A three-dimensional depth perception method for coal rock identification

Abstract

The application discloses a three-dimensional depth perception method for coal rock identification, comprising the following steps: a blue light speckle projector projects a blue light speckle coding pattern to coal rock; a binocular camera simultaneously collects the projected blue light speckle coding pattern; the input blue light speckle pattern of the left or right camera is subjected to monocular structured light depth decoding with a blue light speckle reference pattern, and a rough parallax map of the irradiated coal rock is output; a rough depth value of the coal rock is calculated according to a monocular structured light depth calculation formula and the rough parallax map; the rough depth value is substituted into a binocular depth calculation formula to inversely calculate a corresponding binocular parallax value, and the input blue light speckle patterns of the left and right cameras are subjected to binocular accurate matching calculation according to the reference parallax value to obtain an accurate parallax map; and an accurate depth value of the coal rock is calculated according to a binocular depth calculation formula and the accurate parallax map. The application obtains accurate depth information through monocular and binocular structured light fusion depth decoding, and greatly improves the accuracy and detail resolution capability of depth ranging.

Description

Technical Field

[0001] This invention belongs to the field of mine geological exploration and intelligent coal mine technology, and relates to a three-dimensional depth perception method for coal and rock identification. Background Technology

[0002] In my country's current coal industry, intelligent and digital transformation has become crucial for promoting coal mine safety, improving production efficiency, and reducing environmental impact. In particular, sensing technologies in the mine environment, especially coal and rock identification technology, are a core component of smart coal mine construction and are essential for enhancing operational safety, automation levels, and emergency response capabilities.

[0003] The working environment in coal mines is characterized by low light, low texture, dust interference, and dynamic occlusion, making traditional depth sensing technologies inadequate for practical applications. For example, traditional Time-of-Flight (ToF) cameras face severe multipath effects in mine environments, especially in the confined working spaces where multiple signal reflections distort depth data. Furthermore, the absorption by black coal and rock significantly increases depth errors, and the high dust and humidity further degrade their ranging performance. Another common depth sensing technology is binocular stereo vision, which, while offering advantages in structural flexibility and lower cost, also faces significant challenges in mine environments. Insufficient lighting and the lack of distinct texture features in coal and rock cross-sections make it difficult for binocular vision systems to acquire stable matching points, leading to holes and mismatches in the depth map. Particularly in low-light and low-texture environments, traditional image feature matching-based depth estimation methods are ineffective, impacting the overall system's accuracy and reliability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a three-dimensional depth perception method for coal and rock identification, solving the problems of insufficient accuracy and reliability of existing underground depth perception in coal mines. The present invention is based on active-passive fusion visual perception technology, actively projecting blue light speckle structured light encoded patterns onto the coal and rock, and then obtaining accurate depth information through binocular reception and mono- and binocular structured light fusion depth decoding.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] A three-dimensional depth sensing method for coal and rock identification includes the following steps:

[0007] S100: Blue light speckle projector projects blue light speckle coded patterns onto coal and rock;

[0008] S200: Dual-lens cameras simultaneously capture the projected blue light speckle pattern;

[0009] S300: Monocular structured light depth decoding: Select the input blue light speckle map from the left or right camera, perform monocular structured light depth decoding with a blue light speckle reference map at a known reference distance, and output a coarse parallax map of the illuminated coal and rock.

[0010] S400: Monocular structured light depth calculation: Based on the monocular structured light depth calculation formula, the intrinsic and extrinsic parameters of the monocular camera and blue light speckle projector, and the coarse parallax map, the coarse depth value of the coal and rock is calculated.

[0011] S500: Binocular structured light depth decoding: Substitute the coarse depth value into the binocular depth calculation formula to obtain the corresponding binocular disparity value as the reference disparity value for binocular disparity calculation. The input blue light speckle map of the left and right binoculars is precisely matched and calculated within the effective range based on the reference disparity value to obtain the accurate disparity map.

[0012] S600: Binocular structured light depth calculation: Based on the binocular depth calculation formula, binocular camera intrinsic and extrinsic parameters, and accurate disparity map, the accurate depth value of coal and rock is calculated.

[0013] The present invention also includes the following technical features:

[0014] Specifically, the blue light speckle projector includes a blue light LD laser source, a collimating lens, and a DoE optical diffraction device. The projected blue light speckle coding pattern is a speckle patch composed of several randomly distributed blue light speckles and is unique.

[0015] Specifically, the binocular camera and the blue light speckle projector are on the same baseline, and the projection and reception are performed synchronously; the binocular camera is symmetrically distributed on both sides of the blue light speckle projector and maintains a baseline distance S, and the baseline distance between the binocular cameras is 2S.

[0016] Specifically, the blue light speckle projector works synchronously with the binocular camera, with projection and reception working together to collect data;

[0017] Alternatively, a blue speckle projector can be switched to project white LED lighting. A binocular camera simultaneously uses a time-division multiplexing method to alternately acquire blue speckle encoded patterns, i.e., input blue speckle images and RGB images. The input blue speckle images are used for depth decoding, and the RGB images are used for texture information acquisition.

[0018] Specifically, S300 includes:

[0019] S301: Select the left or right eye camera, and preprocess its input blue light speckle image through histogram enhancement or binarization to extract blue light speckle spots in the image, so that the acquired input blue light speckle image is highly similar to the speckle image patch of the blue light speckle reference image at a known reference distance of the camera.

[0020] S302: Perform block matching disparity calculation between the preprocessed input blue speckle map and the blue speckle reference map with a known reference distance to obtain a coarse disparity map corresponding to the preprocessed input blue speckle map.

[0021] Specifically, the blue light speckle reference pattern is generated by projecting a blue light speckle coded pattern onto a point perpendicular to the optical central axis Z of the projector and at a distance d from the projector. ref On the reference plane, a static image is captured by a camera, and after image preprocessing, a standard speckle pattern is stored and solidified in the memory.

[0022] Specifically, in S400, the monocular structured light depth calculation formula calculates d, using the horizontal offset Δx as the input parameter:

[0023]

[0024] Where, d ref Δx is the known distance of the monocular blue speckle reference image, Δx is the disparity value corresponding to the optimal matching block obtained by block matching disparity calculation between the preprocessed input blue speckle image and the preprocessed blue speckle reference image at the known distance, S is the baseline distance between the left or right eye camera and the blue speckle projector, f is the focal length of the left or right eye camera, and μ is the dot distance of the image sensor of the left or right eye camera.

[0025] Specifically, in S500, the binocular depth calculation formula is used to calculate the binocular disparity value Δx' as a reference disparity value for binocular disparity calculation, with the coarse depth value d as the input parameter:

[0026]

[0027] Where S is the baseline distance between the left or right eye camera and the blue light speckle projector, f is the focal length of the left or right eye camera, and μ is the dot pitch of the image sensor of the left or right eye camera.

[0028] Specifically, in S500, the input blue light speckle map of the left and right binoculars is used for binocular precise matching calculation within the effective range based on the reference disparity value to obtain a precise disparity map, including:

[0029] The input blue light speckle images from the left and right cameras are preprocessed using histogram enhancement or binarization.

[0030] The input speckle image block in the left-eye input blue light speckle image is used to search for the optimal speckle image matching block in the right-eye input blue light speckle image based on the reference disparity value; or the input speckle image block in the right-eye input blue light speckle image is used to search for the optimal speckle image matching block in the left-eye input blue light speckle image based on the reference disparity value; thus obtaining the accurate disparity map corresponding to the input blue light speckle image.

[0031] Specifically, in S600, d is calculated according to the binocular depth calculation formula. l,r With horizontal offset Δx l,r As input parameters:

[0032]

[0033] In the formula, the horizontal offset Δx l,r Input blue light speckle pattern I for left eye l The input speckle image patch B' and its corresponding right eye input blue light speckle image I r The optimal offset of the optimal speckle image matching block B in the X direction; S is the baseline distance between the left or right eye camera and the blue light speckle projector; f is the focal length of the left or right eye camera; and μ is the dot distance of the image sensor of the left or right eye camera.

[0034] Compared with the prior art, the present invention has the following technical effects:

[0035] This invention uses the sparse depth map obtained from monocular structured light depth decoding under blue light speckle projection as guiding information for high-resolution binocular structured light depth decoding. This significantly reduces the computational load of binocular structured light depth decoding (both the matching block and search range are greatly reduced), and the resulting depth map is more accurate and richer in detail. This method combines the advantages of monocular structured light cameras and binocular stereo ranging, overcoming the challenges of black absorption in coal and rock, low illumination, and weak texture. Through monocular and binocular structured light fusion depth decoding, it greatly improves the accuracy of depth ranging and the ability to resolve details. Attached Figure Description

[0036] Figure 1 This is a flowchart of a three-dimensional depth sensing method for coal and rock identification according to an embodiment of the present invention.

[0037] Figure 2 This is a schematic diagram of monocular structured light depth calculation according to an embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram of the input speckle image block and the search for the optimal speckle image matching block in an embodiment of the present invention.

[0039] Figure 4 This is a schematic diagram of the binocular block matching depth calculation according to an embodiment of the present invention. Detailed Implementation

[0040] The following are specific embodiments of the present invention. It should be noted that the present invention is not limited to the following specific embodiments. All equivalent modifications made based on the technical solutions of this application fall within the protection scope of the present invention.

[0041] Example 1:

[0042] This embodiment provides a three-dimensional depth perception method for coal and rock identification. Based on active-passive fusion visual perception technology, a blue light speckle structured light encoded pattern is actively projected onto the coal and rock. The speckle encoded pattern is received by left and right binoculars on the same baseline. After monocular and binocular structured light fusion depth decoding, accurate depth information is obtained.

[0043] Figure 1 The schematic diagram illustrates the overall flow of the three-dimensional depth sensing method for coal and rock identification according to an embodiment of the present invention. For clarity, the following will combine... Figure 2 , Figure 3 , Figure 4 To describe this method.

[0044] The present invention provides a three-dimensional depth sensing method for coal and rock identification, comprising the following steps:

[0045] S100: Blue light speckle projector projects blue light speckle coded patterns onto coal and rock.

[0046] The blue speckle projector includes a blue LD laser source, a collimating lens, and a DoE (optical diffraction device). The projected blue speckle pattern remains unchanged and is unique within a certain range, consisting of a speckle patch composed of several randomly distributed blue speckle spots.

[0047] S200: Dual-lens cameras simultaneously capture the projected blue light speckle pattern.

[0048] The binocular camera and the blue light speckle projector are on the same baseline, and the projection and reception are synchronized. The binocular cameras are of the same specifications and are symmetrically distributed on both sides of the blue light speckle projector, maintaining a certain baseline distance S, that is, the baseline distance between the two cameras is 2S.

[0049] One working mode is that the blue light speckle projector and the binocular camera work simultaneously, that is, the projection and reception work together to collect data;

[0050] Another working mode is that the blue light speckle projector and white LED lighting can be switched. The binocular camera can simultaneously use time-division multiplexing to alternately acquire the blue light speckle encoded pattern (i.e., the input blue light speckle image) and the RGB image. The input blue light speckle image is used for depth decoding, and the RGB image is used for texture information acquisition.

[0051] S300: Monocular Structured Light Depth Decoding

[0052] The monocular structured light depth decoding module selects the blue light speckle image from the left or right camera, performs monocular structured light depth decoding with a blue light speckle reference image at a known reference distance, and outputs a coarse parallax image of the illuminated coal rock.

[0053] Specifically, monocular structured light depth decoding includes the following steps:

[0054] S301: Select the left or right eye camera, and preprocess its input blue light speckle image through histogram enhancement or binarization to extract blue light speckle spots in the image, so that the acquired input blue light speckle image is highly similar to the speckle image patch of the blue light speckle reference image at a known reference distance of the camera.

[0055] S302: Perform block matching disparity calculation between the preprocessed input blue speckle map and the blue speckle reference map with a known reference distance to obtain a coarse disparity map corresponding to the preprocessed input blue speckle map;

[0056] In this step, the preprocessed input blue speckle image is matched with a blue speckle reference image at a known reference distance, and block matching disparity is calculated using speckle image blocks of a certain size (the size of the speckle image block is k×l, where k and l are positive integers). (The specific calculation is as follows...) Figure 2 As shown, the specific strategy is to search for the optimal matching block with the highest similarity within a K×L search window (K and L are integers, greater than k and l) of the monocular blue light speckle reference image for an input speckle image patch of size k×l (the search process can refer to the traditional image patch matching method) to obtain the disparity value Δx corresponding to the optimal matching block, and then calculate the coarse disparity map corresponding to the input blue light speckle image block by block.

[0057] Before deployment, blue light speckle reference images from either the left or right eye camera must be collected and archived as a reference benchmark for matching and comparison. The blue light speckle reference image is generated by projecting a blue light speckle coded pattern onto a point perpendicular to the projector's optical central axis (Z-axis) and at a distance d from the projector. ref On a plane (which can be composed of a projection screen, a flat plate, etc., used to present a clear and stable image, this plane can be called a reference plane), a static image is captured by a camera, and after image preprocessing, it is stored and solidified in a memory for use as a standard pattern for matching the reference and monocular structured light block matching parallax calculation. Preferably, the blue light speckle reference pattern of the blue light speckle projector is composed of multiple speckles at a known distance d. ref The standard speckle pattern.

[0058] S400: Monocular Structured Light Depth Calculation

[0059] Based on the monocular structured light depth calculation formula, the intrinsic and extrinsic parameters of the monocular camera and blue light speckle projector, and the coarse parallax map, the coarse depth value d of the coal and rock is calculated.

[0060] In this embodiment, d is calculated according to the following monocular depth calculation formula (1), taking the horizontal offset Δx as an example:

[0061]

[0062] like Figure 2 In, among which, d ref Δx is the known distance of the monocular blue speckle reference image, Δx is the disparity value corresponding to the optimal matching block obtained by block matching disparity calculation between the preprocessed input blue speckle image and the preprocessed blue speckle reference image at the known distance, S is the baseline distance between the left or right eye camera and the blue speckle projector, f is the focal length of the left or right eye camera, and μ is the dot distance of the image sensor of the left or right eye camera.

[0063] S500: Binocular Structured Light Depth Decoding

[0064] The binocular structured light depth decoding module takes the coarse depth value obtained in step S400 and substitutes it into the binocular depth calculation formula to obtain the corresponding binocular disparity value (as a reference disparity value for binocular disparity calculation). The input blue light speckle map of the left and right binoculars is precisely matched and calculated within the effective range based on the reference disparity value to obtain a more detailed and accurate disparity map.

[0065] In this embodiment, the binocular disparity value Δx' is calculated in reverse according to the following binocular depth calculation formula (2) (as a reference disparity value for binocular disparity calculation), taking the coarse depth value d as an example as the input parameter:

[0066]

[0067] Where S is the baseline distance between the left or right eye camera and the blue light speckle projector, f is the focal length of the left or right eye camera, and μ is the dot pitch of the image sensor of the left or right eye camera.

[0068] The input blue light speckle patterns for both left and right binoculars are precisely matched within the effective range based on the reference disparity value Δx'. One mode involves calculating the input speckle image patch in the left-eye input blue light speckle pattern and matching it with the right-eye input blue light speckle pattern based on the reference disparity value Δx'. Figure 1 The optimal speckle image matching block is searched within a certain range; another mode is to use the input speckle image block in the right eye input blue light speckle image, and then match it with the reference disparity value Δx' in the left eye input blue light speckle image. Figure 1 Search for the optimal speckle image matching block within a defined range. Specifically:

[0069] S501: Apply the same preprocessing method as step S301 to the input blue light speckle images of the left and right cameras, and preprocess the input blue light speckle images by histogram enhancement or binarization.

[0070] S502: One mode is binocular input blue light speckle pattern I l I r For the calculation of mutual stereo block matching, see Figure 3 This allows us to obtain the blue light speckle pattern input to the left eye (I). l The corresponding precise disparity map. Specifically:

[0071] Input blue light speckle pattern I into the left eye l Extract a speckle image patch B'block of a certain size from the input image. m×n The center point is o'; based on the reference parallax value Δx', input the blue light speckle pattern I in the right eye. r Extract a matching search window of a certain size, centered at the position o corresponding to the center point o' of the input speckle image patch (the horizontal disparity between o' and o is Δx'). M×N The size is M×N, where M and N are integers and can be equal or unequal. Generally, M≥N, M>m, and N≥n; then, in the Match search window... M×N Extract all speckle matching blocks that are the same size as the input speckle image patch. k The size is m×n, and the center point of the matching block is o' k k is an integer representing the number of matching blocks.

[0072] Preferably, the input speckle image patch size m×n is smaller than the speckle image patch size k×l selected in step 302 monocular structured light depth decoding; the matching search window Match M×N The size M×N is also smaller than that of direct binocular matching calculation.

[0073] Next, the input speckle image patches are calculated separately. Match with k speckle blocks k The similarity value between them is match_value k This value serves as a similarity metric for image patch matching.

[0074] Finally, all similarity values ​​match_value k Find the minimum value in the middle, and the corresponding speckle matching block. k That is, the input speckle image block B'block. m×n The optimal speckle image matching block B to be searched is the location information corresponding to the minimum value of the input speckle image block. m×n The offset of the center point o (Δx)l,r Δy l,r ), that is, the motion vector of the input speckle image patch B'. Figure 3 As shown, the input speckle image patch is the blue light speckle image I input to the left eye. l The medium gray area represents the region, and the optimal matching block is the right eye input blue light speckle map I. r In the matching search window, the diagonal lines represent regions, with their center point o. k Matching search window block M×N The optimal offset of the center point o (which corresponds to the position of the center point o of the input speckle image patch) is (Δx) l,r Δy l,r ), representing the displacement in the X and Y axes respectively. The offset value is calculated by subtracting the coordinates of the center point o of the matching search window (x, y) from the coordinates of the center point of the optimal matching block (x', y') along the X and Y axes respectively, and then taking the absolute value. It is represented by the number of pixels.

[0075] S600: Binocular structured light depth calculation:

[0076] The depth calculation module calculates the precise depth value of coal and rock based on the binocular depth calculation formula, binocular camera intrinsic and extrinsic parameters, and accurate disparity map.

[0077] In this embodiment, d is calculated according to the following binocular depth calculation formula (3). l,r With horizontal offset Δx l,r For example, see [example input parameter] Figure 4 :

[0078]

[0079] In the formula, the horizontal offset Δx l,r Input blue light speckle pattern I for left eye l The input speckle image patch B' and its corresponding right eye input blue light speckle image I r The optimal offset of the optimal speckle image matching block B in the X direction, i.e., the right eye input blue light speckle image I. r The x-coordinate of the center point of the matching search window is subtracted from the x-coordinate of the center point of the optimal matching block B found in the matching search window, and the absolute value is taken, which is expressed as the number of pixels; S is the baseline distance between the left or right eye camera and the blue light speckle projector, f is the focal length of the left or right eye camera, and μ is the dot distance of the image sensor of the left or right eye camera.

[0080] Preferably, the size of the input speckle image patch is selected based on the fact that the input speckle image patch has relative uniqueness within a certain range in the horizontal or vertical direction, that is, the features of the input speckle image patch are different from the features of other input speckle image patches of the same size, and can be distinguished from other input speckle image patches of the same size.

[0081] Preferably, the similarity value calculation method of the present invention uses the method of calculating the sum of absolute differences (SAD) between corresponding pixels of the input image block and the matching block, but is not limited to this method.

[0082] Move the center point o of the projected image patch to the next pixel in the same row, and repeat step 502 to calculate the depth value corresponding to the next pixel. Continue this process, calculating point by point from left to right and top to bottom, to obtain the left eye input blue light speckle map I. l The corresponding precise depth value for the entire image.

[0083] Similarly, the right eye inputs blue light speckle pattern I r Alternatively, blue light speckle pattern I can be input into the left eye according to steps S502 and S600. l Search for the optimal speckle image matching block within a certain range to obtain the corresponding precise depth value of the entire image.

[0084] As an example, the binocular camera of the present invention uses two independent cameras with the same performance indicators (the same optical lens and image sensor), which are arranged symmetrically and equidistantly on both sides of the blue light speckle projector. Their optical axes are parallel to the optical axis of the coded pattern projector and are on the same baseline. However, the baseline of the two cameras can be adjusted according to different needs, or two cameras with different focal lengths or models can be used.

[0085] As an example, the search strategy for matching blocks in this invention adopts traditional full-search block matching, but various other improved search strategies can also be used; the similarity value calculation method adopts the sum of absolute differences (SAD) method, but is not limited to this method; all methods that adopt a similar process to the content of this invention should be included within the scope of the claims of this invention.

[0086] Although the above embodiments are implemented in specific systems, they are not intended to limit the invention. The invention can be similarly applied to similar coded pattern projection and image sensor systems. The invention supports not only structured light modes generated by different laser sources, such as infrared, visible light, ultraviolet light, and invisible light, but also projection schemes for different patterns, such as dotted, blocky, cross-shaped, and striped patterns. Therefore, any modifications and improvements that do not depart from the spirit and scope of the invention should be included within the scope of the foregoing claims.

[0087] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0088] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0089] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A three-dimensional depth sensing method for coal and rock identification, characterized in that, Includes the following steps: S100: Blue light speckle projector projects blue light speckle coded patterns onto coal and rock; S200: Dual-lens cameras simultaneously capture the projected blue light speckle pattern; S300: Monocular structured light depth decoding: Select the input blue light speckle map from the left or right camera, perform monocular structured light depth decoding with a blue light speckle reference map at a known reference distance, and output a coarse parallax map of the illuminated coal and rock. S400: Monocular structured light depth calculation: Based on the monocular structured light depth calculation formula, the intrinsic and extrinsic parameters of the monocular camera and blue light speckle projector, and the coarse parallax map, the coarse depth value of the coal and rock is calculated. S500: Binocular structured light depth decoding: Substitute the coarse depth value into the binocular depth calculation formula to obtain the corresponding binocular disparity value as the reference disparity value for binocular disparity calculation. The input blue light speckle map of the left and right binoculars is precisely matched and calculated within the effective range based on the reference disparity value to obtain the accurate disparity map. S600: Binocular structured light depth calculation: Based on the binocular depth calculation formula, binocular camera intrinsic and extrinsic parameters, and accurate disparity map, the accurate depth value of coal and rock is calculated; In S400, the monocular structured light depth calculation formula is used to calculate... With horizontal offset As input parameters: （1） in, For the known distances in the monocular blue light speckle reference image, To obtain the disparity value corresponding to the optimal matching block, block matching disparity is calculated between the preprocessed input blue light speckle map and the preprocessed blue light speckle reference map with known distances. The baseline distance between the left or right eye camera and the blue light speckle projector. The focal length of the left or right eye camera. The dot pitch of the image sensor for the left or right eye camera; In S500, the binocular depth calculation formula is used to inversely calculate the binocular disparity value. As a reference disparity value for binocular disparity calculation, a coarse depth value is used. As input parameters: （2） in, The baseline distance between the left or right eye camera and the blue light speckle projector. The focal length of the left or right eye camera. The dot pitch of the image sensor for the left or right eye camera; In S600, the depth is calculated based on the binocular depth calculation formula. With horizontal offset As input parameters: （3） In the formula, the horizontal offset Input blue light speckle pattern for left eye Input speckle image patch The corresponding right eye input blue light speckle pattern Optimal speckle image matching patch The optimal offset in the X direction; The baseline distance between the left or right eye camera and the blue light speckle projector. The focal length of the left or right eye camera. This refers to the dot pitch of the image sensor for the left or right eye camera.

2. The three-dimensional depth sensing method for coal and rock identification as described in claim 1, characterized in that, The blue light speckle projector includes a blue light LD laser source, a collimating lens, and a DoE optical diffraction device. The projected blue light speckle coding pattern is a speckle patch composed of several randomly distributed blue light speckles and is unique.

3. The three-dimensional depth sensing method for coal and rock identification as described in claim 1, characterized in that, The binocular cameras and the blue light speckle projector are on the same baseline, and projection and reception are performed synchronously; the binocular cameras are symmetrically distributed on both sides of the blue light speckle projector while maintaining a baseline distance. The baseline distance between the two cameras is .

4. The three-dimensional depth sensing method for coal and rock identification as described in claim 1, characterized in that, The blue light speckle projector works synchronously with the binocular camera, and the projection and reception work together to collect data. Alternatively, a blue speckle projector can be switched to project white LED lighting. A binocular camera simultaneously uses a time-division multiplexing method to alternately acquire blue speckle encoded patterns, i.e., input blue speckle images and RGB images. The input blue speckle images are used for depth decoding, and the RGB images are used for texture information acquisition.

5. The three-dimensional depth sensing method for coal and rock identification as described in claim 1, characterized in that, The S300 includes: S301: Select the left or right eye camera, and preprocess its input blue light speckle image through histogram enhancement or binarization to extract blue light speckle spots in the image, so that the acquired input blue light speckle image is highly similar to the speckle image patch of the blue light speckle reference image at a known reference distance of the camera. S302: Perform block matching disparity calculation between the preprocessed input blue speckle map and the blue speckle reference map with a known reference distance to obtain a coarse disparity map corresponding to the preprocessed input blue speckle map.

6. The three-dimensional depth sensing method for coal and rock identification as described in claim 5, characterized in that, The blue light speckle reference image is obtained by projecting a blue light speckle coded pattern onto a surface perpendicular to the optical center axis (Z-axis) of the projector and at a distance from the projector. On the reference plane, a static image is captured by a camera, and after image preprocessing, a standard speckle pattern is stored and solidified in the memory.

7. The three-dimensional depth sensing method for coal and rock identification as described in claim 1, characterized in that, In step S500, the input blue light speckle maps of the left and right binoculars are calculated using precise binocular matching within an effective range based on reference disparity values ​​to obtain a precise disparity map, including: The input blue light speckle images from the left and right cameras are preprocessed using histogram enhancement or binarization. The input speckle image block in the left-eye input blue light speckle image is used to search for the optimal speckle image matching block in the right-eye input blue light speckle image based on the reference disparity value; or the input speckle image block in the right-eye input blue light speckle image is used to search for the optimal speckle image matching block in the left-eye input blue light speckle image based on the reference disparity value; thus obtaining the accurate disparity map corresponding to the input blue light speckle image.

Images