Explosion-proof full-color low-light night vision camera for mine

Abstract

The invention discloses an explosion-proof full-color low-light night vision camera for mines, which comprises an explosion-proof shell, an explosion-proof lens, an image sensor, an image processing module and an image transmission module, and is characterized in that the image processing module comprises an image acquisition module used for acquiring multiple frames of original low-light images in a low-illumination environment based on an enhanced image sensor; the image noise processing module is used for processing the multiple frames of original low-light images through a noise suppression algorithm so as to remove noisy points in the images; the image color processing module is used for generating a full-color night vision image which is accurate in color and rich in details based on an image fusion and color compensation technology; and the image optimization module is used for performing real-time optimization processing on the full-color night vision image to improve the frame rate and the recognition effect of the target in the low-contrast environment. The camera can solve the problem that a high-definition full-color picture cannot be displayed even if light supplement is adopted in a night vision environment.

Description

Technical Field

[0001] The present invention relates to the technical field of cameras, and in particular to an explosion-proof full-color low-light-level night vision camera for use in mines. Background Art

[0002] In dangerous places such as mines, the requirements for monitoring equipment are extremely high, requiring such cameras to have functions such as explosion-proof, night vision, and high definition. Although existing night vision cameras use fill light to increase the camera's vision, it is difficult to achieve high-definition and clear images. Summary of the invention

[0003] The purpose of the present invention is to provide a flameproof full-color low-light night vision camera for use in mines to overcome the above-mentioned deficiencies in the prior art.

[0004] To achieve the above technical objectives, the technical solution of the present invention is implemented as follows:

[0005] A full-color low-light-level night vision imaging system, comprising:

[0006] An image acquisition module is used to acquire multiple frames of original low-light images in a low-light environment based on an enhanced image sensor;

[0007] An image noise processing module, used for processing the above-mentioned multiple frames of original low-light images by using a noise suppression algorithm to remove noise points in the images;

[0008] Image color processing module, used to generate full-color night vision images with accurate colors and rich details based on image fusion and color compensation technology;

[0009] The image optimization module is used to improve the frame rate and the recognition effect of the target in a low-contrast environment by performing real-time optimization processing on the full-color night vision image.

[0010] Furthermore, the image color processing module includes:

[0011] An extraction module, used for extracting high-frequency components and low-frequency components of each multi-frame original low-light image;

[0012] A combining module, used for combining the high-frequency components of the multiple frames of original low-light images with a low-frequency component of a reference image;

[0013] Color correction module, used to perform color correction operations to improve the color saturation and brightness level of the image, and automatically adjust the color balance according to the lighting conditions of the environment;

[0014] The preliminary color formation module integrates high-frequency details and low-frequency contour information based on multi-scale fusion technology to form a preliminary color map.

[0015] Furthermore, the extraction module includes:

[0016] Transformation module, used to perform wavelet transformation on each single frame to obtain different resolution levels;

[0017] Enhancement module, used to improve the image detail level based on contrast enhancement operator and enhance the visibility of weak signal areas;

[0018] The weight assignment module is used to assign different weights according to the strength of features at each level;

[0019] The pixel intensity enhancement coefficient setting module sets the corresponding pixel intensity enhancement coefficient α=75*TH / current layer mean value σ when the multi-frame average signal intensity value is higher than the predefined threshold TH;

[0020] Where TH represents the signal threshold, σ represents the signal-to-noise ratio of the current layer of the image, and TH represents the reference standard set in advance by the user.

[0021] Furthermore, the combined module includes:

[0022] An arrangement module, used for arranging low-frequency component results of multiple frames in time sequence;

[0023] A homogenization processing module is used to perform homogenization processing on each frame in the time domain using a smoothing algorithm to eliminate inter-frame fluctuations;

[0024] An overlay module is used to use a reference low-frequency template image as an overlay base map to keep the background stable;

[0025] An alignment module, used to ensure precise alignment between components based on spatial registration;

[0026] The conditional expression is: (T_α*T_β) / D>θ. When this inequality holds, the exact matching is considered to be completed. T_α represents the time alignment factor; T_β represents the matching index after spatial transformation; D represents the data dimension difference; θ is the system error limit constant, and the default value is determined by the application scenario requirements to ensure the consistency and comparability of the final result image.

[0027] Furthermore, the color correction module includes:

[0028] A color mapping module, used to configure a color lookup table to map to the real environment color gamut to improve color reproduction accuracy;

[0029] Color balance module, used to adaptively adjust the color balance value of each color area in the image to ensure smooth transition of natural tones;

[0030] A color stabilization module that applies feedback control mechanisms to maintain stable viewing under changing lighting conditions;

[0031] A color gain module calculates a color saturation gain coefficient CS based on a color temperature and a contrast level adjustment coefficient.

[0032] If the ambient color temperature Temp_C < Temp_STD and the contrast CD is lower than the minimum acceptable contrast level, then set CS = (Min_Contrast + (Min_Brightness) * Max_Saturation) / Temp_STD;

[0033] Where: Min_Contrast is the bottom line of the picture contrast threshold expected by the user; Min_Brightness is the lower limit of the picture brightness most desired to be achieved; Max_Saturation is the maximum colorfulness limit allowed; Temp_STD is the standard laboratory ambient light source.

[0034] Furthermore, the color preliminary formation module includes:

[0035] A model construction module constructs different image detail expression models with a multi-level pyramid.

[0036] A model enhancement module enhances the details of the edge region using a Laplacian sharpening strategy.

[0037] A color filling module combines a color mapping algorithm to add details to the contour to form color filling.

[0038] A significant level setting module determines the significant level of each local part of the image based on the quality score matrix F of the fused image, the image block complexity evaluation index C, and the block position identifier P.

[0039] The decision rule is that if F(x,y) * C(z) > Threshold, then set P[(x + y) / 2] = [z * 5 + F(x,y) * 1] / (C(z) * 1.25);

[0040] Here, Threshold is a fixed threshold, x,y are the pixel index points in the matrix F, z represents the matrix block number, F(x,y) measures the detail intensity level of a specific coordinate, and C() measures the diversity of features within the image block.

[0041] A flameproof full-color low-light night vision camera for mines includes a flameproof housing, a flameproof lens, a high-definition image sensor, an image transmission module, an intelligent monitoring module, and the full-color low-light night vision imaging system as described above. The full-color low-light night vision imaging system is electrically connected to the high-definition image sensor and the image transmission module respectively, and the intelligent monitoring module is electrically connected to the high-definition image sensor and the remote intelligent terminal respectively.

[0042] Preferably, the material of the explosion-proof housing is 304 stainless steel, the material of the explosion-proof lens is special explosion-proof glass coated with nano invisible wipers, and the image sensor is a 1 / 1.8 high-definition sensor.

[0043] A full-color low-light-level night vision imaging method, the method comprising:

[0044] Acquire multiple frames of original low-light images in low-light environments based on enhanced image sensors;

[0045] Processing the multiple frames of original low-light images by a noise suppression algorithm to remove noise in the images;

[0046] Generates full-color night vision images with accurate colors and rich details based on image fusion and color compensation technology;

[0047] The frame rate and the recognition effect of the target in a low-contrast environment are improved by performing real-time optimization processing on the full-color night vision image.

[0048] Furthermore, the method comprises:

[0049] Acquire multiple frames of original low-light images in low-light environments based on enhanced image sensors;

[0050] Processing the multiple frames of original low-light images by a noise suppression algorithm to remove noise in the images;

[0051] Perform wavelet transform on each single frame to obtain different resolution levels;

[0052] Improve the image detail level based on the contrast enhancement operator and enhance the visibility of weak signal areas;

[0053] Assign different weights according to the strength of features at each level;

[0054] When the multi-frame average signal strength value is higher than the predefined threshold TH, the corresponding pixel intensity enhancement coefficient α is set to 75*TH / current layer average σ.

[0055] Where TH represents the signal threshold, σ represents the signal-to-noise ratio of the current layer of the image, and TH represents the reference standard set in advance by the user;

[0056] Arrange the low-frequency component results of multiple frames in time order;

[0057] Use smoothing algorithm to perform homogenization processing on each frame in the time domain to eliminate inter-frame fluctuations;

[0058] The reference low-frequency template image is used as an overlay base map to keep the background stable;

[0059] Based on spatial registration, we ensure precise alignment between components.

[0060] Its conditional expression is: (T_α * T_β) / D > θ. When this inequality holds, it is considered that the precise pairing is completed. Here, T_α represents the time alignment factor; T_β represents the matching index after spatial transformation, D represents the data dimension difference, and θ is the system error bound constant. The default value is determined by the application scenario requirements to ensure the consistency and comparability of the final result image;

[0061] Configure the color lookup table to map to the real - environment color gamut to improve the color - restoration accuracy;

[0062] Adaptive - adjust the color - balance values of each color area in the image to ensure a smooth transition of natural tones;

[0063] Apply a feedback control mechanism under changing lighting conditions to maintain a stable visual perception;

[0064] Calculate the color - saturation gain coefficient CS based on the color - temperature and contrast - level adjustment coefficients,

[0065] If the environmental color temperature Temp_C < Temp_STD and the contrast CD is lower than the minimum acceptable contrast level, then set CS = (Min_Contrast+(1 - Min_Brightness)*Max_Saturation) / Temp_STD,

[0066] where: Min_Contrast is the user - expected picture - contrast threshold bottom line, Min_Brightness is the minimum picture - brightness limit that is most desired to achieve, Max_Saturation is the maximum color - richness limit allowed, and Temp_STD represents the standard laboratory - environment light source;

[0067] And automatically adjust the color balance according to the lighting conditions of the environment;

[0068] Construct different image - detail - expression models with multi - level pyramids;

[0069] Use the Laplacian sharpening strategy to enhance the details of the edge area;

[0070] Combine the color - mapping algorithm to add details to the contour to form color filling;

[0071] Determine the significant level of each local part of the image based on the quality - score matrix F of the fused image, the image - block complexity evaluation index C, and the block - position identifier P,

[0072] The decision rule is that if F(x,y)*C(z)>Threshold, then set P[(x + y) / 2]=[z*5+F(x,y)*1] / (C(z)*1.25),

[0073] Here, Threshold is a fixed threshold, x, y are the pixel index points in the matrix F, z represents the matrix block number, F(x, y) measures the level of detail intensity at a specific coordinate, and C() measures the diversity of features within the image block;

[0074] The frame rate and the recognition effect of the target in a low-contrast environment are improved by performing real-time optimization processing on the full-color night vision image.

[0075] The beneficial effects of the present invention are as follows: the present invention can solve the problem of image noise caused by insufficient light in a dim environment; the problem of inaccurate color reproduction caused by complex light in low-light scenes; the problem of missing color information due to weak dim light signals under full-color display requirements; the problem of difficulty in target recognition due to low contrast at night or in dim conditions; and the problem of decreased image frame rate due to large data processing volume during real-time imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a schematic diagram of the functional module structure of a flameproof full-color low-light-level night vision camera for use in mines according to an embodiment of the present invention;

[0077] Figure 2 This is a specific application picture of the explosion-proof full-color low-light night vision camera for mines according to the embodiment of the present invention;

[0078] Figure 3 This is a design model diagram of a flameproof full-color low-light night vision camera for use in mines according to an embodiment of the present invention;

[0079] Figure 4 This is the effect picture taken by the existing fill light camera;

[0080] Figure 5 This is a picture showing the effect of the explosion-proof full-color low-light night vision camera for mines of the present invention taking pictures at night. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the present invention are described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0082] like Figure 1As shown, according to an embodiment of the present invention, a flameproof full-color low-light level night vision camera for use in mines is disclosed, comprising an explosion-proof housing, an explosion-proof lens, an image sensor, an image processing module and an image transmission module, wherein the image processing module comprises: an image acquisition module, for acquiring multiple frames of original low-light level images in a low-light environment based on an enhanced image sensor; an image noise processing module, for processing the multiple frames of original low-light level images through a noise suppression algorithm to remove noise in the image; an image color processing module, for generating a full-color night vision image with accurate color and rich details based on image fusion and color compensation technology; and an image optimization module, for improving the frame rate and the recognition effect of the target in a low-contrast environment by performing real-time optimization processing on the full-color night vision image.

[0083] The present invention proposes solutions to various problems in image quality and imaging performance in existing low-light imaging systems. It mainly includes obtaining original low-light images in low-light environments, improving the quality of these images through denoising algorithms to eliminate noise interference, using a variety of means to restore colors and rich details to generate clear and natural-looking night vision images, and real-time enhancement of the full-color images after imaging to ensure that the ideal frame rate and better recognition can be achieved even in low-light conditions. The following is a detailed description of the process.

[0084] First, the developed image capture sensor device with improved sensitivity continuously collects several images of the dark environment. The design principle of this image sensor is to adapt to low light conditions. It can be more sensitive to the weak light source information in the surroundings, while reducing the impact of background noise such as infrared rays, to ensure that the information obtained is accurate and maintains the original state as much as possible. For example: In one implementation case, the technology uses a small camera with a sensor installed during the tracking of wild animals in the forest at night. It can obtain continuous video records of animal movements with almost no artificial light supplement; because the forest environment itself is insufficiently lit and changeable, traditional camera equipment may encounter great challenges in such occasions; however, by using this sensor, even in very harsh conditions, it can ensure that the received signal has high clarity and credibility, laying a solid foundation for the subsequent noise removal steps.

[0085] Then, adaptive multi-frame fusion technology is used to reduce the noise in the above sequence of photos. The core advantage of adaptive filtering technology is that it can adjust the filtering parameters according to the characteristics of different areas in each frame; therefore, it can more accurately and efficiently locate and remove unnecessary content such as spots, stripes, block shadows and other common interference factors, thereby significantly improving the clarity and making observation more convenient, simple, fast and accurate. For example, when a large feline carnivore is monitored in a wild reserve late at night and passes in front of a monitoring device equipped with such an automatically adjustable lens; if there is no such process, then the things captured may only be a bunch of unclear and worthless blurry images that cannot identify the target's identity information; but if our solution is used, the details of the outer contour lines, texture, texture and color of the animal's body can be retained while making the hair and skin color and texture characteristics appear extremely real and natural, allowing species experts to make accurate judgments based on these precious and hard-to-obtain data.

[0086] Then, in order to obtain the same color visual effect as during the day under dim light, subsequent operations are required - color enhancement and mixing, so as to make up for the problem of color information loss caused by insufficient light in low-light environments and give the overall picture a stronger three-dimensional level of richness and other visual aesthetic effects. The main working logic of this program is to refer to the standard color table library to find the color number with the highest matching degree and fill it into the corresponding pixel unit to achieve a realistic reproduction. At the same time, it will also adjust the point, line, shape layout and the relationship between the cold and warm tones according to factors such as the position and distance of the foreground and background objects to create a visual atmosphere close to the real world. For example: when a night ranger is watching his jurisdiction through a camera monitor equipped with the above technology, he suddenly sees a suspected man-made illegal crossing activity, but he can't see the face of the intruder. At this time, if the device in ordinary black and white mode is difficult to distinguish the specific situation, it may even miss key information and cause false alarms and distortion; but when we use the color enhancement function, it can be easier to identify the color style of the person's clothes and even some facial details, so as to quickly respond to the incident and prevent unnecessary damage to forest resources.

[0087] Finally, we have one last hurdle to overcome, which is to improve the overall operating efficiency of the system, especially when processing a large amount of data, we must prevent the image transmission frame rate from slowing down to avoid stuttering and intermittent playback, which affects the user experience and practical experience. This stage mainly relies on software algorithm design to achieve a technical solution that accelerates the computing process, increases processing speed, reduces resource consumption, and achieves stable and efficient output. For example, when a natural disaster occurs in a certain place, search and rescue dogs need to be dispatched to assist in searching for survivors, but time is urgent and no time is allowed to waste. At this time, if we directly pass the video material back, it may become abnormally slow due to network transmission problems and cannot meet the requirements of timely communication; however, when the optimization program is enabled, the processing time can be greatly compressed, so that the search and rescue team can quickly grasp the latest on-site dynamic information and find the trapped as soon as possible to increase the success rate of treatment.

[0088] Next, the present invention describes the generation of full-color night vision images with accurate colors and rich details based on image fusion and color compensation technology. The process of extracting high-frequency components and low-frequency components of each multi-frame original low-light image is carried out by separating the collected continuous or non-continuous multi-frame night images into two parts using a specific algorithm: one part represents the details in the image, that is, high-frequency information, such as boundaries, textures, and subtle differences; the other part represents the more vague global information or low-frequency components, which contain the main outline or overall appearance of the object but do not have the accuracy of details. In one embodiment, when a camera is used to shoot a person walking under dim street lights at night, the sequence of images shot from multiple perspectives and times will first be decomposed into two components, the former is used to capture information on subtle differences such as clothing patterns or facial lines, while the latter focuses on recording changes in the overall appearance of the human body.

[0089] The subsequent step mentions combining the high-frequency components of the multiple frames of original low-light images with the low-frequency components of a reference image. This step aims to integrate the various image information obtained in the above processing process to create a more comprehensive, balanced, and highly recognizable final result. This can be achieved through a weighted averaging or selective fusion strategy. Specifically, continuing with the above example, during the synthesis process, the image of a person walking will retain the most prominent textures such as any logos on clothing, and will be embedded in a photo with the most stable outline.

[0090] The next step is to perform color correction to increase the color saturation and brightness level of the image, and automatically adjust the most natural color balance ratio according to the set or perceived light conditions, so that the generated picture can be visually closer to the effect when viewed in daylight. In order to ensure the authenticity of the effect and the satisfaction of the perception, the tone of each pixel can be enhanced and corrected by establishing a linear model or a lookup table. If the previous night scene video diagram is continued to this point, then after this step is completed, you can see that not only the dynamic effect of the pedestrians is completely preserved, but the shadows produced under the light become clearer and clearer, and the true color of the clothes is restored as much as possible.

[0091] Based on multi-scale fusion technology, integrating high-frequency details and low-frequency contour information to form a preliminary color image is a summary and refinement process of all the above work. This process involves using mathematical tools such as pyramids or Laplace transforms to analyze the data that have been improved by the previous process at different scales, and finally merge a whole high-quality full-color night imaging result as the output of the entire system operation. Such a set of program design not only greatly improves the usable range and efficiency level of the night monitoring system, but also has extremely broad practical value and application prospects in important fields such as military detection and civil security. In the framework of the above example, it is assumed that this final step will make the image of pedestrians and environmental factors appear on the screen together, and give the screen a more intuitive and realistic expression. However, it should be noted that in this process, there is no need to list specific formulas or formula parameter settings in detail.

[0092] Next, the invention describes the extraction of high-frequency components and low-frequency components of each multi-frame original low-light image. This process mainly includes the steps of performing wavelet transform on each single frame to obtain different resolution levels, improving the image detail level based on the contrast enhancement operator to enhance the visibility of the weak signal area, and allocating different weights according to the feature strength of each level.

[0093] The first step is to perform a wavelet transform on each single frame to obtain information at different resolution levels. As a multi-resolution analysis tool, wavelet transform is widely used in the field of image processing to decompose low-frequency and high-frequency components in an image. In this process, each image is converted from a spatial domain to a time-frequency representation containing details of multiple scales and directions by using a discrete wavelet transform (DWT). For example, in one embodiment, if there is a sequence of image frames of mice nocturnal activities recorded under low light conditions, these frames need to be converted into information blocks at multiple resolution levels through a specific basis (such as db4 wavelet) to better retain texture details.

[0094] Next, a contrast enhancement operator is used to improve the perceptibility of details in the image, especially in areas with low light levels. This step aims to make the image more clearly visible if it is not clear due to poor lighting conditions. For example, if the outline of a mouse moving in a dark environment is blurred in some video frames, the outline can be made clearer by applying a local linear contrast algorithm or other schemes suitable for enhancing signal details in low light conditions.

[0095] Then, the corresponding weights are assigned according to the strength of the features at each level for further processing. This means that data in different frequency bands will be strengthened or suppressed to varying degrees according to the importance of their respective characteristics. For example, in the case of mouse video analysis mentioned earlier, since low-frequency information often contains a larger proportion of object contours and main parts while high-frequency parts carry more detailed features, it may be possible to give low frequencies a greater weight to ensure that the overall shape is not distorted, while using smaller weights to process higher frequency bands to enhance the perception of local subtle changes.

[0096] When the overall average brightness of multiple images after the above processing and comprehensive consideration is higher than the predetermined standard value (TH), a mathematical relationship will be used to adjust the corresponding pixel brightness magnification parameter, the so-called intensity increase coefficient α. The formula shows that the coefficient α is calculated by taking 75 times TH and then dividing the result by the signal-to-noise ratio σ actually detected at this level; TH refers to the reference threshold value set by the system developer as a judgment on whether the image is too dim to be well observed. Assuming that the preset signal threshold value is 50 units of grayscale and the signal-to-noise level of the specific sub-interval obtained through multi-level analysis is 25 units of grayscale, in this context, 75*50 / 25 will be used to obtain a 3:1 magnification as the basis for subsequent processing. The purpose of setting this ratio is to automatically adapt to different background brightness to ensure that key objects can be seen even when the background is extremely dark, while avoiding unnecessary enhancement of those areas that are already bright enough to maintain the authenticity of the image quality.

[0097] Next, the present invention is described to combine the high-frequency components of the multiple frames of original low-light images with the low-frequency components of a reference image. The method first involves arranging the low-frequency component results of multiple frames in chronological order. In this step, all processed low-frequency component frames are sequenced according to the capture order for subsequent time domain processing. This operation ensures that each frame of the image enters the processing pipeline in the correct order. For example, in one embodiment, after a night camera device captures multiple consecutive pictures, the low-frequency parts of the static elements (such as trees and buildings) in each picture are extracted and sorted according to the chronological order of these images.

[0098] Next, a smoothing algorithm is used to homogenize each frame in the time domain to eliminate fluctuations between frames, that is, to remove or reduce the image noise caused by camera shaking or small movements of the subject. The specific smoothing algorithm can be implemented in many ways, such as mean filtering or Gaussian blur technology. The key to this step is to maintain the continuity of the sequence, improve the overall clarity of the image, and make the separation of dynamic and static content more obvious and effective. Specifically, in a scene where a vehicle is continuously shot slowly passing in front of the camera, after such smoothing, even if the vehicle is moving slowly, the image noise effect caused by the vehicle will be reduced.

[0099] Afterwards, the baseline low-frequency template image is set as the overlay background image to ensure the consistency of the background environment throughout the imaging process and maintain the stability of the image without changing with the appearance of dynamic targets. The low-frequency background can be understood as the low-frequency component of the first captured image or other pre-selected reference images. This setting makes the basic level of the entire image more stable and unchanged, thereby highlighting the high-frequency dynamic details. For example, if there is a constantly flashing road sign as one of the backgrounds in the night environment, this part should be fixed as a static object as much as possible when generating the final image.

[0100] Finally, based on the consideration of spatial position relationship, the alignment between various components is guaranteed, and a mathematical criterion is introduced to confirm the accuracy of the final splicing. Its conditional expression is (T_α*T_β) / D>θ. Among them, T_α is used to evaluate the temporal alignment level between image segments, and the ideal value is close to 1; T_β calculates the fit index of the image after spatial geometric changes, and the value tends to approach the full score of 1 for perfect overlap; the variable D measures the deviation in size specifications between different layers, and the threshold θ is the error limit pre-set within the system, usually ranging from 0 to 0.1. The specific selection value depends on the specific visual application scenario and the expected effect. By setting such a threshold check standard, it is ensured that the coordinated synchronization state of all image elements meets the expected minimum error requirements before the actual combination operation. For example, when the monitoring equipment monitors the flashing of forest fires, the application of this mathematical judgment helps to accurately capture the instantaneous bright spots in the fire diffusion path, and prevents the occurrence of image distortion and dislocation due to subtle differences in time or space in the process of superimposing background images, ensuring the authenticity of the final generated composite image.

[0101] Next, the color correction operation of the present invention is described to enhance the color saturation and brightness level of the image.

[0102] First, configure the color lookup table (Color LookUp Table). This step is to map the colors in the image to a color gamut that is closer to the real environment, thereby improving the accuracy of color reproduction and reducing the deviation caused by the limitations of the sensor or imaging system itself. Specifically, in one embodiment, a nighttime outdoor surveillance video image obtained by a full-color low-light night vision imaging system may contain a large number of blue and purple dark areas and some dim green vegetation images. Due to the weak light during actual nighttime shooting, the plants in these images will appear to lack vitality and are not bright enough. At this time, a specific color lookup table configuration can make the green leaves and flowers appear closer to the bright state seen in the real world during the day when the image is displayed, while reducing the excessive impact of blue tones. The color lookup table generally includes multiple one-to-one matches from the collected colors to the processed color outputs. These data points are pre-calibrated and can effectively improve the performance of the entire color space.

[0103] Next, the color balance values ​​of each color area in the image are adaptively adjusted to make the transition between colors smooth without causing unnatural colors due to over-correction. For example, in a city street scene taken at night, the illuminated area is usually yellowish or incandescent, while the dark area in the distance may be more blue-gray. By using differentiated adaptive calibration methods for these different areas, the transition between each color segment can appear continuous and natural.

[0104] The third step is to apply a feedback control strategy to cope with changes in imaging effects caused by changes in ambient lighting. For example, if the camera is quickly moved to a place where the light intensity increases or decreases significantly (a rapid change from street lighting to a dark background), in order to ensure that the overall visual experience remains relatively consistent in both bright and dark fields, this mechanism can dynamically correct the color intensity and brightness level. Feedback control usually involves detecting changes in light intensity and then adjusting key parameters in the subsequent correction process based on actual conditions.

[0105] Finally, based on the measured or assumed temperature chromaticity (Temperature ColorTone) of the current ambient light, the corresponding color saturation gain coefficient (CS) is calculated and set to enhance the color contrast. If the currently measured color temperature value (Temp_C) is lower than the standard set point (Temp_STD) (which may be D65 daylight color temperature or other universal reference colorimetry), and it is found that the contrast level in the overall scene is also in a relatively poor range, the color saturation factor is updated using the following expression: CS = (Min_Contrast + (1-Min_Brightness) * Max_Saturation) / Temp_STD. Here, Temp_STD can be understood as the idealized standard ambient light source color temperature (for example, D65 color temperature is approximately equal to 5000 to 6500K); the Min_Brightness parameter reflects the lower limit of the lowest brightness of the picture that the operation expects to achieve, which may be a ratio between 5% and 50% in some applications. In order to achieve a satisfactory display, it is generally set above 30%. Similarly, Min_Contrast represents the minimum acceptable picture clarity (or contrast), ranging from 0.5 to 1.5; Max_Saturation is the highest degree of color richness allowed, which determines the upper limit of the enhancement effect, and is ideally between 0.8 and 1. This design allows the color expression to be enhanced but not excessive, especially in a cooler light environment, providing a better viewing experience.

[0106] In this way, high-quality night vision picture output with rich colors and appropriate brightness levels can be obtained even in low-light conditions.

[0107] Next, the present invention is described to integrate high-frequency details and low-frequency contour information to form a preliminary color image based on multi-scale fusion technology. Multi-level pyramids construct different image detail expression models, which refers to using a multi-level pyramid structure (such as a Gaussian pyramid or a Laplacian pyramid) to process the input low-light image, decompose the original image at different scales, and obtain a contour image containing low-frequency information and a high-frequency information image that retains detail texture features. This structure can better characterize multiple characteristics of the same image from multiple scales, so that subsequent processing can better maintain and restore the level of details.

[0108] Using the Laplace sharpening strategy to enhance edge area details is to apply the Laplace sharpening operation on the basis of extracting high-frequency detail images to highlight the contrast at the object boundary and increase the expressiveness of subtle features. Enhancing the boundary information of the image through the Laplace operator can make the contour appear clearer. This step improves the recognition of small targets or tiny features in the image, which is particularly important for low-light night imaging, because such images usually have a blurred effect due to limited light sources.

[0109] Adding details to the outline to form a color fill in combination with a color mapping algorithm involves fusing a sharpened high-resolution detail map with coarse low-resolution outline information containing basic shapes and structures, while introducing color components. This is particularly effective in night vision system applications because it provides more realistic and understandable information to the observer. In this way, data obtained even in low-light conditions can be converted into a form that can be directly interpreted by the human eye, thereby improving operability and accuracy.

[0110] Determining the local saliency level of each image based on the quality score matrix F of the fused image, the image block complexity evaluation index C, and the block position identifier P means evaluating the quality of each block in the image (marked by the identifier P) based on two aspects: first, the matrix F records the clarity (or the presence of edges and details) at each pixel position (i.e., x, y position), and second, the function C measures the degree of change (complexity or diversity) of the visual content of each interval. If the calculation result exceeds a set threshold (Threshold), the block is considered to be more significant and its importance is emphasized. This threshold is a key control point preset by humans, and may be appropriately adjusted according to the background noise level and other external factors in actual application scenarios.

[0111] Specifically, in a night vision system, when trying to reconstruct a low-light photograph, in one embodiment, it is assumed that Threshold is set to 15 (the numerical value selection needs to be adjusted according to actual conditions, and the possible values ​​are between 0 and 255, depending on the specific 8-bit color depth); x and y in the matrix F are 100 and 20 respectively, indicating that the clarity of the point is a higher value; the block number Z (corresponding to the position of the matrix block) is assumed to be 42, representing a small area on the image; after calculation, F(100,20) = 75 (the higher the value, the stronger the texture); and for this specific area, assuming that the result of C() calculation is 32, this block is considered to have more features and a more complex area.

[0112] The judgment formula is: if \[F(x,y)\times C(z)>Threshold\], then set \[P[(x+y) / 2]=[z\times5+F(x,y)] / (C(z)\times 1.25). \] The significance of this formula is that for those blocks that are considered to be particularly visually prominent or have a large amount of information, more attention should be paid to them, perhaps by enlarging them in the final presentation or weighting the visual information of these areas to highlight them.

[0113] The logic behind the formula setting is to emphasize the local image parts that are meaningful or have high contrast and diversity, so that they can be displayed more clearly and distinctively in the process of full-color night vision imaging.

[0114] The full-color low-light-level night vision imaging method of the present invention comprises: based on an enhanced image sensor, capturing multiple frames of original images in a low-light environment, i.e., under low illumination conditions; then using an advanced noise suppression algorithm to remove noise from these original images, this process effectively solves the image noise problem caused by insufficient illumination under low-light conditions; then using image fusion and color compensation technology to enhance the expressiveness of image details while ensuring that the image colors truly reflect the actual situation, solve the problem of inaccurate color restoration caused by complex lighting conditions under low illumination, and supplement the color information lost due to weak light signals, thereby ensuring richer and more accurate color representation under the requirements of full-color display.

[0115] In addition, during the processing, the generated full-color night vision images are further optimized and adjusted to increase the frame update speed of the dynamic image sequence and enhance the effect of object detection, especially in the face of low-contrast backgrounds, the main target objects can still be distinguished with high quality. In this way, the invention also specifically solves the problem that target recognition becomes difficult in dim conditions and the practical application obstacles such as reduced image quality and reduced refresh rate caused by the large amount of computing and processing required for real-time imaging. Overall, the full-color low-light night vision imaging method has overcome a series of key technical problems that have plagued low-light photography through a series of technological innovations, and provides a more optimized practical solution with higher commercial value and social benefits.

[0116] like Figure 2-3 As shown, the present invention also discloses a special explosion-proof full-color low-light-level night vision camera for mines, comprising an explosion-proof housing, an explosion-proof lens, a high-definition image sensor, an image transmission module, an intelligent monitoring module and the full-color low-light-level night vision imaging system as described above, wherein the full-color low-light-level night vision imaging system is electrically connected to the high-definition image sensor and the image transmission module respectively, and the intelligent monitoring module is electrically connected to the high-definition image sensor and the remote intelligent terminal respectively.

[0117] Specifically, the material of the explosion-proof housing is 304 stainless steel, which can be used in corrosive environments to extend the service life of the equipment. The material of the explosion-proof lens is special explosion-proof glass with high light transmittance to improve image quality. It is coated with nano invisible wipers, which are non-sticky to water and oil, and repel dust to ensure the clarity of image acquisition. The image sensor uses a 1 / 1.8 high-definition sensor with improved photosensitivity. The intelligent monitoring module can integrate human intelligent detection, electronic fences, cross-border detection, off-post detection, and passenger flow statistics to improve the intelligence level of monitoring.

[0118] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0119] Any process or method description in the flow chart or otherwise described herein can be understood to represent a module, fragment or portion of a code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present application includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

[0120] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0121] It should be understood that the various parts of the present application can be implemented with hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method can be completed by instructing the relevant hardware through a program, which can be stored in a computer-readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0122] In addition, each functional unit in each embodiment of the present application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium can be a read-only memory, a disk or an optical disk, etc.

[0123] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A full-color low-light-level night vision imaging system, characterized in that: The system comprises: An image acquisition module is used to acquire multiple frames of original low-light images in a low-light environment based on an enhanced image sensor; An image noise processing module, used for processing the above-mentioned multiple frames of original low-light images by using a noise suppression algorithm to remove noise points in the images; Image color processing module, used to generate full-color night vision images with accurate colors and rich details based on image fusion and color compensation technology; The image optimization module is used to improve the frame rate and the recognition effect of the target in a low-contrast environment by performing real-time optimization processing on the full-color night vision image.

2. The full-color low-light-level night vision imaging system according to claim 1, characterized in that: Image color processing module includes: An extraction module, used for extracting high-frequency components and low-frequency components of each multi-frame original low-light image; A combining module, used for combining the high-frequency components of the multiple frames of original low-light images with a low-frequency component of a reference image; Color correction module, used to perform color correction operations to improve the color saturation and brightness level of the image, and automatically adjust the color balance according to the lighting conditions of the environment; The preliminary color formation module integrates high-frequency details and low-frequency contour information based on multi-scale fusion technology to form a preliminary color map.

3. The full-color low-light-level night vision imaging system according to claim 2, characterized in that: The extraction module includes: Transformation module, used to perform wavelet transformation on each single frame to obtain different resolution levels; Enhancement module, used to improve the image detail level based on contrast enhancement operator and enhance the visibility of weak signal areas; The weight assignment module is used to assign different weights according to the strength of features at each level; The pixel intensity enhancement coefficient setting module sets the corresponding pixel intensity enhancement coefficient α=75*TH / current layer mean value σ when the multi-frame average signal intensity value is higher than the predefined threshold TH; Where TH represents the signal threshold, σ represents the signal-to-noise ratio of the current layer of the image, and TH represents the reference standard set in advance by the user.

4. The full-color low-light-level night vision imaging system according to claim 3, characterized in that: Combined modules include: An arrangement module, used for arranging low-frequency component results of multiple frames in time sequence; A homogenization processing module is used to perform homogenization processing on each frame in the time domain using a smoothing algorithm to eliminate inter-frame fluctuations; An overlay module is used to use a reference low-frequency template image as an overlay base map to keep the background stable; An alignment module, used to ensure precise alignment between components based on spatial registration; The conditional expression is: (T_α*T_β) / D>θ. When this inequality holds, the exact matching is considered to be completed. T_α represents the time alignment factor; T_β represents the matching index after spatial transformation; D represents the data dimension difference; θ is the system error limit constant, and the default value is determined by the application scenario requirements to ensure the consistency and comparability of the final result image.

5. The full-color low-light-level night vision imaging system according to claim 4, characterized in that: The color correction module includes: A color mapping module, used to configure a color lookup table to map to the real environment color gamut to improve color reproduction accuracy; Color balance module, used to adaptively adjust the color balance value of each color area in the image to ensure smooth transition of natural tones; A color stabilization module that applies feedback control mechanisms to maintain stable viewing under changing lighting conditions; A color gain module, which calculates a color saturation gain coefficient CS based on a color temperature and a contrast level adjustment coefficient; If the environmental color temperature Temp_C < Temp_STD and the contrast CD is lower than the minimum acceptable contrast level, then set CS = (Min_Contrast + (Min_Brightness) * Max_Saturation) / Temp_STD; Where: Min_Contrast is the bottom line of the picture contrast threshold expected by the user; Min_Brightness is the lower limit of the picture brightness most desired to be achieved; Max_Saturation is the maximum color richness limit allowed; Temp_STD is the standard laboratory environmental light source.

6. The full-color low-light-level night vision imaging system according to claim 5, characterized in that: The color preliminary formation module includes: The model construction module constructs different image detail expression models with a multi-level pyramid; The model enhancement module enhances the details of the edge area using the Laplacian sharpening strategy; The color filling module combines the color mapping algorithm to add details to the contour to form color filling; The significant level setting module determines the significant level of each local part of the image based on the quality score matrix F of the fused image, the image block complexity evaluation index C, and the block position identifier P; The determination rule is that if F(x, y) * C(z) > Threshold, then set P[(x + y) / 2] = [z * 5 + F(x, y) * 1] / (C(z) * 1.25); Here, Threshold is a fixed threshold, x, y are the pixel index points in the matrix F, z represents the matrix block number, F(x, y) measures the detail intensity level of a specific coordinate, and C() measures the diversity of features within the image block.

7. A flameproof full-color low-light night vision camera for mines, characterized in that: It includes an explosion-proof housing, an explosion-proof lens, a high-definition image sensor, an image transmission module, an intelligent monitoring module, and the full-color low-light night vision imaging system as described in any one of claims 1-6. The full-color low-light night vision imaging system is electrically connected to the high-definition image sensor and the image transmission module respectively, and the intelligent monitoring module is electrically connected to the high-definition image sensor and the remote intelligent terminal respectively.

8. The explosion-proof full-color low-light night vision camera for mines according to claim 7 is characterized in that: The material of the explosion-proof housing is 304 stainless steel, the material of the explosion-proof lens is special explosion-proof glass, and it is coated with a nano invisible wiper. The image sensor uses a 1 / 1.8 high-definition sensor.

9. A full-color low-light-level night vision imaging method, characterized in that: The method includes: Obtaining multiple frames of original low-light images in a low-light environment based on an enhanced image sensor; Processing the multiple frames of original low-light images through a noise suppression algorithm to remove the noise in the images; Generating a full-color night vision image with accurate color and rich details based on image fusion and color compensation technology; Improving the frame rate and the recognition effect of the target in a low-contrast environment by performing real-time optimization processing on the full-color night vision image.

10. The method according to claim 9, characterized in that The method includes: Obtaining multiple frames of original low-light images in a low-light environment based on an enhanced image sensor; Processing the multiple frames of original low-light images through a noise suppression algorithm to remove the noise in the images; Performing wavelet transform on each single frame to obtain different resolution levels; Enhancing the image detail level based on a contrast enhancement operator to enhance the visibility of weak signal areas; Assigning different weights according to the intensity of the features at each level; When the multi-frame average signal intensity value is higher than the predefined threshold TH, set its corresponding pixel intensity enhancement coefficient α = 75 * TH / the mean value σ of the current layer, where TH represents the signal threshold, σ represents the signal-to-noise ratio of the current layer of the image, and TH represents the reference standard set by the user in advance; Arrange the low-frequency component results of multiple frames in chronological order; Use a smoothing algorithm to perform normalization processing in the time domain for each frame to eliminate inter-frame fluctuations; Use the reference low-frequency template image as the superimposed background image to keep the background stable; Based on spatial registration to ensure precise alignment between each component, whose conditional expression is: (T_α * T_β) / D > θ. When this inequality holds, it is considered that the precise pairing is completed, where T_α represents the time alignment factor; T_β represents the matching index after spatial transformation, D represents the data dimension difference, and θ is the system error bound constant, and the default value is determined by the application scenario requirements to ensure the consistency and comparability of the final result image; Configure the color lookup table to map to the real environment color gamut to improve the color restoration accuracy; Adaptive adjustment of the color balance value of each color area in the image to ensure a smooth transition of natural tones; Apply a feedback control mechanism under changing lighting conditions to maintain a stable visual perception; Calculate the color saturation gain coefficient CS based on the color temperature and contrast level adjustment coefficient; If the ambient color temperature Temp_C < Temp_STD and the contrast CD is lower than the minimum acceptable contrast level, then set CS = (Min_Contrast + (1 - Min_Brightness) * Max_Saturation) / Temp_STD, where: Min_Contrast is the threshold bottom line of the picture contrast expected by the user, Min_Brightness is the lower limit of the picture brightness most desired to be achieved, Max_Saturation is the maximum color fullness limit allowed, and Temp_STD represents the standard laboratory environmental light source; And automatically adjust the color balance according to the lighting conditions of the environment; The multi-level pyramid constructs different image detail expression models; Use the Laplacian sharpening strategy to enhance the details of the edge area; Combine the color mapping algorithm to add details to the contour to form color filling; Based on the quality score matrix F of the fused image, the image block complexity evaluation index C, and the block position identifier P, determine the significant level of each local part of the image, The determination rule is that if F(x, y) * C(z) > Threshold, then set P[(x + y) / 2] = [z * 5 + F(x, y) * 1] / (C(z) * 1.25), Here, Threshold is a fixed threshold, x, y are the pixel index points in the matrix F, z represents the matrix block number, F(x, y) measures the detail intensity level of a specific coordinate, and C() measures the diversity of features within the image block; Improve the frame rate and the recognition effect of the target in a low-contrast environment by performing real-time optimization processing on the full-color night vision image.

Images