A brain vision detection method based on neurofeedback for scientific research laboratories

Abstract

The present invention discloses a brain vision detection method based on neurofeedback for scientific research laboratories, comprising the following steps: fusing the depth information of a visual target on the basis of detecting the two-dimensional position information of an image to achieve visual target detection in a three-dimensional space and obtain the three-dimensional coordinates of the visual target; fitting a response model of brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target. The present invention takes a deep learning algorithm as the core and integrates the depth information of traditional machine vision, solves the deficiency that a single deep learning algorithm can only obtain two-dimensional coordinate information, and obtains complete three-dimensional coordinates; fitting a response model of brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target for visual detection of the brain vision to be predicted, and the model recognition improves the efficiency of visual detection and reduces the complexity.

Description

Technical Field

[0001] The present invention relates to the technical field of visual detection, and particularly relates to a method for cerebral vision detection based on neurofeedback for scientific research laboratories. Background Art

[0002] "Cerebral vision" in English is Cerebral vision. Vision is translated as "vision", which refers to the subjective sensation obtained through the peripheral sensory organ (eye) of the visual system, receiving electromagnetic wave stimulation within a certain wavelength range in the external environment, and being encoded, processed, and analyzed by relevant parts of the central nervous system. It is the sensation and cognition of visible light and the most important means for humans to perceive the world and obtain information. Cerebral is translated as "of the brain" or "cerebral". In terms of the visual system, it should be the system for processing and analyzing after the conversion of light signals into nerve signals. The retina not only has the function of converting light signals but also participates in the processing of the converted information. Moreover, due to the retina having a multi-layered ordered structure similar to the cerebral cortex and the same embryonic development origin as the brain, in the postgraduate textbook "Neurobiology" used in national higher medical colleges and universities, the retina is also called the "peripheral brain". Therefore, "cerebral vision" should be the most important functional manifestation of the entire visual nervous system from the retina to the visual center, and the detection of cerebral vision is also crucial.

[0003] The prior art CN201910704737.4 discloses a device and method for cerebral vision detection and analysis based on neurofeedback, including a detection device, a first server, and an electro-acupuncture device; the detection device is used to obtain the physiological data of the user and the current visual function parameters of the user; the first server is used to update the current training samples and the current neural network model according to the current visual function parameters and physiological data; through the updated neural network model, the type and level of visual impairment of the user are output; the first server is also used to determine electro-acupuncture data according to the type and level of visual impairment; according to the electro-acupuncture data, the corresponding electro-acupuncture device operation data is matched for the user; the electro-acupuncture device is used to perform electro-acupuncture for the user according to the electro-acupuncture device operation data. This application can not only adjust the electro-acupuncture parameters in real time to adapt to the current visual condition of the user but also optimize the acupuncture points and combinations of acupuncture points.

[0004] Although the prior art has achieved the detection of cerebral vision using neurofeedback to a certain extent, it is difficult to achieve visual recognition in three-dimensional space, which does not conform to the three-dimensional characteristics of brain imaging, resulting in low reliability of the detection effect. Summary of the Invention

[0005] The object of the present invention is to provide a brain vision detection method based on neurofeedback for scientific research laboratories, so as to solve the technical problems in the prior art that it is difficult to achieve visual recognition in three-dimensional space, which does not conform to the three-dimensional characteristics of brain imaging, resulting in low reliability of the detection effect.

[0006] To solve the above technical problems, the present invention specifically provides the following technical solutions:

[0007] A brain vision detection method based on neurofeedback for scientific research laboratories includes the following steps:

[0008] Step S1: Select a group of sample individuals, and collect the optic nerve feedback signals of the sample individuals and the left view and right view of the visual detection scene in the visual detection scene. Use the deep learning object detection algorithm to learn the visual detection scene in the left view of the scene to obtain an autonomous visual detection model, obtain a detection image, and obtain the two-dimensional position information of the visual target.

[0009] Step S2: Perform stereo matching on the left view and right view of the scene through the semi-global stereo matching algorithm SGBM to complete the global depth perception of the visual detection scene, obtain a depth image, obtain the depth information of the visual target, and fuse the depth information of the visual target on the basis of the two-dimensional position information of the detection image to realize the detection of the visual target in the three-dimensional space and obtain the three-dimensional coordinates of the visual target.

[0010] Step S3: Fit a response model of brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target, so as to be used for visual detection of the brain vision to be predicted.

[0011] As a preferred solution of the present invention, the step of using the deep learning object detection algorithm to learn the visual detection scene in the left view of the scene to obtain an autonomous visual detection model includes:

[0012] Predict in the area where the visual target may appear in the left view of the scene according to the anchor box anchor according to the preset size ratio relationship, fine-tune the position of the visual target prediction box according to the position regression algorithm, and then use the non-maximum suppression algorithm to screen out the optimal visual target prediction box to obtain a detection image representing the visual target.

[0013] Select the left view of the scene as the input image of the deep learning object detection algorithm, select the detection image as the output image of the deep learning object detection algorithm, and use the deep learning object detection algorithm to perform algorithm training based on the input image and the output image to obtain the autonomous visual detection model. The model expression of the autonomous visual detection model is:

[0014] G = YOLO V5(g);

[0015] In the formula, g is the left view of the scene, G is the detection image, and YOLO V5 is a deep learning object detection algorithm.

[0016] As a preferred solution of the present invention, obtaining the depth information of the visual target includes:

[0017] In the semi-global stereo matching algorithm SGBM, horizontal gradient preprocessing is performed on the input left view and right view of the scene. According to the respective corresponding positions of the same pixel point in the visual detection scene in the left view and right view of the scene, cost aggregation calculation, disparity calculation, and post-processing disparity optimization are sequentially performed to complete the global depth perception of the obstacle avoidance scene, obtain a depth image, and obtain the depth information of the visual target from the depth image.

[0018] As a preferred solution of the present invention, obtaining the three-dimensional coordinates of the visual target includes:

[0019] Based on the one-to-one mapping relationship between the detection image and the depth image in the two-dimensional space, the two-dimensional position information of the detection image is decoded to obtain the two-dimensional position coordinates of the visual target under the left view of the scene;

[0020] Through the two-dimensional position coordinates of the detection image, the depth information corresponding to the two-dimensional position coordinates of the visual target is obtained in the depth image by coordinate indexing, the depth coordinates corresponding to the two-dimensional position coordinates are decoded, and the two-dimensional position coordinates of the visual target and the depth coordinates of the visual target are combined to obtain the three-dimensional coordinates of the visual target.

[0021] As a preferred solution of the present invention, fitting the response model of the brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target includes:

[0022] Taking the three-dimensional coordinates of the visual target as the input item of the BP neural network and the optic nerve feedback signal as the output item of the BP neural network, and using the BP neural network to perform convolutional training on the input item and the output item to obtain the response model of the brain vision representing the mapping relationship between the three-dimensional coordinates of the visual target and the optic nerve feedback signal. The model expression of the response model of the brain vision is:

[0023] S = BP([X, Y, Z]);

[0024] In the formula, S is the optic nerve feedback signal, [X, Y, Z] is the three-dimensional coordinates of the visual target, and BP is the BP neural network;

[0025] Among them, the three-dimensional coordinates of the visual target belonging to the left view and right view of the scene and the optic nerve feedback signal are derived from the simultaneous acquisition of the same visual target.

[0026] As a preferred embodiment of the present invention, the visual detection of the brain vision to be predicted includes:

[0027] Placing the individual to be detected in a visual detection scenario, and collecting the true value of the optic nerve feedback signal of the individual to be detected and the left view and right view of the visual detection scenario in the visual detection scenario;

[0028] Using the autonomous visual detection model to obtain a detection image representing the visual target in the left view of the scenario, performing stereo matching on the left view and right view of the scenario through the semi-global stereo matching algorithm SGBM to obtain a depth image and the depth information of the visual target, and fusing the depth information of the visual target based on the two-dimensional position information of the detection image to obtain the three-dimensional coordinates of the visual target;

[0029] Using the response model of the brain vision to obtain the predicted value of the optic nerve feedback signal of the individual to be detected based on the three-dimensional coordinates of the visual target, and performing a similarity detection on the true value of the optic nerve feedback signal of the individual to be detected and the predicted value of the optic nerve feedback signal of the individual to be detected, where

[0030] If the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal is lower than a preset threshold, the brain vision of the individual to be detected is marked as an abnormal state;

[0031] If the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal is higher than or equal to the preset threshold, the brain vision of the individual to be detected is marked as a normal state;

[0032] The calculation formula for the similarity detection is:

[0033] ;

[0034] In the formula, I is the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal, S r is the vector form of the true value of the optic nerve feedback signal, S is the vector form of the predicted value of the optic nerve feedback signal, and T is the transpose operator.

[0035] As a preferred embodiment of the present invention, when the brain vision of the individual to be detected is in an abnormal state, the degree of abnormality is quantified using the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal, and the calculation formula for the degree of abnormality is:

[0036] ;

[0037] In the formula, P is the degree of abnormality, |S r -S| is S rThe vector distance between and S, and |S| is the vector distance between S and the 0 vector.

[0038] As a preferred embodiment of the present invention, the left view of the scene and the right view of the scene are respectively collected by the left camera and the right camera of the binocular camera. The binocular camera is consistent with the binocular positions of the sample individual or the individual to be detected, so as to realize the visual detection simulation of the sample individual or the individual to be detected in the visual detection scene.

[0039] As a preferred embodiment of the present invention, the left view of the scene and the right view of the scene are both adjusted to the same size specification.

[0040] As a preferred embodiment of the present invention, the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal are normalized before calculation.

[0041] The present invention has the following beneficial effects compared with the prior art:

[0042] The present invention uses deep learning algorithms and binocular stereo matching to realize the detection of visual objects in a three-dimensional space, avoiding the tediousness of extracting features in the manual extraction of visual object feature algorithms. At the same time, it improves the accuracy and speed of image processing, enabling the brain visual detection to have an end-to-end output ability and making obstacle avoidance more intelligent. With deep learning algorithms as the core, it integrates the depth information of traditional machine vision, solves the deficiency that a single deep learning algorithm can only obtain two-dimensional coordinate information, and obtains complete three-dimensional coordinates. According to the optic nerve feedback signal and the three-dimensional coordinates of the visual object, a response model of the brain vision is fitted to be used for visual detection of the brain vision to be predicted. The model recognition improves the efficiency of visual detection and reduces the tediousness. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can also be obtained according to the provided drawings.

[0044] Figure 1 It is a flowchart of the brain vision detection method based on neural feedback provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] As Figure 1 shown, the present invention provides a brain vision detection method based on neurofeedback for a scientific research laboratory, including the following steps:

[0047] Step S1: Select a group of sample individuals, and collect the optic nerve feedback signals of the sample individuals and the left scene view and right scene view of the visual detection scene in the visual detection scene. Use the deep learning object detection algorithm to learn the visual detection scene in the left scene view to obtain an autonomous visual detection model, obtain a detection image, and obtain the two-dimensional position information of the visual target. The visual target is an item placed in the visual detection scene for detecting brain vision;

[0048] Using the deep learning object detection algorithm to learn the visual detection scene in the left scene view to obtain an autonomous visual detection model, including:

[0049] Predict in the area where the visual target may appear in the left scene view according to the anchor box anchor according to the preset size ratio relationship, fine-tune the position of the visual target prediction box according to the position regression algorithm, and then use the non-maximum suppression algorithm to screen out the optimal visual target prediction box to obtain a detection image representing the visual target;

[0050] Select the left scene view as the input image of the deep learning object detection algorithm, select the detection image as the output image of the deep learning object detection algorithm, and use the deep learning object detection algorithm to perform algorithm training based on the input image and the output image to obtain an autonomous visual detection model. The model expression of the autonomous visual detection model is:

[0051] G = YOLO V5(g);

[0052] In the formula, g is the left scene view, G is the detection image, and YOLO V5 is the deep learning object detection algorithm.

[0053] The left scene view and the right scene view are respectively collected by the left camera and the right camera of the binocular camera. The binocular camera is in the same binocular position as the sample individual or the individual to be detected to realize the visual detection simulation of the sample individual or the individual to be detected in the visual detection scene.

[0054] The binocular camera is used to simulate the human eyes of an individual, identify the position information of the visual target in the visual detection scene, and construct an autonomous visual detection model. The two-dimensional position information of the visual target can be directly obtained from the left view of the scene captured by the binocular camera. When mapping to a sample individual or an individual to be detected, the position information of the visual target at the eyes of the sample individual or the individual to be detected can be accurately obtained, avoiding the cumbersome extraction of features in the manual extraction of the visual target feature algorithm. At the same time, the accuracy and speed of image processing are improved, enabling the brain visual detection to have an end-to-end output ability.

[0055] Step S2: Perform stereo matching on the left view and the right view of the scene through the semi-global stereo matching algorithm SGBM to complete the global depth perception of the visual detection scene, obtain a depth image, and get the depth information of the visual target. Based on the two-dimensional position information of the detection image, fuse the depth information of the visual target to achieve the detection of the visual target in the three-dimensional space and obtain the three-dimensional coordinates of the visual target.

[0056] Obtaining the depth information of the visual target includes:

[0057] In the semi-global stereo matching algorithm SGBM, perform horizontal gradient preprocessing on the input left view and right view of the scene. According to the respective corresponding positions of the same pixel point in the left view and the right view of the visual detection scene, perform cost aggregation calculation, disparity calculation, and post-processing disparity optimization in sequence to complete the global depth perception of the obstacle avoidance scene, obtain a depth image, and obtain the depth information of the visual target from the depth image.

[0058] Obtaining the three-dimensional coordinates of the visual target includes:

[0059] [[ID=@15]]Based on the one-to-one mapping relationship between the detection image and the depth image in the two-dimensional space, perform coordinate decoding on the two-dimensional position information of the detection image to obtain the two-dimensional position coordinates of the visual target under the left view of the scene;

[0060] Through the two-dimensional position coordinates of the detection image, obtain the depth information corresponding to the two-dimensional position coordinates of the visual target in the depth image in a coordinate indexing manner, decode to obtain the depth coordinates corresponding to the two-dimensional position coordinates, and combine the two-dimensional position coordinates of the visual target and the depth coordinates of the visual target to obtain the three-dimensional coordinates of the visual target.

[0061] Obtaining the three-dimensional coordinates of the visual target can restore the three-dimensional features in the real scene, making up for the deficiency that the deep learning model, i.e., the autonomous visual detection model, can only obtain two-dimensional coordinate information. Obtaining the three-dimensional features of the real scene conforms to the stereoscopic characteristics of human eye recognition. Thus, it realizes the process of obtaining the three-dimensional features of the visual target by the human eye through the three-dimensional feature recognition and mapping of the left scene image and the right scene view collected by the binocular camera. Furthermore, by fitting the three-dimensional coordinates of the visual target with the optic nerve feedback signal, a universal mapping relationship between the three-dimensional coordinates of the visual target and the optic nerve feedback signal can be obtained. That is, under the normal state of the brain vision, the three-dimensional coordinates of the visual target are [X, Y, Z], and the corresponding optic nerve feedback signal should be Sa. If the measured optic nerve feedback signal is not Sa, it indicates that the brain vision of the individual to be detected is abnormal, and the detection of the brain vision can be completed.

[0062] Step S3: Fit a response model of the brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target for visual detection of the to-be-predicted brain vision.

[0063] Fitting a response model of the brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target includes:

[0064] Taking the three-dimensional coordinates of the visual target as the input item of the BP neural network and the optic nerve feedback signal as the output item of the BP neural network, and using the BP neural network to perform convolutional training on the input item and the output item to obtain a response model of the brain vision representing the mapping relationship between the three-dimensional coordinates of the visual target and the optic nerve feedback signal. The model expression of the response model of the brain vision is:

[0065] S = BP([X, Y, Z]);

[0066] In the formula, S is the optic nerve feedback signal, [X, Y, Z] are the three-dimensional coordinates of the visual target, and BP is the BP neural network;

[0067] Among them, the sum of the left scene view and the right scene view of the scene to which the three-dimensional coordinates of the visual target belong and the optic nerve feedback signal are derived from the simultaneous acquisition of the same visual target.

[0068] Performing visual detection on the to-be-predicted brain vision includes:

[0069] Placing the individual to be detected in the visual detection scene, and collecting the true value of the optic nerve feedback signal of the individual to be detected and the left scene view and the right scene view of the visual detection scene in the visual detection scene;

[0070] Use the autonomous vision detection model to obtain the detection image representing the visual target in the left view of the scene. Perform stereo matching on the left view and the right view of the scene through the semi-global stereo matching algorithm SGBM to obtain the depth image and the depth information of the visual target. Based on the two-dimensional position information of the detection image, fuse the depth information of the visual target to obtain the three-dimensional coordinates of the visual target;

[0071] Use the response model of the brain vision to obtain the predicted value of the optic nerve feedback signal of the individual to be detected based on the three-dimensional coordinates of the visual target. Perform similarity detection on the true value of the optic nerve feedback signal of the individual to be detected and the predicted value of the optic nerve feedback signal of the individual to be detected. Among them,

[0072] If the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal is lower than the preset threshold, then mark the brain vision of the individual to be detected as an abnormal state;

[0073] If the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal is higher than or equal to the preset threshold, then mark the brain vision of the individual to be detected as a normal state;

[0074] The calculation formula for similarity detection is:

[0075] ;

[0076] In the formula, I is the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal, S r is the vector form of the true value of the optic nerve feedback signal, S is the vector form of the predicted value of the optic nerve feedback signal, and T is the transpose operator.

[0077] When the brain vision of the individual to be detected is in an abnormal state, then quantify the degree of abnormality using the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal. The calculation formula for the degree of abnormality is:

[0078] ;

[0079] In the formula, P is the degree of abnormality, |S r -S| is the vector distance between S r and S, and |S| is the vector distance between S and the 0 vector.

[0080] The response model of the brain vision can predict the optic nerve feedback signal under normal brain vision conditions only through the three-dimensional position coordinates of the visual target, possessing the ability of autonomous learning in the visual detection scenario and avoiding the complicated three-dimensional reconstruction process. That is, through a certain amount of data training, the response model can automatically learn the optic nerve feedback signals of the visual target in all visual detection scenarios. Moreover, the higher the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal, the closer the brain vision of the individual to be detected is to the normal state. On the contrary, the lower the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal, the farther the brain vision of the individual to be detected is from the normal state, that is, the abnormal state. And the degree of abnormality can be obtained by calculating the deviation between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal, thus realizing the detection of the brain vision condition of the individual to be detected.

[0081] Both the left view and the right view of the scene are adjusted to the same size specification.

[0082] The true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal are normalized before calculation.

[0083] The present invention uses deep learning algorithms and binocular stereo matching to realize the detection of visual targets in three-dimensional space, avoiding the tediousness of feature extraction in the manual extraction of visual target feature algorithms, while improving the accuracy and speed of image processing, enabling the brain vision detection to have end-to-end output capabilities and making obstacle avoidance more intelligent; with deep learning algorithms as the core, integrating the depth information of traditional machine vision, solving the deficiency that a single deep learning algorithm can only obtain two-dimensional coordinate information, and obtaining complete three-dimensional coordinates; fitting a response model of the brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target for visual detection of the brain vision to be predicted, and the model recognition improves the efficiency of visual detection and reduces the tediousness.

[0084] In this embodiment, the visual detection is changed from two-dimensional space to three-dimensional space, and various models are used for accurate recognition and operation, which more meets the requirements of accuracy corresponding to scientific research in the laboratory. Therefore, it can be applied to scientific research laboratories for use.

[0085] The above embodiments are only exemplary embodiments of the present application and are not used to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements within the essence and protection scope of the present application, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present application.

Claims

1. A brain vision detection method based on neurofeedback for scientific research laboratories, characterized in that: Including the following steps: Step S1: Select a group of sample individuals, and collect the optic nerve feedback signals of the sample individuals and the left scene view and right scene view of the visual detection scene in the visual detection scene. Use the deep learning object detection algorithm to learn the visual detection scene in the left scene view to obtain an autonomous visual detection model, obtain a detection image, and obtain the two-dimensional position information of the visual target; Step S2: Perform stereo matching on the left scene view and the right scene view through the semi-global stereo matching algorithm SGBM to complete the global depth perception of the visual detection scene, obtain a depth image, obtain the depth information of the visual target, and fuse the depth information of the visual target on the basis of the two-dimensional position information of the detection image to achieve visual target detection in the three-dimensional space and obtain the three-dimensional coordinates of the visual target; Step S3: Fit a response model of the brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target for visual detection of the brain vision to be predicted; Fitting a response model of the brain vision according to the optic nerve feedback signal and the three-dimensional coordinates of the visual target includes: Taking the three-dimensional coordinates of the visual target as the input item of the BP neural network, taking the optic nerve feedback signal as the output item of the BP neural network, and using the BP neural network to perform convolutional training on the input item and the output item to obtain a response model of the brain vision representing the mapping relationship between the three-dimensional coordinates of the visual target and the optic nerve feedback signal. The model expression of the response model of the brain vision is: S = BP([X, Y, Z]); In the formula, S is the optic nerve feedback signal, [X, Y, Z] is the three-dimensional coordinates of the visual target, and BP is the BP neural network; Among them, the sum of the left scene view and the right scene view of the three-dimensional coordinates of the visual target and the optic nerve feedback signal are derived from the simultaneous acquisition of the same visual target; Performing visual detection on the brain vision to be predicted includes: Placing the individual to be detected in the visual detection scene, and collecting the true value of the optic nerve feedback signal of the individual to be detected and the left scene view and right scene view of the visual detection scene in the visual detection scene; Using the autonomous visual detection model to obtain a detection image representing the visual target in the left scene view, performing stereo matching on the left scene view and the right scene view through the semi-global stereo matching algorithm SGBM, obtaining a depth image and the depth information of the visual target, and fusing the depth information of the visual target on the basis of the two-dimensional position information of the detection image to obtain the three-dimensional coordinates of the visual target; Using the response model of the brain vision to obtain the predicted value of the optic nerve feedback signal of the individual to be detected based on the three-dimensional coordinates of the visual target, and performing a similarity detection on the true value of the optic nerve feedback signal of the individual to be detected and the predicted value of the optic nerve feedback signal of the individual to be detected. Among them, If the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal is lower than the preset threshold, the brain vision of the individual to be detected is marked as an abnormal state; If the similarity between the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal is higher than or equal to the preset threshold, the brain vision of the individual to be detected is marked as a normal state; The calculation formula for similarity detection is as follows: ； where I is the similarity between the true value and the predicted value of the optic nerve feedback signal, S r is the vector form of the true value of the optic nerve feedback signal, S is the vector form of the predicted value of the optic nerve feedback signal, and T is the transpose operator.

2. The method for brain vision detection based on neurofeedback for a scientific research laboratory according to claim 1, wherein: The use of the deep learning object detection algorithm to learn the visual detection scene in the left view of the scene to obtain an autonomous visual detection model includes: Predict in the area where the visual target object may appear in the left view of the scene according to the anchor box anchor with a preset size ratio relationship, fine-tune the position of the visual target object prediction box according to the position regression algorithm, and then use the non-maximum suppression algorithm to screen out the optimal visual target object prediction box to obtain a detection image representing the visual target object; Select the left view of the scene as the input image of the deep learning object detection algorithm, select the detection image as the output image of the deep learning object detection algorithm, and use the deep learning object detection algorithm to perform algorithm training based on the input image and the output image to obtain the autonomous visual detection model. The model expression of the autonomous visual detection model is: G = YOLO V5(g); In the formula, g is the left view of the scene, G is the detection image, and YOLO V5 is the deep learning object detection algorithm.

3. The method for brain vision detection based on neurofeedback for a scientific research laboratory according to claim 2, characterized in that: The obtaining of the depth information of the visual target object includes: In the semi-global stereo matching algorithm SGBM, perform horizontal gradient preprocessing on the input left view of the scene and the right view of the scene. According to the respective corresponding positions of the same pixel point in the visual detection scene in the left view of the scene and the right view of the scene, perform cost aggregation calculation, disparity calculation, and post-processing disparity optimization in sequence to complete the global depth perception of the obstacle avoidance scene, obtain a depth image, and obtain the depth information of the visual target object from the depth image.

4. A brain vision detection method based on neurofeedback for scientific research laboratories according to claim 3, characterized in that: The obtaining of the three-dimensional coordinates of the visual target object includes: Based on the one-to-one mapping relationship between the detection image and the depth image in the two-dimensional space, decode the two-dimensional position information of the detection image to obtain the two-dimensional position coordinates of the visual target object under the left view of the scene; Obtain the depth information corresponding to the two-dimensional position coordinates of the visual target object in the depth image through the two-dimensional position coordinates of the detection image in a coordinate indexing manner, decode to obtain the depth coordinates corresponding to the two-dimensional position coordinates, and combine the two-dimensional position coordinates of the visual target object and the depth coordinates of the visual target object to obtain the three-dimensional coordinates of the visual target object.

5. A brain vision detection method based on neurofeedback for scientific research laboratories according to claim 4, characterized in that, When the brain vision of the individual to be detected is in an abnormal state, the abnormal degree is quantified using the true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal. The calculation formula for the abnormal degree is: ； where P is the degree of abnormality, |S r - S| is the vector distance between S r and S, and |S| is the vector distance between S and the 0 vector.

6. A brain vision detection method based on neurofeedback for a scientific research laboratory according to claim 5, characterized in that, The left view of the scene and the right view of the scene are respectively collected by the left camera and the right camera of the binocular camera. The binocular camera is in the same binocular position as the sample individual or the individual to be detected to realize the visual detection simulation of the sample individual or the individual to be detected in the visual detection scene.

7. A brain vision detection method based on neurofeedback for a scientific research laboratory according to claim 6, characterized in that, The left view of the scene and the right view of the scene are both adjusted to the same size specification.

8. A brain vision detection method based on neurofeedback for scientific research laboratories according to claim 7, characterized in that, The true value of the optic nerve feedback signal and the predicted value of the optic nerve feedback signal are normalized before calculation.

Images