Cordyceps sinensis locating system based on binocular stereo and deep learning technology

Abstract

The invention discloses a cordyceps sinensis locating system based on a binocular stereo and deep learning technology, and relates to the technical field of cordyceps sinensis collection, the cordyceps sinensis locating system comprises a hardware structure and a software system, the hardware structure comprises a mountaineering walking stick, a supporting cross beam, a CCD camera, a controller, a battery, a camera support, a controller supporting frame, a battery supporting frame and a tablet personal computer; the software system comprises an image acquisition module, a training learning module, an identification module and an interaction module; the image acquisition module acquires a worm grass image based on a CCD camera and uploads the acquired worm grass image to the training learning module. The identified image is sent to the tablet personal computer in a wireless communication mode, display software is deployed on the tablet personal computer, when cordyceps sinensis is found, the tablet personal computer can give out a prompt tone, meanwhile, the cordyceps sinensis is selected on the image, and compared with a traditional cordyceps sinensis collecting method, the method has the advantages of being high in identification efficiency and accuracy and good in practicability. The labor intensity of cordyceps sinensis collecting personnel can be effectively reduced, and the collecting efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cordyceps collection, and specifically to a cordyceps positioning system based on binocular stereo and deep learning technologies. Background Art

[0002] Cordyceps is a plant of the Clavicipitaceae family, also known as "Dongchong Xiacao" (Cordyceps sinensis), a special product of China. It is a strange thing where an insect and a plant grow together. In winter, it is an insect, and in summer, a plant grows out of the insect. The insect is the larva of the Hepialus armoricanus Oberthur, and the plant is a kind of Cordyceps fungus. In summer, the insect lays eggs on the flower leaves of the grass, and with the leaves falling to the ground, after about a month of hatching, it becomes a larva, which then burrows into the damp and soft soil layer. The ascospores of the Cordyceps fungus in the soil layer invade those fat and well-developed larvae. After being invaded by the spores, the larvae drill towards the shallow ground layer, and the spores grow inside the larvae. The internal organs of the larvae slowly disappear, and the body becomes a shell filled with hyphae, buried in the soil layer. After a winter, when spring comes the next year, the hyphae begin to grow and grow out of the ground in summer, growing into a small grass. In this way, the shell of the larva and the small grass together form a complete "Dongchong Xiacao" (Cordyceps sinensis). Cordyceps is mainly produced on the Qinghai-Tibet Plateau within Sichuan and Qinghai provinces in China. As a traditional precious Chinese herbal medicine in China, Cordyceps sinensis is favored by people for its extremely high medicinal value and therapeutic effects. However, due to its harsh growth environment, it is difficult to pick, resulting in a high price when purchased. In recent years, with the improvement of people's living standards, the demand for Cordyceps has shown an upward trend. In the existing technology, the collection of wild Cordyceps mainly relies on manual identification and excavation. Since the above-ground part of Cordyceps is not easily distinguishable from the environmental weeds, when collecting manually, people usually lie on the ground or squat on the ground to work. This not only has a large labor intensity and very low efficiency, but also often causes large-scale damage to the vegetation. In view of this, we propose a cordyceps positioning system based on binocular stereo and deep learning technologies. Summary of the Invention

[0003] To solve the above technical problems, a cordyceps positioning system based on binocular stereo and deep learning technologies is provided. This technical solution solves the problems of large labor intensity, low efficiency, and large-scale damage to vegetation in manual collection.

[0004] To achieve the above object, the technical solution adopted by the present invention is: a cordyceps positioning system based on binocular stereo and deep learning technologies, including: a hardware structure and a software system, wherein the hardware structure includes a hiking stick, a support crossbeam, a CCD camera, a controller, a battery, a camera bracket, a controller support frame, a battery support frame, and a tablet computer;

[0005] The software system includes an image acquisition module, a training and learning module, an identification module, and an interaction module;

[0006] The image acquisition module collects cordyceps images based on a CCD camera and uploads the collected cordyceps images to the training and learning module;

[0007] The training and learning module trains the cordyceps image data based on a deep learning framework, learns and optimizes the CA attention mechanism, adds the CA attention mechanism in front of SPPF in the BackBone of the yolov5s model, and constructs a cordyceps recognition model;

[0008] The recognition module inputs the collected image data based on the constructed cordyceps recognition model, reconstructs the cordyceps point cloud based on the deep learning model, performs three-dimensional positioning, extracts the position of the cordyceps from the color image, and judges and analyzes the position of the cordyceps;

[0009] The interaction module performs human-computer interaction based on a tablet computer. As the handheld terminal of the cordyceps collectors, it is deployed with client software, receives the returned cordyceps images in real time, instructs the collectors to collect cordyceps, and emits an indication sound when cordyceps is recognized.

[0010] Preferably, one end of the support cross beam is a semi-circular structure that is in transitional fit with the hiking stick and is fixed to the hiking stick by an internal hexagonal bolt. The support cross beam is provided with a CCD camera, a controller and a battery. The battery is used to supply power to the controller. The CCD camera is connected to the controller through USB or network cable. The tablet computer is held by hand.

[0011] Preferably, the CA attention mechanism structure first performs average pooling operations on the obtained image in the horizontal and vertical directions, integrates the feature information in the horizontal and vertical directions and performs convolution processing on it. The specific steps are as follows:

[0012] Given the input layer X, with a size of H×W×C, pooling kernels with sizes of (H, 1) and (1, W) are used to encode each channel along the horizontal and vertical directions respectively. The outputs of the c channels with a height of H and the c-th channel with a width of W are respectively:

[0013]

[0014] In the formula, H and W respectively represent the height and width of the feature layer, which are known, and respectively represent the output results of the c-th channel in the H and W directions, x c (h,i) and x c (j,w) represent the inputs of the input feature X along the H direction and the W direction.

[0015] Preferably, the feature maps in two independent directions output by the above two groups of formulas are subjected to splicing operations, 1×1 convolution operations and activation operations. The formula is:

[0016] f = σ(F1 [Z h ,Z w )

[0017] In the formula, f represents the intermediate feature map after encoding spatial information in the horizontal and vertical directions, [.,.] represents the splicing operation in the spatial dimension, F1 represents the convolution operation, and δ represents the Sigmoid function;

[0018] Split f along the spatial dimension, and then use two 1×1 convolution operations to restore the number of channels of the feature map to the same as that of the input feature X. Calculate the attention weight of the feature map using the Sigmoid function, and the result is:

[0019] g h =σ(F h (f h ))

[0020] g w =σ(F w (f w ))

[0021] In the formula, g h and g w represent the attention weights of the convolution and activation after splitting f. f h and f w respectively represent the feature maps of f along the H direction and the W direction after splitting. F h and F w represent the convolution operation.

[0022] Preferably, multiply g h and g w by the input feature map weight and perform weighted output to obtain a feature map with coordinate attention weight. The formula is expressed as:

[0023]

[0024] In the formula, y c (i,j) and x c (i,j) respectively represent the input and output of the c-th channel. and respectively represent the attention weights along the H direction and the W direction on the c-th channel.

[0025] Preferably, the identification module identification method inputs the collected image data into the cordyceps identification model, and uses the bounding box box(x, y, width, height) to represent the position of the cordyceps. Here, x is the value along the horizontal direction of the image coordinate at the upper left point of the rectangular box, y is the value along the vertical direction of the image coordinate at the upper left point of the rectangular box, width and height are the lengths along the horizontal and vertical directions of the image coordinate of the rectangular box respectively, and the bounding box box(x, y, width, height) is the rectangular box output after algorithm identification.

[0026] Preferably, based on the image acquisition module, the depth map of the cordyceps is obtained, the rectangular box box(x, y, width, height) is drawn, the region of interest of the depth map is extracted, and it is obtained by using the algorithm of image processing based on the comprehensive rectangular box. Then, using the formula of binocular vision single-point reconstruction, the pixel points of the region of interest are reconstructed, and the point cloud after reconstruction is the point cloud of the cordyceps, that is, the positioning of the cordyceps is realized. The formula is:

[0027] x ccs =(u - cx) * z ccs / f x

[0028] y ccs =(v - cy) * z ccs / f y

[0029] z ccs = fb / d, d = u l - u r

[0030] In the formula, (x ccs , y ccs , z ccs ) represents the three-dimensional point in the camera coordinate system after reconstruction, (u, v) is the image coordinate of the pixel in the depth map, f is the focal length of the camera, (f x , f y ) is the component of the camera focal length, b is the baseline distance of the camera, d is the disparity of the camera, and u l , u r are the projection points of the reconstructed point on the left camera and the right camera respectively.

[0031] Preferably, the tablet computer is used as the carrier of the interaction module, and the collected cordyceps picture is transmitted to the client software on the tablet computer through the network transmission link, and the analysis result is visually displayed to show the position of the cordyceps.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] The present invention proposes a system that combines binocular stereo vision technology, image fusion technology, deep learning technology, and embedded technology. The deep learning is used to identify and record the positions of cordyceps, achieving fast identification and positioning of cordyceps. The identified images are sent to a tablet computer via wireless communication. A display software is deployed on the tablet computer. When cordyceps is detected, the tablet computer will emit a prompt sound, and at the same time, the cordyceps will be boxed on the picture. Compared with the traditional cordyceps collection method, it has the advantages of fast identification efficiency and high accuracy, can effectively reduce the labor intensity of cordyceps collectors, and improve the collection efficiency.

[0034] The present invention solves the problems of high labor intensity, low efficiency, and large loss of cordyceps during the collection by workers at present, thereby improving the collection efficiency and accuracy of cordyceps, realizing the position indication and prompt function, improving the collection efficiency, reducing the labor intensity, reducing the eye fatigue of workers, and the hiking stick can also be used as a support for workers to reduce the fatigue of climbing mountains. Brief Description of the Drawings

[0035] Figure 1 It is a schematic structural diagram of the cordyceps position searching system of the present invention;

[0036] Figure 2 It is a structural diagram of the CA attention mechanism of the present invention;

[0037] Figure 3 It is a flow chart of cordyceps identification and positioning of the present invention;

[0038] Figure 4 It is the cordyceps reconstruction and positioning point cloud of the present invention;

[0039] Figure 5 It is a working principle diagram of the cordyceps position searching system of the present invention.

[0040] In the figure: 1, hiking stick; 2, support crossbeam; 3, CCD camera; 4, controller; 5, battery; 6, camera bracket; 7, controller support frame; 8, battery support frame; 9, tablet computer. Detailed Embodiments

[0041] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0042] Referring to Figures 1 to 5 As shown, a cordyceps position searching system based on binocular stereo and deep learning technology includes: a hardware structure and a software system, wherein the hardware structure includes a hiking stick 1, a support crossbeam 2, a CCD camera 3, a controller 4, a battery 5, a camera bracket 6, a controller support frame 7, a battery support frame 8, and a tablet computer 9;

[0043] The software system includes an image acquisition module, a training and learning module, an identification module, and an interaction module;

[0044] The image acquisition module acquires cordyceps images based on the CCD camera 3 and uploads the acquired cordyceps images into the training and learning module;

[0045] The training and learning module trains the cordyceps image data based on the deep learning framework, learns and optimizes the CA attention mechanism, adds the CA attention mechanism in front of SPPF in the BackBone of the yolov5s model, and constructs a cordyceps identification model;

[0046] The identification module inputs the acquired image data based on the constructed cordyceps identification model, reconstructs the cordyceps point cloud based on the deep learning model, performs three-dimensional positioning, extracts the position of the cordyceps from the color image, and judges and analyzes the position of the cordyceps;

[0047] The interaction module conducts human-computer interaction based on the tablet computer 9. As the handheld terminal of the cordyceps collectors, it is deployed with client software, receives the transmitted cordyceps images in real time, instructs the collectors to collect cordyceps, and emits an indication sound when a cordyceps is recognized.

[0048] This application utilizes the deep learning framework, which can fully explore the deep features and laws contained in the cordyceps image data. The powerful non-linear modeling ability of deep learning can automatically learn the unique appearance features of cordyceps, such as the texture at the combined part of the worm body and the grass body of cordyceps, the overall shape contour, and the difference from the surrounding environment. Compared with traditional identification methods, it greatly improves the accuracy and efficiency of identification. Using the constructed cordyceps identification model, it can not only identify the cordyceps in the image, but also, based on the powerful functions of deep learning, reconstruct the cordyceps point cloud and perform three-dimensional positioning. This means that the position information of cordyceps in the actual space can be accurately obtained. For cordyceps growing in different terrain environments such as mountainous areas with undulating terrain and deep grass, the collectors can more accurately find its specific position through three-dimensional positioning, avoiding the positioning deviation problem that may occur when relying solely on two-dimensional images for judgment.

[0049] One end of the support cross beam 2 is a semi-circular structure and is in transitional fit with the hiking stick 1 and is fixed on the hiking stick 1 through an inner hexagon bolt. The support cross beam 2 is provided with a CCD camera 3, a controller 4, and a battery 5. The battery 5 is used to supply power to the controller 4. The CCD camera 3 is connected to the controller 4 through USB or network cable. The tablet computer 9 is held manually.

[0050] One end of the supporting beam 2 of the present application adopts a semicircular structure to transitionally cooperate with the trekking crutch 1. This design can enable the supporting beam 2 to be tightly and firmly connected to the trekking crutch 1. The semicircular structure can fit the shape of the trekking crutch 1 well, reducing the gap between the two, avoiding looseness, shaking and instability during use, and ensuring that the entire device can maintain a reliable connection when walking and climbing in complex terrain in the wild.

[0051] The CA attention mechanism structure first performs average pooling operations on the acquired image in the horizontal and vertical directions, integrates the feature information in the horizontal and vertical directions, and performs convolution processing on it. The specific steps are:

[0052] Given an input layer X of size H×W×C, using pooling kernels of size (H, 1) and (1, W), each channel is encoded in the horizontal and vertical directions respectively, and the outputs of the c channels of height H and the cth channel of width W are:

[0053]

[0054] Where H and W represent the height and width of the feature layer, respectively, which are known. and Respectively represent the output results of the cth channel in the H and W directions, x c (h,i) and x c (j,w) represents the input feature X along the H direction and the W direction.

[0055] The present application can focus on and capture key information in the input feature layer from different dimensions by encoding each channel horizontally (H direction) and vertically (W direction) respectively. For example, in image data, the input layer X is regarded as an image-related feature representation, H and W correspond to the height and width of the image respectively, and C is the number of channels. For example, in the RGB channel, horizontal pooling can extract the features of the horizontal distribution of the object in the image, such as the morphological and texture continuity features of the Cordyceps main body in the Cordyceps image along the horizontal direction; vertical pooling can capture the key features of the object in the vertical direction, such as the relative position and height features of the Cordyceps in the vertical direction with the surrounding vegetation. This multi-directional feature extraction helps to identify and analyze the object more accurately in the future, dig out more discriminative features, and lay the foundation for accurate classification and positioning tasks.

[0056] The feature maps of the two independent directions output by the above two sets of formulas are concatenated, 1×1 convolutional and activated. The formula is:

[0057] f=σ(F 1 [Z h ,Z w ])

[0058] In the formula, f represents the intermediate feature map after encoding spatial information in the horizontal and vertical directions, [.,.] represents the splicing operation in the spatial dimension, F1 represents the convolution operation, and δ represents the Sigmoid function;

[0059] Split f along the spatial dimension, and then use two 1×1 convolution operations to restore the number of channels of the feature map to the same as that of the input feature X. Calculate the attention weight of the feature map using the Sigmoid function, and the result is:

[0060] g h = σ(F h (f h ))

[0061] g w = σ(F w (f w ))

[0062] In the formula, g h and g w represent the attention weights after convolution and activation of the split f, f h and f w represent the feature maps of the split f along the H direction and the W direction respectively, F h and F w represent the convolution operation.

[0063] Multiply g h and g w by the input feature map weights and output the feature map with coordinate attention weights through weighted summation. The formula is expressed as:

[0064]

[0065] In the formula, y c (i, j) and x c (i, j) represent the input and output of the c-th channel respectively, and represent the attention weights along the H direction and the W direction on the c-th channel respectively.

[0066] The recognition method of the recognition module inputs the collected image data into the Cordyceps recognition model, and uses the bounding box box(x, y, width, height) to represent the position of Cordyceps. x is the value along the horizontal direction of the image coordinate of the upper left point of the rectangle, y is the value along the vertical direction of the image coordinate of the upper left point of the rectangle, width and height are the lengths along the horizontal direction and the vertical direction of the image coordinate of the rectangle respectively, and the bounding box box(x, y, width, height) is the rectangle output after algorithm recognition.

[0067] This application uses bounding boxes to mark the positions of cordyceps, which can very intuitively display the specific areas where cordyceps are located on the image. For cordyceps collectors or relevant monitoring personnel, there is no need for a complex interpretation process. At a glance, they can clearly see the approximate orientation of cordyceps in the picture through the rectangular boxes on the image, clearly know where cordyceps exist, which is convenient for quickly locating and taking corresponding collection or subsequent analysis actions, making the position information of cordyceps obvious at a glance and greatly improving the efficiency of information transmission.

[0068] Based on the depth map of cordyceps obtained by the image acquisition module, a rectangular box (x, y, width, height) is drawn, and the region of interest is extracted from the depth map. It is obtained by using the algorithm of image processing based on the comprehensive rectangular box, and then the formula of binocular vision single-point reconstruction is used to reconstruct the pixel points of the region of interest. The point cloud after reconstruction is the point cloud of cordyceps, that is, the positioning of cordyceps is realized. The formula is:

[0069] x ccs =(u - cx)*zccs / f x

[0070] y ccs =(v - cy)*z ccs / f y

[0071] z ccs =fb / d, d = u l -u r

[0072] In the formula, (x ccs , y ccs , z ccs ) represents the three-dimensional points in the camera coordinate system after reconstruction, (u, v) are the image coordinates of the pixels in the depth map, f is the focal length of the camera, (f x , f y ) are the components of the camera focal length, b is the baseline distance of the camera, d is the disparity of the camera, u l , u r are the projection points of the reconstructed point on the left camera and the right camera respectively.

[0073] By using the binocular vision single-point reconstruction formula, this application can convert the two-dimensional information of cordyceps in the depth map into three-dimensional point cloud information in the camera coordinate system, realizing the leap from a plane to a three-dimensional space. This enables us to accurately know the position of cordyceps in the actual three-dimensional space. Compared with positioning only relying on two-dimensional images, it can more accurately reflect the true location of cordyceps. Whether it is in a mountainous environment with undulations or in different depth positions of the grass, it can be accurately captured, effectively avoiding the positioning error caused by the lack of depth perception in two-dimensional images, and providing extremely accurate position guidance for the subsequent work of cordyceps collection.

[0074] The tablet computer 9, as the carrier of the interaction module, transmits the collected cordyceps images to the client software on the tablet computer 9 through the network transmission link, and intuitively displays the analysis results to show the position of the cordyceps.

[0075] This application timely transmits the collected cordyceps images to the client software on the tablet computer 9 through the network transmission link, enabling the collection workers to obtain the latest cordyceps image information in the first time. Whether the cordyceps images are collected by the cameras or drone devices installed in the surrounding areas, they can be quickly presented on the screen of the tablet computer 9. This means that the collection workers do not need to wait and can know the distribution of cordyceps in each collection area in real time, reducing the time spent on searching for cordyceps and making the collection work more time-efficient and targeted. This intuitive way of showing the position of cordyceps can assist the collection workers to quickly make positioning decisions. In the vast cordyceps collection area, facing the complex terrain and numerous vegetation, it is very difficult to quickly and accurately find cordyceps with the naked eye. However, the cordyceps position information displayed on the tablet computer 9 guides the workers to quickly move towards the target direction and accurately locate the place where the cordyceps is located, avoiding blind search, optimizing the collection path, and further improving the efficiency and accuracy of cordyceps collection.

[0076] The working principle of the cordyceps location system of the present invention is to use a hiking stick for positioning, collect cordyceps images, train a cordyceps recognition model by using a mainstream deep learning framework for the collected cordyceps images to identify and locate the position of the cordyceps, obtain the cordyceps ROI of the depth map based on the rectangular frame, and reconstruct the cordyceps point cloud to achieve three-dimensional positioning.

[0077] The above shows and describes the basic principle, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A Cordyceps positioning system based on binocular stereo and deep learning technology, characterized in that: include: A hardware structure and a software system, wherein the hardware structure includes a mountaineering crutch (1), a supporting beam (2), a CCD camera (3), a controller (4), a battery (5), a camera bracket (6), a controller bracket (7), a battery bracket (8) and a tablet computer (9); The software system includes image acquisition module, training and learning module, recognition module and interaction module; The image acquisition module acquires the Cordyceps image based on the CCD camera (3), and uploads the acquired Cordyceps image to the training and learning module; The training learning module trains Cordyceps image data based on the deep learning framework, learns and optimizes the CA attention mechanism, adds the CA attention mechanism before SPPF in the BackBone of the yolov5s model, and builds a Cordyceps recognition model; The recognition module inputs the collected image data based on the constructed Cordyceps recognition model, reconstructs the Cordyceps point cloud based on the deep learning model, performs three-dimensional positioning, extracts the position of the Cordyceps from the color image, and makes judgment and analysis on the position of the Cordyceps; The interactive module performs human-computer interaction based on a tablet computer (9), which serves as a handheld terminal for Cordyceps collectors and is equipped with client software. It receives the transmitted Cordyceps images in real time, instructs the collectors to collect Cordyceps, and emits an instruction sound when the Cordyceps is identified.

2. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 1 is characterized in that: One end of the support crossbeam (2) is a semicircular structure that transitionally matches the mountaineering crutch (1) and is fixed to the mountaineering crutch (1) via a hexagon socket bolt. A CCD camera (3), a controller (4) and a battery (5) are arranged on the support crossbeam (2). The battery (5) is used to supply power to the controller (4). The CCD camera (3) is connected to the controller (4) via a USB or a network cable. The tablet computer (9) is manually held.

3. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 1 is characterized in that: The CA attention mechanism structure first performs average pooling operations on the acquired image in the horizontal and vertical directions, integrates the feature information in the horizontal and vertical directions, and performs convolution processing on it. The specific steps are: Given an input layer X of size H×W×C, using pooling kernels of size (H, 1) and (1, W), each channel is encoded in the horizontal and vertical directions respectively, and the outputs of the c channels of height H and the cth channel of width W are: Where H and W represent the height and width of the feature layer, respectively, which are known. and Respectively represent the output results of the cth channel in the H and W directions, x c (h,i) and x c (j,w) represents the input feature X along the H direction and the W direction.

4. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 3 is characterized in that: The feature maps of the two independent directions output by the above two sets of formulas are concatenated, 1×1 convolutional and activated. The formula is: f=σ(F1[Z h ,Z w ]) Where f represents the intermediate feature map after encoding the spatial information in the horizontal and vertical directions, [.,.] represents the concatenation operation of the spatial dimension, F1 represents the convolution operation, and δ represents the Sigmoid function; Split f along the spatial dimension, and then use two 1×1 convolution operations to restore the number of channels of the feature map to the same number of channels as the input feature X. Use the Sigmoid function to calculate the attention weight of the feature map, and the result is: g h =σ(F h (f h )) g w =σ(F w (f w )) In the formula, g h and g w represents the attention weight of the convolution and activation after f split, f h and f w They represent the feature maps along the H direction and W direction after f is split, respectively. h and F w Represents a convolution operation.

5. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 4 is characterized in that: G h and g w Multiply the input feature map weights to output a feature map with coordinate attention weights. The formula is: In the formula, y c (i,j) and x c (i,j) represent the input and output of the cth channel respectively, and They represent the attention weights along the H direction and the W direction on the cth channel respectively.

6. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 1 is characterized in that: The recognition module recognition method inputs the collected image data into the Cordyceps recognition model, and uses the bounding box box(x,y,width,height) to represent the position of the Cordyceps. x is the value of the upper left point of the rectangular box along the horizontal direction of the image coordinate, y is the value of the upper left point of the rectangular box along the vertical direction of the image coordinate, width and height are the length of the rectangular box in the horizontal direction and the height in the vertical direction along the image coordinate, respectively. The bounding box box(x,y,width,height) is the rectangular box output after the algorithm recognizes it.

7. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 6 is characterized in that: Based on the image acquisition module, the depth map of Cordyceps is obtained, and a rectangular box (x, y, width, height) is drawn. The region of interest is extracted from the depth map. Based on the comprehensive rectangular box, the image processing algorithm is used to extract it, and then the formula of binocular vision single-point reconstruction is used to reconstruct the pixel points of the region of interest. The reconstructed point cloud is the point cloud of Cordyceps, which means that the location of Cordyceps is realized. The formula is: x ccs =(u-cx)*z ccs / f x y ccs =(v-cy)*z ccs / f y s ccs =fb / d,d=u l -you r In the formula (x ccs ,y ccs , z ccs ) represents the reconstructed 3D point in the camera coordinate system, (u,v) is the image coordinate of the pixel in the depth map, f is the focal length of the camera, (f x , f y ) is the component of the camera focal length, b is the camera baseline distance, d is the camera parallax, u l 、u r are the projection points of the reconstructed points on the left camera and the right camera respectively.

8. The Cordyceps positioning system based on binocular stereo and deep learning technology according to claim 1 is characterized in that: The tablet computer (9) serves as a carrier of the interactive module, transmits the collected Cordyceps images to the client software on the tablet computer (9) through a network transmission link, and intuitively displays the analysis results to show the location of the Cordyceps.

Images