Mineral machinery visual identification device with multiple light sources combined

Abstract

The invention belongs to the technical field of mineral classification, and relates to a multi-light-source combined mineral machinery visual identification device, which comprises a plurality of door-shaped mounting frames, a plurality of light source modules, a plurality of shooting modules and an identification module, and is used for identifying a plurality of ore images to obtain the ore identification rate and the ore type of each ore image. The ore recognition rates of the multiple ore images are ranked from high to low, and the ore type with the highest ore recognition rate is used as the ore type of the marked area on the ore conveying belt. According to the invention, the plurality of shooting modules can shoot a plurality of ore images irradiated by the combined light sources in different combination forms in the same marked area, and the plurality of ore images are identified to obtain the ore identification rate and the ore type of each ore image; the ore type with the highest recognition rate is selected as the ore type of the marked area on the ore conveying belt, and the problem that the accuracy of feature recognition is reduced due to the fact that a single light source irradiates the ore is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of mineral classification, and relates to a mineral mechanical visual identification device with a multi-light source combination. Background Art

[0002] In the mining industry, ore identification is a crucial process, which involves the identification and classification of various minerals in the ore. Through ore identification, the types of major minerals in the ore can be determined, so as to select appropriate mineral processing methods and process flows, improve mineral processing efficiency and concentrate grade recovery rate, and at the same time more accurately assess the impact of ore mining on the environment, formulate effective environmental protection measures, and reduce damage to the environment.

[0003] At present, mechanical vision recognition devices are often used in ore identification. The mechanical vision recognition device is set on the ore conveyor belt, illuminates the ore with a single light source, and then takes a photo for image recognition, extracting features that can represent the characteristics of the ore, such as color, texture, shape, size, etc., to identify the ore.

[0004] However, since mechanical vision recognition devices have high requirements for light sources, if a single light source is used to illuminate the ore, it is easy to cause a decrease in the accuracy of feature recognition, resulting in a decrease in the accuracy of ore identification, affecting the mineral processing efficiency and reducing the concentrate grade recovery rate. Summary of the invention

[0005] The purpose of the present invention is to provide a mineral mechanical visual identification device with a multi-light source combination, which adopts a plurality of combined light sources with different combination forms, can avoid the problem of reduced accuracy of feature identification caused by a single light source irradiating the ore, and improve the accuracy of ore identification.

[0006] To achieve the above purpose, the specific technical solution of a multi-light source combined mineral mechanical visual identification device provided by the present invention is as follows:

[0007] A mineral mechanical visual identification device with a multi-light source combination, comprising a plurality of door-type mounting frames, the plurality of door-type mounting frames being uniformly arranged in sequence above an ore conveyor belt, a plurality of marking areas being arranged in sequence along the length direction of the ore conveyor belt, and further comprising:

[0008] Multiple light source modules correspond to multiple door-type mounting frames one by one. Each light source module is arranged at the top of the corresponding door-type mounting frame. Each light source module can emit combined light of different wavelengths. The combined forms of the multiple combined lights emitted by the multiple light source modules are different.

[0009] A plurality of shooting modules correspond one to one with a plurality of gantry mounting frames, each shooting module is arranged on the inner side wall of the corresponding gantry mounting frame, and the plurality of shooting modules are used to shoot the ore image of the same marked area on the ore conveyor belt entering the corresponding gantry mounting frame.

[0010] The recognition module is electrically connected to the multiple shooting modules, and is used to receive multiple ore images of ore in the same marked area taken by the multiple shooting modules, and recognize the multiple ore images to obtain the ore recognition rate and ore type of each ore image, sort the ore recognition rates of the multiple ore images from high to low, and use the ore type with the highest ore recognition rate as the ore type of the marked area on the ore conveyor belt.

[0011] The present invention is also characterized in that:

[0012] Each light source module includes:

[0013] Two groups of light sources are arranged side by side along the length direction of the corresponding door-type mounting frame. Each group of light sources includes six light sources, and the six light sources can emit white light, red light, green light, blue light, purple light and ultraviolet light in sequence.

[0014] Each shooting module includes:

[0015] Two mounting plates are symmetrically arranged on two side walls in the corresponding door-type mounting frame.

[0016] Two high-speed cameras are respectively arranged at the lower parts of the two mounting plates, and each high-speed camera is electrically connected to the recognition module.

[0017] The identification module includes:

[0018] The image segmentation unit is used to segment multiple ore images, segment each ore image into image units of different sizes, and group the image units of different sizes into an image unit set.

[0019] The gray-level co-occurrence matrix unit is used to perform GLCM processing on image units of different sizes in each image unit set, extract the texture features of each image unit, and the texture features extracted from image units of different sizes form a texture feature set.

[0020] The ore recognition model is used to recognize each texture feature in each texture feature set to obtain the recognition rate and ore type of the ore corresponding to each ore image.

[0021] The output unit is used to sort the recognition rates of multiple ores from high to low, use the ore type with the highest ore recognition rate as the ore type of the marked area, and output the ore type of the marked area.

[0022] The ore recognition model is trained through the following steps:

[0023] Receive multiple ore images of the same marked area taken by multiple shooting modules, and receive them multiple times, and use the multiple ore images received each time as a training set;

[0024] Segment each ore image in each training set and extract texture features of the segmented images, and combine multiple texture features extracted from multiple ore images into a texture feature set;

[0025] Each texture feature in each texture feature set is input into the convolutional neural network VGG-19 model for training to obtain the ore recognition model.

[0026] It also includes a control module, which includes:

[0027] The controller is electrically connected to each light source and each high-speed camera respectively. The controller is used to control the start-up of each light source and each high-speed camera, and control the multiple light source modules to emit multiple combined lights with different combined forms.

[0028] A plurality of infrared transmitters correspond to a plurality of marking areas one by one, each infrared transmitter is arranged in a corresponding marking area on the ore conveyor belt, and each infrared transmitter is used to transmit an infrared signal.

[0029] Multiple infrared receivers correspond one to one with multiple door-type mounting frames, each infrared receiver is arranged in the corresponding door-type mounting frame, each infrared receiver is electrically connected to the controller, and each infrared receiver is used to receive infrared signals emitted by multiple infrared transmitters and feed them back to the controller, and the controller controls the corresponding light source and the high-speed camera to start.

[0030] Each infrared receiver is a cone-shaped structure.

[0031] A fluorescence detector is respectively arranged at a position on the top of each door-shaped mounting frame near the corresponding two groups of light sources, and the fluorescence detector is electrically connected to the corresponding controller.

[0032] The mineral mechanical visual identification device with a multi-light source combination of the present invention has the following advantages:

[0033] First, through the cooperation of multiple light source modules, multiple shooting modules, a control module and an identification module, multiple shooting modules can be used to shoot multiple ore images under the illumination of combined light sources in different combinations of the same marking area, and multiple ore images can be identified to obtain the ore recognition rate and ore type of each ore image, and the ore type with the highest recognition rate is selected as the ore type of the marked area on the ore conveyor belt, thereby enhancing the characteristics of the mineral, avoiding the problem of decreased accuracy of feature recognition caused by a single light source irradiating the ore, and improving the accuracy of ore recognition.

[0034] Second, by setting up a combination of light sources in a variety of different combinations, the complexity of the recognition module algorithm can be reduced, making mineral recognition more efficient and accurate.

[0035] Third, through the cooperation of the controller, multiple infrared transmitters and multiple infrared receivers, multiple light source modules can emit multiple combinations of light in different forms, and control multiple shooting modules to shoot the ore in the same marked area, avoiding errors caused by different marked areas or different mineral surfaces irradiated by the shooting modules, and further improving the accuracy of ore identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of the structure of multiple door-type mounting frames in the present invention.

[0037] Figure 2 It is a schematic diagram of the working principle of the present invention.

[0038] Figure 3 It is a schematic diagram of the main structure inside the door-type mounting frame of the present invention.

[0039] Reference numerals:

[0040] 1. Door-type mounting frame, 2. Light source module, 3. Light source adjustment assembly, 4. Mounting plate, 5. Camera adjustment assembly, 6. High-speed camera, 7. Controller, 8. Infrared receiver. DETAILED DESCRIPTION

[0041] The technical solution in the present application will be described clearly and in detail below in conjunction with the accompanying drawings. Among them, in the description of the embodiments of the present application, unless otherwise specified, " / " means or, for example, A / B can mean A or B: "and / or" in the text is only a description of the association relationship of associated objects, indicating that there can be three relationships, such as A and / or B, which can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, "multiple" refers to two or more than two. The following terms "first" and "second" are used for descriptive purposes only and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0042] like Figure 1 , Figure 2As shown, the present invention provides a mineral mechanical visual recognition device with a multi-light source combination, including multiple gantry mounting frames 1, multiple light source modules, multiple shooting modules, and recognition modules. The multiple gantry mounting frames 1 are uniformly arranged above the ore conveyor belt in sequence, and the ore conveyor belt can pass through each gantry mounting frame 1. Multiple marking areas are sequentially arranged on the ore conveyor belt along its length direction. Multiple light source modules correspond to the multiple gantry mounting frames 1 one by one, and each light source module is arranged at the top of the corresponding gantry mounting frame 1. Each light source module can emit a combination of light rays of different wavelengths. The combination forms of the multiple combination lights emitted by the multiple light source modules are different. Multiple shooting modules correspond to the multiple gantry mounting frames 1 one by one, and each shooting module is arranged on the inner side wall of the corresponding gantry mounting frame 1. The multiple shooting modules are used to shoot the ore image of the same marking area on the ore conveyor belt entering the corresponding gantry mounting frame 1. The recognition module and The multiple shooting modules are electrically connected, and the recognition module is used to receive multiple ore images of the ore in the same marked area taken by the multiple shooting modules, and to recognize the multiple ore images to obtain the ore recognition rate and ore type of each ore image, and to sort the ore recognition rates of the multiple ore images from high to low, and to use the ore type with the highest ore recognition rate as the ore type of the marked area on the ore conveyor belt. Through the cooperation of the multiple light source modules, the multiple shooting modules, the control module and the recognition module, the multiple shooting modules can be used to shoot multiple ore images of the same marked area under the illumination of combined light sources in different combinations, and to recognize the multiple ore images to obtain the ore recognition rate and ore type of each ore image, and to select the ore type with the highest recognition rate as the ore type of the marked area, so as to enhance the characteristics of the mineral, avoid the problem of decreased accuracy of feature recognition caused by irradiating the ore with a single light source, and improve the accuracy of ore recognition.

[0043] like Figure 2 As shown, the recognition module includes an image segmentation unit, a grayscale co-occurrence matrix unit, an ore recognition model and an ore recognition model. The image segmentation unit is used to segment multiple ore images, divide each ore image into image units of different sizes, and form an image unit set with image units of different sizes. The grayscale co-occurrence matrix unit is used to perform GLCM processing on image units of different sizes in each image unit set, extract texture features of each image unit, and the texture features extracted from image units of different sizes form a texture feature set. The ore recognition model is used to identify each texture feature in each texture feature set to obtain the recognition rate and ore type of the ore corresponding to each ore image. The output unit is used to sort the recognition rates of multiple ores from high to low, use the ore type with the highest ore recognition rate as the ore type of the marked area, and output the ore type of the marked area.

[0044] In summary, first of all, when identifying multiple ore images, each ore image is divided into image units of different sizes, and the image texture feature extraction research method based on gray-level co-occurrence matrix (GLCM) is adopted to calculate the texture features to convert the RGB image into a grayscale image. A 5×5 sliding window is selected for eigenvalue calculation. After the sliding window traverses all the image pixels, the entire image forms a texture eigenvalue matrix composed of texture eigenvalues, and then this texture eigenvalue matrix is ​​converted into a texture feature image to improve the generalization ability and robustness of the image.

[0045] Secondly, after extracting the features, a mineral identification model is built. The convolutional neural network VGG-19 model is used to build the mineral identification model. The convolutional neural network architecture proposed in the present invention is designed to process single-channel images with an input size of 224×224 pixels. Regularization technology is used here to prevent image overfitting. The network uses multiple 3×3 convolution kernels and 2×2 pooling kernels with a stride of 2 to gradually extract image features, and finally classifies them through a fully connected layer.

[0046] Finally, the ore texture features are input into the trained ore recognition model for processing to obtain the recognition rate and ore type of the ore in the same marked area under different light sources. Thus, the light source with the best ore recognition effect in the marked area can be obtained by comparison.

[0047] Among them, the ore recognition model is trained through the following steps:

[0048] Receive multiple ore images of the same marked area taken by multiple shooting modules, and receive them multiple times, and use the multiple ore images received each time as a training set;

[0049] Segment each ore image in each training set and extract texture features of the segmented images, and combine multiple texture features extracted from multiple ore images into a texture feature set;

[0050] Each texture feature in each texture feature set is input into the convolutional neural network VGG-19 model for training to obtain the ore recognition model.

[0051] In summary, the specific process of constructing an ore recognition model through a convolutional neural network VGG-19 model in the present invention is as follows:

[0052] Initial convolutional block: The network first passes through two stacked convolutional layers, both of which increase the number of output channels to 64 while keeping the spatial dimension unchanged, that is, the output size is 64×224×224. Subsequently, a maximum pooling layer halves the spatial dimension to 112×112, and the number of output channels remains unchanged, that is, 64×112×112.

[0053] Second convolutional block: This is followed by two stacked convolutional layers that increase the number of output channels to 128 while keeping the spatial dimension at 112 × 112. After that, another max pooling layer halves the spatial dimension to 56 × 56, giving an output channel number of 128 × 56 × 56.

[0054] Dilated convolution block: The network continues through four stacked convolutional layers to increase the number of output channels to 256 and further extract features while keeping the spatial dimension at 56 × 56. Subsequently, a maximum pooling layer halves the spatial dimension to 28 × 28, and the number of output channels is 256 × 28 × 28.

[0055] Deep convolutional block: This is followed by four stacked convolutional layers, increasing the number of output channels to 512, while keeping the spatial dimension at 28 × 28. After that, a max pooling layer is applied again, halving the spatial dimension to 14 × 14, with an output channel number of 512 × 14 × 14. This process is repeated once, but the last max pooling layer reduces the spatial dimension to 7 × 7, with the output channel number still being 512 × 7 × 7.

[0056] Feature flattening: Flatten the output of the last convolutional block (512×7×7) into a one-dimensional vector with a length of 512×7×7=25088.

[0057] Fully connected layers: The flattened vector is fed into the first fully connected layer, which contains 4096 neurons and applies the activation function ReLU. This is followed by the second fully connected layer, which also contains 4096 neurons and ReLU activation. These two fully connected layers are designed to further integrate and abstract features.

[0058] Classification layer: The last fully connected layer contains 1000 neurons but no activation function is applied. The output of this layer is a vector of length 1000 representing the unnormalized log-probability of each class.

[0059] Probability Output: Finally, the softma× function is applied to convert the output of the classification layer into a probability distribution, where each element represents the predicted probability that the input image belongs to the corresponding category.

[0060] like Figure 2As shown, each light source module includes two groups of light sources 2, and the two groups of light sources 2 are arranged side by side along the length direction of the corresponding door-type mounting frame 1. Each group of light sources 2 includes six light sources 2, and the six light sources 2 can emit white light, red light, purple light and ultraviolet light, X-rays and infrared rays in sequence. A fluorescence detector is respectively arranged at the top of each door-type mounting frame 1 near the corresponding two groups of light sources. Some minerals have a fluorescence effect, such as emitting light of a specific color under the irradiation of ultraviolet rays, X-rays, and cathode rays. This effect is due to the electrons inside the minerals jumping to a high energy level under the excitation of external energy, and then releasing energy when falling back to a low energy level to produce fluorescence. For example, fluorite emits blue or purple fluorescence under ultraviolet irradiation. Silica ore can emit bright green fluorescent sticks under ultraviolet irradiation. Calcium uranium ore emits very weak yellow-green fluorescence under ultraviolet irradiation. Calcite emits blue, yellow, green or pink fluorescence under ultraviolet irradiation. Fluorite emits strong green fluorescence under ultraviolet irradiation. Calcium feldspar emits white fluorescence under ultraviolet irradiation. Amethyst will emit purple fluorescence under ultraviolet light. Coralite will emit pink or orange fluorescence under ultraviolet light. Beryl will emit green fluorescence under ultraviolet light. Rhodochrosite will emit orange-yellow fluorescence under ultraviolet light. Scheelite will emit light blue fluorescence under ultraviolet light. Therefore, a fluorescence detection device is set up to better identify the characteristics of these minerals.

[0061] like Figure 3 As shown, the upper part of each light source 2 is connected to the top of the corresponding door-type mounting frame 1 through a light source adjustment component, and the light source adjustment component is used to adjust the irradiation angle of the corresponding light source to facilitate better irradiation of the ore.

[0062] like Figure 3 As shown, each shooting module includes two mounting plates 4 and two high-speed cameras 6. The two mounting plates 4 are symmetrically arranged on the two side walls of the corresponding door-type mounting frame 1. The two high-speed cameras 6 are respectively arranged at the lower part of the two mounting plates 4. Each high-speed camera 6 is electrically connected to the control module and the identification module respectively. Each high-speed camera 6 uses a color model with more than 2 million pixels at 500 frames per second.

[0063] like Figure 3 As shown, a displacement assembly is provided between each mounting plate 4 and the side wall of the corresponding door-type mounting frame 1, and the displacement assembly is used to drive the corresponding mounting plate 4 to move up and down along the side wall of the door-type mounting frame 1. Each high-speed camera 6 is connected to the corresponding mounting plate 4 through a camera adjustment assembly 5, and the camera adjustment assembly 5 is used to adjust the shooting angle of the high-speed camera 6 to ensure that all minerals entering the door-type mounting frame 1 are within the optimal shooting range of the high-speed camera 6, so that the high-speed camera 6 can better capture the picture. The position change of the high-speed camera 6 can shoot minerals at different angles, which helps to reduce the shadows and reflections of irregular minerals and provide clearer and high-contrast images.

[0064] like Figure 2 , Figure 3 As shown, the present invention is a mineral mechanical visual identification device with a multi-light source combination, which also includes a control module. The control module is electrically connected to the multiple light source modules and the multiple shooting modules respectively. The control module is used to control the start-up of the multiple light source modules and the multiple shooting modules, and make the combination forms of the multiple combination lights emitted by the multiple light source modules different, and also make the multiple shooting modules all shoot the ore image of the same marked area. The control module includes a controller 7, multiple infrared transmitters and multiple infrared receivers 8. The controller 7 is electrically connected to each light source 2, each high-speed camera 6, and each fluorescence detector respectively. The controller 7 is used to control the start-up of each light source 2, each high-speed camera 6 and each fluorescence detector, and control the multiple light source modules to emit multiple combination lights with different combination forms. The multiple infrared transmitters correspond to the multiple marking areas one by one. Each infrared transmitter is arranged in the corresponding marking area on the ore conveyor belt. Each infrared transmitter is used to transmit an infrared signal. Two adjacent infrared transmitters The distance between the emitters is 40cm, and a plurality of infrared receivers 8 correspond to a plurality of gantry mounting frames 1 one by one. Each infrared receiver 8 is arranged in a corresponding gantry mounting frame 1, and each infrared receiver 8 is electrically connected to a controller 7. Each infrared receiver 8 is used to receive infrared signals emitted by a plurality of infrared transmitters, and feed them back to the controller 7. The controller 7 controls the corresponding light source 2 and the high-speed camera 6 to start. Each infrared transmitter moves along with the movement of the ore conveyor belt, and each time it enters a gantry mounting frame 1, the infrared receiver 8 inside the gantry mounting frame 1 receives the infrared signal of the infrared transmitter and forms an electrical signal, which is fed back to the controller 7. The controller 7 controls the corresponding light source 2 and the high-speed camera 6 to start, so that the marking area photographed by the high-speed camera 6 in each gantry mounting frame 1 is the same, thereby better highlighting the influence of combined light sources of different combinations on the mineral, and reducing the error caused by different marking areas or different mineral surfaces illuminated by the camera.

[0065] like Figure 3 As shown, each infrared receiver is a conical structure, which can effectively prevent the ore from falling onto the infrared sensor and blocking the infrared sensor signal, thereby facilitating the controller 7 to control the start-up of each light source, each high-speed camera 6 and the fluorescence detector in the corresponding gantry mounting frame 1, thereby facilitating the identification of the type of mineral.

[0066] like Figure 1 As shown, the present invention is a mineral mechanical visual identification device with a multi-light source combination, which also includes a mineral spreading mechanism. The mineral uniform spreading device is arranged on the ore conveyor belt. The mineral uniform spreading device is used to evenly arrange the ore on the ore conveyor belt so that each ore pile is a quadrilateral.

[0067] The mineral feeding device includes a plurality of feeding troughs and a plurality of partitions. The plurality of feeding troughs are arranged side by side at the upper end of the ore conveyor belt and close to one end of the ore conveyor belt. The plurality of partitions are arranged on the ore conveyor belt and are evenly arranged along the length direction of the ore conveyor belt. Each partition is arranged along the width direction of the ore conveyor belt. Each infrared transmitter is arranged between two adjacent partitions, so that the ore conveyor belt can be divided into a plurality of squares by the plurality of partitions, so that the ore falls on the ore conveyor belt after coming down from the plurality of feeding troughs, forming a plurality of quadrilateral ore piles. At the same time, it is also convenient for a plurality of shooting modules to shoot ore images of the same marked area.

[0068] Working principle: When in use, the mineral is placed on the ore conveyor belt, and the ore conveyor belt transports the mineral on it into multiple gantry mounting frames 1 in sequence. When entering the first gantry mounting frame 1, the infrared signal emitted by the first infrared transmitter on the ore conveyor belt is received by the infrared receiver 8 in the first gantry mounting frame 1, and converted into an electrical signal and fed back to the controller 7. The controller 7 controls the two groups of light sources 2 and the two high-speed cameras 6 in the first gantry mounting frame 1 to start. The two groups of light sources 2 form a combined light to illuminate the ore in the first marked area corresponding to the first infrared transmitter on the ore conveyor belt, and then pass through the multiple gantry mounting frames 1 behind in sequence. Multiple images of the ore in the first marked area are taken under different combinations of combined light. By analogy, multiple images of the ore in the second marked area corresponding to the second infrared transmitter and the ore in the third marked area corresponding to the third infrared transmitter are taken in sequence.

[0069] Then, multiple images under the illumination of combined light of different combinations of the same marked area are segmented, each ore image is segmented into image units of different sizes, and the image units of different sizes are combined into an image unit set, and then the image units of different sizes in each image unit set are processed by GLCM through the gray level co-occurrence matrix unit to extract the texture features of each image unit, and then the multiple texture features extracted from each image unit set are identified through the ore recognition model to obtain the recognition rate and ore type of the ore corresponding to each ore image, and finally the recognition rates of multiple ores are sorted from high to low through the output unit, and the ore type with the highest ore recognition rate is used as the ore type of the marked area, and the ore type of the marked area is output.

[0070] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

Claims

1. A mineral mechanical visual identification device with a multi-light source combination, comprising a plurality of door-type mounting frames (1), wherein the plurality of door-type mounting frames (1) are uniformly arranged in sequence above an ore conveyor belt, and a plurality of marking areas are arranged in sequence along the length direction of the ore conveyor belt, characterized in that: Also includes: A plurality of light source modules correspond one to one with the plurality of door-shaped mounting frames (1), each light source module is arranged at the top of the corresponding door-shaped mounting frame (1), each light source module can emit a combination of light rays of different wavelengths, and the combination forms of the plurality of combination light rays emitted by the plurality of light source modules are different; A plurality of camera modules corresponding one to the plurality of door-shaped mounting frames (1), each camera module being arranged on the inner side wall of the corresponding door-shaped mounting frame (1), and the plurality of camera modules being used to capture images of ore in the same marked area on the ore conveyor belt entering the corresponding door-shaped mounting frame (1); The identification module is electrically connected to the multiple shooting modules, and is used to receive multiple ore images of the ore in the same marked area taken by the multiple shooting modules, and identify the multiple ore images to obtain the ore recognition rate and ore type of each ore image, sort the ore recognition rates of the multiple ore images from high to low, and use the ore type with the highest ore recognition rate as the ore type of the marked area.

2. The mineral mechanical visual identification device with multiple light sources according to claim 1 is characterized in that: Each of the light source modules comprises: Two groups of light sources (2) are arranged side by side along the length direction of the corresponding door-shaped mounting frame (1), each group of light sources (2) comprises six light sources (2), and the six light sources (2) can emit white light, red light, green light, blue light, purple light and ultraviolet light in sequence.

3. The mineral mechanical visual identification device with multiple light sources according to claim 2 is characterized in that: Each of the shooting modules comprises: Two mounting plates (4) are symmetrically arranged on two side walls of the corresponding door-shaped mounting frame (1); Two high-speed cameras (6) are respectively arranged at the bottom of the two mounting plates (4), and each of the high-speed cameras (6) is electrically connected to the recognition module.

4. The mineral mechanical visual identification device with multiple light sources according to claim 1 is characterized in that: The identification module comprises: An image segmentation unit is used to segment a plurality of ore images, segment each ore image into image units of different sizes, and group the image units of different sizes into an image unit set; Gray-level co-occurrence matrix unit, used to perform GLCM processing on image units of different sizes in each image unit set, extract texture features of each image unit, and the texture features extracted from image units of different sizes form a texture feature set; The ore recognition model is used to recognize each texture feature in each texture feature set, and obtain the recognition rate and ore type of the ore corresponding to each ore image; The output unit is used to sort the recognition rates of multiple ores from high to low, use the ore type with the highest ore recognition rate as the ore type of the marked area, and output the ore type of the marked area.

5. The mineral mechanical visual identification device with multiple light sources according to claim 4 is characterized in that: The ore identification model is obtained by training through the following steps: Receive multiple ore images of the same marked area taken by multiple shooting modules, and receive them multiple times, and use the multiple ore images received each time as a training set; Segment each ore image in each training set and extract texture features of the segmented images, and combine multiple texture features extracted from multiple ore images into a texture feature set; Each texture feature in each texture feature set is input into the convolutional neural network VGG-19 model for training to obtain the ore recognition model.

6. The mineral mechanical visual identification device with multiple light sources according to claim 3 is characterized in that: Also included is a control module, the control module comprising: A controller (7) is electrically connected to each light source (2) and each high-speed camera (6) respectively, and the controller (7) is used to control the start-up of each light source (2) and each high-speed camera (6), and control the multiple light source modules to emit multiple combined light beams in different combined forms; A plurality of infrared transmitters corresponding to the plurality of marking areas one by one, each infrared transmitter being arranged in a corresponding marking area on the ore conveyor belt, and each infrared transmitter being used to transmit an infrared signal; A plurality of infrared receivers (8) correspond one to one with a plurality of door-type mounting frames (1), each infrared receiver (8) is arranged in a corresponding door-type mounting frame (1), each infrared receiver (8) is electrically connected to a controller (7), each infrared receiver (8) is used to receive infrared signals emitted by a plurality of infrared transmitters, and feed the infrared signals back to the controller (7), and the controller (7) controls the corresponding light source (2) and the high-speed camera (6) to start.

7. The mineral mechanical visual identification device with multiple light sources according to claim 6 is characterized in that: Each infrared receiver (8) is of a conical structure.

8. The mineral mechanical visual identification device with multiple light sources according to claim 6 is characterized in that: A fluorescence detector is provided at the top of each door-shaped mounting frame (1) near the corresponding two groups of light sources, and the fluorescence detector is electrically connected to the corresponding controller (7).

Images