized in that The improved near-infrared absorption signal should be evaluated based on the feature point distribution of regional statistical variance. The specific evaluation is as follows: the evaluation of the uniformity of the distribution of signal feature points is mainly used to estimate whether the number of feature points is the same in signals of different regions. If the number of feature points is the same or similar, it means that the uniformity of the distribution of feature points in the signal is good. If the number of feature points is quite different, it means that the uniformity of the distribution of feature points in the signal is poor. In view of the above evaluation method, a method for evaluating the uniformity of the distribution of feature points based on the regional statistical variance value is established. First, the signal of the feature point to be detected is divided into five regions according to vertical, horizontal, center and surrounding, upper left to lower right, and lower left to upper right, and the top, bottom, left, right, lower left, upper right, and left are obtained. The directions of 10 signal areas, such as the top, bottom right, center and surrounding areas, are then counted for the number of feature points in each area of ​​the signal to be detected according to the above division directions, and the data values ​​of the feature points in 10 directions are obtained. Then, the average value of the number of feature points in the 10 divided areas in the signal to be detected is calculated, and the variance value of the number of feature points of the signal to be detected in the 10 divided areas is calculated. It can be inferred from the size relationship of the obtained variance value of the number of feature points that the larger the variance value of the number of feature points, the more obvious the floating of the data, which indirectly reflects that the uniformity of the distribution of feature points in the signal is poor, and the smaller the variance value of the number of feature points, the smaller the floating of the data, which indirectly reflects that the uniformity of the distribution of feature points in the signal is good.

6. The hydrogen near-infrared absorption remote monitoring optimization optical path and signal processing system according to claim 1 is characterized in that The device for improving near-infrared absorption signals includes: a leakage prevention module, an upper and lower computer communication module, a chassis drive module, and a wireless communication module. The system operating environment is a Linux PC configured with an i3 processor and 4.0GB of memory and running Ubuntu 14.

04. When the program is running, high-pressure hydrogen leaks equipped with a Kinect 1.0 depth leakage prevention probe are collected at a linear speed of 0.02m / s and an angular speed of 0.01m / s to achieve high-pressure hydrogen leakage prevention.

7. The hydrogen near-infrared absorption remote monitoring optimization optical path and signal processing system according to claim 1 is characterized in that The improved near-infrared absorption signal, the signal feature points are extracted as follows: the improved near-infrared absorption signal is used to perform a feature point extraction link on the high-pressure hydrogen leakage signal data of the system, and the feature points extracted by the improved near-infrared absorption signal are evenly distributed, and the feature points in a certain area are relatively dense.

Images