Non-contact laser explosive detection method

Abstract

The invention discloses a non-contact laser explosive detection method. The method comprises the following steps: generating a stable infrared radiation light source; performing secondary beam expansion on the infrared radiation light source to obtain uniform and stable laser spots; acquiring an infrared thermal image of the detected explosive based on the laser spot, and synchronously acquiring a visible light image of the detected explosive; and matching the infrared thermal image and the visible light image to form an infrared thermal image scene graph, and identifying and analyzing the infrared thermal image scene graph based on a deep convolutional neural network. According to the invention, the temperature difference characteristics of the detected explosive are extracted through the fusion identification technology of the infrared image and the visible light image, so that non-contact, rapid and safe explosive detection is realized, and the actual combat requirements of explosive detection are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of explosive detection using infrared laser, and particularly relates to a non-contact laser explosive detection method. BACKGROUND

[0003] At present, explosive detection technologies are divided into two categories according to the size of the detection object: one is bulk detection technology, mainly aimed at larger volume explosives, including imaging detection technology (such as millimeter wave imaging, microwave imaging, terahertz wave imaging, X-ray imaging) and nuclear-based detection technology (such as neutron detection, nuclear quadrupole resonance analysis). Such technology devices are mature but bulky, complex in structure, and high in cost, suitable for high-precision detection in fixed places, but difficult to meet the needs of rapid and portable scenes. The second is trace detection technology, aimed at micro explosives (such as steam, particulate matter), including spectral detection technology (ion mobility spectrometry, gas chromatography, Raman spectroscopy, laser-induced breakdown spectroscopy), chemical sensor technology, and biological sensor technology. Among them, spectral detection technology is widely used, but has the problems of limited detection distance, easy environmental interference, and high equipment cost; other technologies are limited by sensitivity, stability or applicable scenarios, and have not yet been widely popularized.

[0004] At present, spectral technology is still in the exploratory stage for non-contact explosive detection. Although some progress has been made in the application of laser spectroscopy technology, terahertz spectroscopy technology, and infrared thermal imaging technology in explosive detection, there are still the following key problems: first, most trace detection technologies (such as Raman spectroscopy, ion mobility spectrometry) have small detection areas (only a few centimeters to several meters), making it difficult to meet the rapid screening needs of large passenger flow places; second, terahertz and laser-induced breakdown spectroscopy technologies are expensive and complex to operate, making it difficult to popularize them at the grassroots level; third, spectral technology is easily affected by light, temperature, and humidity, and its detection accuracy decreases in complex scenes; fourth, infrared thermal imaging technology is less used in trace explosive detection, and traditional methods have problems such as insufficient temperature resolution and inability to distinguish between trace targets and background interference.

[0005] In view of the above problems, the combination of laser irradiation and infrared thermal imaging technology has gradually attracted attention. This technology uses the temperature effect of explosives after absorbing specific wavelength laser, combined with image recognition to achieve non-contact detection, which is expected to break through the limitations of traditional technology. However, in existing research, the application of laser thermal imaging combined with intelligent algorithms is less, and there is a lack of practical detection methods that take into account large-scale, high efficiency, and low cost.

[0006] For example: the application number is 202011069682.3, disclose a kind of non-contact laser explosive fast detection system, using the application scheme can be realized to different types of explosives non-contact detection, enhance the adaptability to different types of explosives, but its scheme simultaneously exists: detection precision is insufficient, it is difficult to accurately identify the weak characteristics of explosive under complex environment;Environment adaptability is poor, in high temperature, high humidity, strong light and other adverse environments, system performance is affected.

[0007] Therefore, a non-contact, fast and safe explosive laser detection method is needed to meet the key direction of operational requirements. SUMMARY

[0008] In view of the above problems, the purpose of the present application is to provide a kind of non-contact laser explosive detection method, solve the many problems existing in detection range, cost, operation difficulty, detection precision and other aspects of explosive detection technology.

[0009] The embodiment scheme of the present application includes:

[0010] In a first aspect, a non-contact laser explosive detection method includes the following steps:

[0011] S1, generate stable infrared radiation light source;

[0012] S2, two-stage beam expansion is carried out on the infrared radiation light source, and uniform and stable laser spot is obtained;

[0013] S3, based on laser spot, infrared thermal image of the detected explosive is collected, and visible light image of the detected explosive is collected synchronously;

[0014] S4, match the infrared thermal image and the visible light image to form an infrared thermal image scene graph, and identify and analyze the infrared thermal image scene graph based on a deep convolutional neural network.

[0015] In one embodiment of the present application, the S1 generates a stable infrared radiation light source, including the following steps:

[0016] S11, the power of CO2 laser is regulated by PLC to generate a continuous / pulsed infrared radiation light source with stable power;

[0017] S12, the pulse frequency of the infrared radiation light source is regulated by a frequency controller to obtain an infrared radiation light source with stable frequency;

[0018] S13, the working temperature of the laser is adjusted in real time by a water chiller to obtain an infrared radiation light source with stable temperature.

[0019] In one embodiment of the present application, the S2 carries out two-stage beam expansion on the infrared radiation light source to obtain uniform and stable laser spot, including the following steps:

[0020] S21, a reverse Galileo telescope system is used to perform primary beam expansion of the light source, and a primary expanded light source with a magnification of not less than 4 is obtained;

[0021] S22, an off-axis Cassegrain telescope system is used to perform secondary beam expansion of the primary expanded light source, and a secondary expanded light source with a magnification of not less than 23 is obtained,

[0022] S23, the secondary expanded light source is subjected to optical path calibration to form a large-diameter and uniform stable laser spot.

[0023] In an embodiment of the present application, the infrared radiation light source forms a uniform and stable laser spot, the total beam expansion magnification is greater than or equal to 92, and a uniform and stable spot with a diameter greater than or equal to 160 mm is formed at a distance of 2000 mm, and the beam divergence angle is less than or equal to 25 mu rad.

[0024] In an embodiment of the present application, the S3 comprises the following steps:

[0025] S31, based on the uniform and stable laser spot, infrared imaging of the detected explosive is performed to obtain an infrared thermal image;

[0026] S32, synchronous triggering of visible light imaging of the detected explosive is performed to collect a visible light image of the detected explosive;

[0027] In an embodiment of the present application:

[0028] The infrared thermal image has a pixel number of not less than 1280*1024, and the working waveband covers 8~14 mu m;

[0029] The visible light image has a pixel number of not less than 1024*1024;

[0030] The time difference of infrared / optical dual-image collection is less than or equal to 10 ms.

[0031] In an embodiment of the present application, the S4 comprises the following steps:

[0032] S41, noise reduction and registration processing of the infrared thermal image and the visible light image are performed to form an infrared thermal image scene graph;

[0033] S42, based on a deep convolutional neural network, image fusion processing of the infrared thermal image scene graph is performed, temperature features of the fused image are extracted, and pixel-level segmentation of the infrared thermal image scene graph is performed;

[0034] S43, based on pre-trained explosive category recognition, the segmented image is subjected to explosive category discrimination.

[0035] In an embodiment of the present application, the deep convolutional neural network adopts a DetectNet model.

[0036] In one embodiment of the present application, the method further comprises the step of: S5, human-computer interaction is carried out through the touch screen, real-time images, detection results and device states are displayed, when a suspicious object is detected, an alarm mechanism is triggered, and a detection log is recorded at the same time.

[0037] In one embodiment of the present application, the alarm mechanism adopts a multi-level response, including: a red warning box is popped up through the touch screen, a built-in buzzer and an external red LED warning light are synchronously triggered, and the alarm lasts for 10 seconds or until an operator manually confirms; if no confirmation is obtained within 3 seconds, the system automatically pushes the alarm information to the associated security terminal.

[0038] The second aspect: an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method provided in the first aspect when executing the program.

[0039] The third aspect: a non-transitory computer readable storage medium, having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the method provided in the first aspect.

[0040] The beneficial effects of the present application are as follows:

[0041] 1. The method of the present application utilizes the temperature effect characteristics caused by the absorption of infrared laser by explosives (such as TNT, RDX, KNO3, etc.), uses a high-collimation optical system to irradiate the detected sample at a long distance, uses an infrared imaging system and a visible light imaging system to simultaneously detect the image of the detected sample, and extracts the temperature difference characteristics of the detected explosive through the fusion recognition technology of the infrared image and the visible light image, so as to realize non-contact, fast and safe explosive detection, and meet the actual combat needs of explosive detection.

[0042] 2. The method of the present application forms a uniform light spot with a diameter of ≥160mm within a distance of 1-3m through the cooperation of a laser and a two-stage beam expansion technology, and can complete the detection without contacting the detected object, thereby avoiding the explosion risk caused by contact operation. At the same time, the laser power is controllable and supports pulse mode, and the water-cooled machine is used to control the temperature in real time, thereby reducing the thermal damage to the detected object, and the method is suitable for security screening in personnel-intensive places such as passenger stations and subway stations, and takes into account safety and scene adaptability.

[0043] 3. The secondary beam expansion system of the method has a total magnification of greater than or equal to 92 times, and the coverage of a single detection is significantly larger than that of a conventional trace detection technology. The infrared thermal image and the visible light image are synchronously collected (with a time difference of less than or equal to 10 ms), and large-area rapid detection is realized. Through the pixel-level segmentation and identification of the fused image by the DetectNet deep convolutional neural network, the single-frame processing time is short, the identification accuracy of common explosives such as TNT and RDX is greater than or equal to 80%, the detection speed and accuracy are effectively balanced, and the high-efficiency screening of large passenger flow scenes is adapted.

[0044] 4. The infrared thermal imaging system of the method can capture the small temperature difference between the explosive and the background. The image fusion and noise reduction technology reduces the environmental interference such as light and humidity. The hardware adopts mature CO2 laser, uncooled infrared detector and other devices, and compared with the terahertz technology, the cost is reduced. The touch screen man-machine interaction, multi-level alarm mechanism operation is simple, supports permission management and log tracing, and is convenient for rapid deployment and use by grassroots units. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 Fig. 1 is a flowchart of a non-contact laser explosive detection method according to the present application;

[0046] Figure 2 Fig. 4 is a structural schematic diagram of an electronic device according to the present application. DETAILED DESCRIPTION

[0047] Embodiments of the present application will be described in detail below, examples of which are shown in the drawings, wherein the same or similar symbols represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation of the present application.

[0048] The existing non-contact laser explosive detection has the following problems: first, the detection precision is insufficient, and it is difficult to accurately identify the weak characteristics of explosives in complex environments; second, the detection speed is slow, and in scenes with large passenger flow and heavy security check tasks, the security check efficiency is low; third, the environmental adaptability is poor, and in harsh environments such as high temperature, high humidity and strong light, the system performance is significantly affected; fourth, the data security guarantee is not perfect, there is a risk of data leakage during transmission and storage, and there is a lack of effective permission management and log tracing mechanism; fifth, there is a lack of effective suspicious object response mechanism, and when a suspicious object is detected, it cannot be timely and accurately alarmed and pushed, which delays the security check opportunity.

[0049] Embodiment 1:

[0050] In view of the above problems, the present application provides a non-contact laser explosive detection method, as shown in Figure 1 The method comprises the following steps:

[0051] S1, a stable infrared radiation light source is generated.

[0052] First, the power of the CO2 laser is regulated by the PLC to generate a continuous / pulsed infrared radiation light source with stable power. A 10.6 μm wavelength CO2 laser can be used, with a power adjustable range of 0-100 W, single frequency output, and a beam quality M² factor ≤ 1.2, to ensure a stable infrared radiation source. The PLC controller, frequency controller, and water-cooled machine are also provided. The PLC controller can precisely adjust the output power of the CO2 laser through a pre-set program, supporting continuous and pulsed output modes. In continuous mode, the power can be adjusted steplessly in the range of 0-100 W, meeting the needs of different detection distances (1-3 m) and explosive concentrations (≤ 1.5 mg / cm²).

[0053] Then, the frequency of the infrared radiation light source is regulated by the frequency controller to obtain a frequency-stable infrared radiation light source. In pulsed mode, the frequency controller can set a pulse frequency of 1 kHz. By controlling the periodic output of the laser, the thermal damage to the detected sample is reduced, while ensuring that the explosive produces a recognizable temperature change after absorbing the laser energy.

[0054] Then, the water-cooled machine is used to adjust the laser operating temperature in real time to obtain a temperature-stable infrared radiation light source. The water-cooled machine works in conjunction with the laser, and the built-in temperature sensor monitors the laser cavity temperature in real time. When the temperature exceeds 50℃, the water circulation cooling system is automatically started to ensure that the laser operating temperature is stable at 25-50℃, avoiding wavelength drift or power fluctuations due to overheating, and ensuring the stability of the infrared radiation source.

[0055] Further, a multiple safety protection mechanism is adopted. When the detected laser output power deviates from the set value by more than ±5%, or the water cooling system fails, the PLC controller immediately sends a signal to cut off the laser output, displays the fault code through human-computer interaction, and triggers an audible and visual alarm to ensure the safety of the equipment. The entire detection system provides a stable and controllable infrared laser radiation source, which is the core energy foundation for non-contact detection of explosives, and its performance directly affects the accuracy of subsequent laser beam expansion, image acquisition, and image analysis.

[0056] S2, the infrared radiation light source is subjected to secondary beam expansion to obtain a uniform and stable laser spot.

[0057] Firstly, a reversed Galileo telescope system is used to expand the light source by one level, and a light source with an expansion ratio of not less than 4 times is obtained. The reversed Galileo telescope system of the one-level expansion unit is composed of a negative lens and a positive lens. The negative lens is made of zinc selenide (ZnSe) material, and the concave curvature radius is optimized and designed to match the initial divergence characteristics of the laser beam. The positive lens is also made of ZnSe material. The distance between the two lenses (24.39 mm) is accurately calculated to ensure that the output beam maintains good collimation while reducing beam aberration and expanding the beam by 4 times.

[0058] Then, an off-axis Cassegrain telescope system is used to expand the one-level expanded light source by not less than 23 times to obtain a two-level expanded light source. In the off-axis Cassegrain telescope system of the two-level expansion unit, the primary mirror is made of fused quartz with an ellipsoidal surface (conic constant k=-0.9595, vertex curvature radius 644.9 mm), and the secondary mirror is a spherical surface (vertex curvature radius 28.31 mm). The two mirrors are axially spaced by 308.305 mm.

[0059] The off-axis design avoids central obstruction and reduces energy loss. The 23 times expansion ratio further expands the beam, and the one-level expansion unit cooperates to realize a total expansion ratio of not less than 92 times.

[0060] Then, the light path of the two-level expanded light source is calibrated to form a large-diameter and uniform stable laser spot.

[0061] A high-precision stepper motor (positioning accuracy ±0.001 mm) is provided and connected to the mirror and lens adjustment mechanism of the laser expansion system. Through the real-time feedback of the spot illumination distribution data by the infrared imaging system, when the illumination tilt caused by off-axis is detected (for example, the difference between the edge and center illumination exceeds 10%), the motor drives the lens to make a slight adjustment, and at the same time, a customized gradient filter (transmittance distribution is complementary to the illumination tilt) is used to effectively compensate for the non-uniformity of the spot. Finally, a uniform spot with a diameter of not less than 160 mm is formed at a distance of 2000 mm, and the beam divergence angle is strictly controlled to be not more than 25 μrad, ensuring that the energy distribution is stable when irradiating at a long distance, and providing a uniform laser irradiation basis for the subsequent imaging system to capture the temperature difference between the explosive and the substrate.

[0062] Further, the surfaces of the optical elements are coated with an antireflection film (transmittance ≥99%) or a high-reflectivity film (reflectivity ≥99.5%) for 10.6 μm wavelength to reduce optical loss. At the same time, the base materials of the primary mirror and the secondary mirror are made of fused quartz with low thermal expansion coefficient, and the precise mechanical structure is used to reduce the influence of environmental temperature change on the stability of the optical path and ensure the long-term reliability of the expansion performance.

[0063] The infrared radiation light source is obtained by laser beam expansion, the total beam expansion ratio is greater than or equal to 92 times, and a uniform light spot with a diameter greater than or equal to 160 mm is formed at a distance of 2000 mm, and the beam divergence angle is less than or equal to 25 mu rad.

[0064] S3, based on the laser spot, collect the infrared thermal image of the detected explosive, and synchronously collect the visible light image of the detected explosive.

[0065] Firstly, based on the uniform and stable laser spot, the infrared imaging of the detected explosive is carried out to obtain the infrared thermal image.

[0066] An infrared imaging system is adopted to collect the infrared thermal image of the detected explosive based on the uniform and stable laser spot; the core of the infrared imaging system is a non-cooled vanadium oxide infrared detector (for example: TWIN1212R-GU20A), the number of pixels is not less than 1280*1024, the pixel size is 12*12 mu m, and the working waveband covers 8-14 mu m, so that the temperature distribution difference generated after the explosive absorbs infrared laser can be accurately captured. The effective focal length of the system is 130.9 mm, the F number is 1.97, and the thermal sensitivity (NETD) is less than or equal to 30 mK@f1.0 / 25 DEG C, so that even if the temperature difference between the trace explosive (concentration less than or equal to 1.5 mg / cm2) and the substrate is only a few degrees Celsius within the detection distance of 1-3 m, the explosive can still be clearly imaged.

[0067] A visible light imaging system is adopted to collect the visible light image of the detected explosive; the visible light imaging system selects a high-sensitivity CCD sensor (for example: PIXIS_1024_eXcelon), the number of pixels is not less than 1024*1024, the pixel size is 13*13 mu m, the photosensitive surface size is 13.312*13.312 mm, and the optical system based on the double-Gaussian objective lens design is matched, the effective focal length is 130.9 mm, the F number is 10.07, the visible light details of the detected area can be synchronously acquired, and scene auxiliary information such as sample material and shape is provided for the infrared thermal image.

[0068] Then, the visible light imaging of the detected explosive is synchronously triggered, the visible light image of the detected explosive is collected, the infrared / visible dual-imaging system is synchronously triggered through the control trigger, and the infrared thermal image scene graph is formed.

[0069] The time synchronization of the dual-imaging system is realized by using a high-precision trigger (response time less than or equal to 1 mu s), the trigger signal is transmitted to the shutter control end of the infrared detector and the CCD sensor through a coaxial cable respectively, and the time difference between the image collection of the two is less than or equal to 10 ms. This strict synchronization avoids the spatial misalignment of the infrared thermal image and the visible light image caused by the movement of the detected sample (for example, the conveying belt conveying), and lays a foundation for subsequent image registration and fusion.

[0070] Further, the environmental adaptive adjustment function in the imaging process is improved, the gain can be dynamically adjusted according to the scene temperature, the visible light imaging unit can automatically optimize the exposure time (50~500μs) according to the ambient light intensity, and high signal-to-noise ratio images can be output under different lighting conditions (for example: natural light in the daytime, weak light at night), so that the accurate extraction of temperature difference characteristics by subsequent image analysis is ensured.

[0071] S4, match the infrared thermal image and the visible light image to form an infrared thermal image scene graph, and identify and analyze the infrared thermal image scene graph based on a deep convolutional neural network.

[0072] Firstly, the infrared thermal image and the visible light image are subjected to noise reduction and registration processing to form an infrared thermal image scene graph.

[0073] In view of the characteristic differences between the infrared thermal image and the visible light image, a hierarchical noise reduction strategy can be adopted. For the infrared thermal image, an adaptive median filter algorithm is adopted to dynamically adjust the filter window (3×3 to 7×7) by judging the gray value distribution of the pixel neighborhood, so as to retain the temperature gradient details while reducing the random noise covering the temperature difference characteristics.

[0074] For the visible light image, a method combining Gaussian filtering and bilateral filtering is adopted to smooth the high-frequency noise while retaining the key scene information such as sample edge and texture.

[0075] In the registration processing stage, the feature points (such as sample edge corner points and texture mutation points) of the two types of images are extracted based on the SIFT (Scale Invariant Feature Transform) algorithm, the feature matching is performed through the FLANN (Fast Library for Approximate Nearest Neighbors), the spatial transformation matrix (including translation, rotation and scaling parameters) of the infrared thermal image and the visible light image is calculated, the visible light image is mapped to the infrared thermal image coordinate system, the sub-pixel level registration is realized, the temperature features and scene details after fusion are accurately corresponded, and finally the infrared thermal image scene graph containing temperature distribution and spatial scene is formed.

[0076] Then, based on a deep convolutional neural network, the infrared thermal image scene graph is subjected to image fusion processing, the temperature features of the fused image are extracted, and the infrared thermal image scene graph is subjected to pixel-level segmentation.

[0077] The deep convolutional neural network deployed by the neural network unit takes the DetectNet model as the core architecture, the encoder part is improved based on the VGG-16 network, multi-scale features (from 256×256 pixels to 16×16 pixels) are extracted through 5 layers of convolution and pooling operations, and residual connection is introduced to solve the gradient vanishing problem of deep network; the decoder part adopts deconvolution operation to realize feature up-sampling, combines with dilated convolution (dilation rate 2, 4) to expand the receptive field, and captures the global temperature difference pattern of the explosive and the base.

[0078] The DetectNet model first performs channel-level fusion on the infrared thermal image scene graph, encodes temperature information and visible light texture features into a multi-dimensional feature vector; then focuses on high temperature difference areas (e.g. TNT hot spot area after laser irradiation) through attention mechanism, extracts key features such as temperature gradient, peak temperature, area ratio, etc.; finally outputs a binary mask (explosive area is 1, background is 0) through a pixel-level segmentation head.

[0079] Then, based on the pre-trained explosive class recognition, the segmented image is classified as an explosive. Using a pre-trained class recognition subnetwork (based on an explosive temperature feature database), the TNT, RDX, KNO3, and other explosive classes are identified, with an identification accuracy of ≥80% and a single-frame processing time of ≤200ms, meeting the system's rapid response requirements.

[0080] S5, human-computer interaction is performed through a touch screen, real-time images, detection results, and device status are displayed, when a suspicious object is detected, an alarm mechanism is triggered, and detection logs are recorded.

[0081] Human-computer interaction can use a 10.1-inch industrial-grade touch screen (resolution 1920x1200), with dustproof and anti-glare features, suitable for complex lighting environments such as passenger transport stations.

[0082] The touch screen can display the interface in four quadrants: the upper left is the real-time flow of the infrared thermal image, with a temperature distribution heat map and the highest / lowest temperature values marked; the upper right is the visible light image, synchronously displaying the appearance details of the detected sample; the lower left is the infrared thermal image scene graph after fusion analysis, with a red box marking the suspicious area and confidence (e.g. TNT, 92%); the lower right is the device status panel, which updates the laser power (current value / set value), laser temperature, imaging system status, remaining storage space, and other key parameters in real time, allowing operators to monitor system operation in real time.

[0083] Further, the alarm mechanism uses a multi-level response design. When the deep convolutional neural network identifies a suspicious object (confidence ≥80%), a red warning box is first popped up on the touch screen, triggering a built-in buzzer and an external red LED warning light simultaneously, lasting for 10 seconds or until the operator manually confirms; if no confirmation is received within 3 seconds, the system automatically pushes the alarm information (including time, location, and suspicious object type) to the associated security terminal (e.g. security personnel's handheld PAD), ensuring rapid response.

[0084] The detection log adopts a distributed storage architecture, each log contains a timestamp, a sample number, an infrared / visible light image thumbnail, a detection result (including a type and a confidence), a device state parameter, and the like, and is stored in an encrypted format in a local solid state disk, while a backup file is generated daily and can be exported through a USB interface or uploaded remotely to a management server. The log query function supports multi-dimensional screening according to time, sample number, detection result, and the like, facilitating later tracing and statistical analysis.

[0085] Further, the method also integrates a permission management function, which distinguishes two levels of permissions of administrators and operators: the administrator can log in through a password to perform advanced operations such as laser power upper limit setting, alarm threshold adjustment, and system parameter calibration; and the operator can only view real-time data, confirm alarms, and export logs to prevent misoperation from affecting system stability. This design not only ensures convenient operation, but also improves the safety of device operation through strict permission control.

[0086] Further, the neural network unit also supports online learning function, which can receive artificial correction results such as false / missed report labels through human-computer interaction, update network parameters using small batch gradient descent algorithm, and continuously optimize recognition accuracy in complex scenes such as multiple interferents and low concentration explosives.

[0087] Further, the system also includes data transmission, which transmits infrared / visible light image data in real time based on data transmission and reduces transmission bandwidth by using a compression algorithm.

[0088] As a key link in the system data flow, data transmission adopts a gigabit Ethernet interface to realize high-speed data interaction between the infrared imaging unit, the visible light imaging unit, and the image analysis.

[0089] Its core function is to transmit infrared thermal images (for example, 1280×1024 pixels, 16-bit grayscale values) and visible light images (for example, 1024×1024 pixels, 24-bit RGB) output by the collected imaging in real time to the backend processing unit, the transmission frame rate is synchronized with the imaging system (≥25 fps), and the time continuity of image data is ensured.

[0090] Further, to solve the bandwidth pressure caused by high-definition image transmission, H.265 / HEVC video compression algorithm is integrated, through techniques such as intra-frame prediction, inter-frame motion compensation, and entropy encoding, the data volume of a single infrared image is compressed from about 3MB to 0.8MB, and the data volume of a single visible light image is compressed from about 3MB to 0.6MB, the total bandwidth occupancy is controlled within 50Mbps, and the analysis delay caused by data congestion is avoided.

[0091] Meanwhile, the method also has data checking and retransmission mechanism, integrity of each frame image data is checked through Cyclic Redundancy Check (CRC), when data packet loss or error is detected, retransmission request is automatically sent to the acquisition imaging system, and transmission accuracy is ensured to be greater than or equal to 99.9%.

[0092] Further, data transmission adopts an encryption protocol, so that image data is prevented from being tampered with or leaked in the transmission process, and security of security check information is ensured.

[0093] The application further provides an electronic device, Figure 2 A structural schematic diagram of the electronic device provided by the embodiment of the application is shown in the figure, Figure 2 As shown in the figure, the electronic device can include a processor, a communications interface, a memory and a communications bus, wherein the processor, the communications interface and the memory complete mutual communication through the communications bus. The processor can call logical instructions in the memory, for example, to execute the following method:

[0094] S1, generating a stable infrared radiation light source;

[0095] S2, performing secondary beam expansion on the infrared radiation light source to obtain a uniform and stable laser spot;

[0096] S3, collecting an infrared thermal image of the detected explosive based on the laser spot, and synchronously collecting a visible light image of the detected explosive;

[0097] S4, matching the infrared thermal image and the visible light image to form an infrared thermal image scene graph, and performing recognition analysis on the infrared thermal image scene graph based on a deep convolutional neural network.

[0098] In addition, the logic instructions in the above-mentioned memory can be realized in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0099] The embodiments of the present application also provide a non-transitory computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method provided by the above-mentioned embodiments, for example: comprising:

[0100] S1, generating a stable infrared radiation light source;

[0101] S2, performing secondary beam expansion on the infrared radiation light source to obtain a uniform and stable laser spot;

[0102] S3, collecting an infrared thermal image of the detected explosive based on the laser spot, and synchronously collecting a visible light image of the detected explosive;

[0103] S4, matching the infrared thermal image and the visible light image to form an infrared thermal image scene graph, and identifying and analyzing the infrared thermal image scene graph based on a deep convolutional neural network.

[0104] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or distributed on a plurality of network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment scheme. Those skilled in the art can understand and implement it without creative labor.

[0105] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0106] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A non-contact laser explosive detection method, characterized by, The method comprises the steps of: S1, generating a stable infrared radiation light source; S2, performing secondary beam expansion on the infrared radiation light source to obtain a uniform and stable laser spot; S3, collecting an infrared thermal image of the detected explosive based on the laser spot, and synchronously collecting a visible light image of the detected explosive; S4, matching the infrared thermal image and the visible light image to form an infrared thermal image scene graph, and performing recognition analysis on the infrared thermal image scene graph based on a deep convolutional neural network.

2. A non-contact laser explosive detection method according to claim 1, wherein, The S1 of generating a stable infrared radiation light source comprises the steps of: S11, controlling the power of the CO2 laser through a PLC to generate a continuous / pulsed infrared radiation light source with stable power; S12, using a frequency controller to control the pulse frequency of the infrared radiation light source to obtain an infrared radiation light source with stable frequency; S13, using a water-cooled machine to adjust the working temperature of the laser in real time to obtain an infrared radiation light source with stable temperature.

3. The non-contact laser explosive detection method of claim 1, wherein, The S2 of performing secondary beam expansion on the infrared radiation light source to obtain a uniform and stable laser spot comprises the steps of: S21, using an inverted Galileo telescope system to perform primary beam expansion on the light source to obtain a primary beam expansion light source with a magnification of not less than 4 times; S22, using an off-axis Cassegrain telescope system to perform secondary beam expansion on the primary beam expansion light source to obtain a secondary beam expansion light source with a magnification of not less than 23 times; S23, performing optical path calibration on the secondary beam expansion light source to form a large-diameter and uniform and stable laser spot.

4. The non-contact laser explosive detection method of claim 1, wherein, The infrared radiation light source forms a uniform and stable laser spot with a total beam expansion ratio of not less than 92 times, and a uniform and stable spot with a diameter of not less than 160 mm is formed at a distance of 2000 mm, and the beam divergence angle is not less than 25 μrad.

5. The non-contact laser explosive detection method of claim 1, wherein, The S3 comprises the steps of: S31, based on the uniform and stable laser spot, performing infrared imaging on the detected explosive to obtain an infrared thermal image; S32, synchronously triggering visible light imaging of the detected explosive to collect a visible light image of the detected explosive.

6. The non-contact laser explosive detection method according to claim 1, wherein: The infrared thermal image has a pixel number of not less than 1280x1024, and the working waveband covers 8-14 μm; The visible light image has a pixel number of not less than 1024x1024; The time difference between the infrared / optical dual-image collection is not more than 10 ms.

7. The non-contact laser explosive detection method of claim 1, wherein, The S4 comprises the steps of: S41, performing noise reduction and registration processing on the infrared thermal image and the visible light image to form an infrared thermal image scene graph; S42, based on a deep convolutional neural network, performing image fusion processing on the infrared thermal image scene graph, extracting the temperature features of the fused image, and performing pixel-level segmentation on the infrared thermal image scene graph; S43, based on a pre-trained explosive category recognition, performing explosive category discrimination on the segmented image.

8. The non-contact laser explosive detection method of claim 1, wherein, The deep convolutional neural network adopts a DetectNet model.

9. The non-contact laser explosive detection method of claim 1, wherein, The method further comprises the step of: S5, performing human-computer interaction through a touch screen to display real-time images, detection results and device states, triggering an alarm mechanism when a suspicious object is detected, and recording detection logs.

10. The non-contact laser explosive detection method of claim 9, wherein, The alarm mechanism adopts multi-level response, including: through the touch screen pop-up red warning box, synchronous trigger built-in buzzer and external red LED warning light, last 10 seconds or until the operator manual confirmation; If not confirmed within 3 seconds, the system automatically pushes the alarm information to the associated security terminal.

Images