Automatic control method and system for animal laboratory platform based on image processing

Abstract

The embodiment of the present invention provides an automatic control method and system for an animal laboratory table based on image processing, the method comprising: when conducting an animal experiment, collecting first real-time image data of the animal laboratory table, the first real-time image data including an image of a target animal object; inputting the first real-time image data into a preset image recognition model to obtain first object recognition information, the first object recognition information including first positioning information and first posture information of the target animal object, the first positioning information being the position information of the target animal object on the animal laboratory table; determining the planned motion trajectory of the operating mechanical arm according to the first object recognition information, so as to control the operating mechanical arm to move according to the planned motion trajectory and perform experimental operations on the target animal object. The automation of the experimental operation of the animal laboratory table is realized, the experimental efficiency is improved, the experimental operation is completed by the operating mechanical arm, the experimental results are less dependent on the operator's skills, and the accuracy and repeatability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an automatic control method and system of an animal laboratory table based on image processing. Background Art

[0002] In the fields of biology, medical research, etc., in order to obtain new knowledge about biology, medicine, etc. or to solve specific problems, it is necessary to perform precise experimental operations on animals (for example, animal injection operations, animal surgery operations, etc.) for scientific research.

[0003] In the process of realizing the present invention, the inventors found that there are at least the following problems in the prior art: traditional animal experimental platforms mostly rely on manual operation, which is not only inefficient, but also the experimental results are greatly affected by the operator's skills, making it difficult to ensure the repeatability and accuracy of the experiment. Summary of the invention

[0004] In view of this, an object of an embodiment of the present invention is to provide an automatic control method and system for an animal laboratory table based on image processing.

[0005] To achieve the above-mentioned purpose, in a first aspect, an embodiment of the present invention provides an automatic control method of an animal laboratory table based on image processing, the method comprising:

[0006] When performing an animal experiment, collecting first real-time image data of the animal experimental table, wherein the first real-time image data includes an image of a target animal object;

[0007] Inputting the first real-time image data into a preset image recognition model to obtain first object recognition information, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, wherein the first positioning information is position information of the target animal object on the animal laboratory table;

[0008] Determining a planned motion trajectory of an operating robot arm according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform an experimental operation on the target animal object;

[0009] During the movement of the operating mechanical arm, second real-time image data of the animal experimental table is collected in real time, wherein the second real-time image data includes the motion image data of the operating mechanical arm and the current image data of the target animal object;

[0010] The real-time motion trajectory of the operating robot arm is adjusted according to the second real-time image data and the planned motion trajectory.

[0011] As an optional embodiment, adjusting the real-time motion trajectory of the operating robot arm according to the second real-time image data and the planned motion trajectory includes:

[0012] Determine a first real-time motion trajectory of the operating robot arm according to the second real-time image data;

[0013] Determining a deviation value between the first real-time motion trajectory and the planned motion trajectory;

[0014] The first real-time motion trajectory of the operating robot arm is adjusted according to the deviation value to obtain a second real-time motion trajectory, so as to control the operating robot arm to move according to the second real-time motion trajectory to perform experimental operations on the target animal object.

[0015] As an optional embodiment, determining the first real-time motion trajectory of the operating robot arm according to the second real-time image data includes:

[0016] Inputting the second real-time image data into the preset image recognition model to obtain second object recognition information, wherein the second object recognition information includes motion information of the operating robot arm, second positioning information and second posture information of the target animal object, wherein the motion information of the operating robot arm includes trajectory control points and motion speed of the operating robot arm;

[0017] A first real-time motion trajectory of the operating robot arm is determined according to the trajectory control points and the motion speed.

[0018] As an optional embodiment, adjusting the first real-time motion trajectory of the operating robot arm according to the deviation value to obtain the second real-time motion trajectory includes:

[0019] The trajectory control points and the motion speed of the first real-time motion trajectory are adjusted according to the deviation value to obtain a second real-time motion trajectory.

[0020] As an optional embodiment, the step of inputting the first real-time image data into a preset image recognition model to obtain first object recognition information of the target animal object includes:

[0021] Extracting key information features of the first real-time image data through a previous layer of the dynamic adjustment layer of the preset image recognition model, wherein the key information features include visual feature information of the target animal object and environmental feature information of the environment where the target animal object is currently located;

[0022] Adjusting the weight parameters of a specific network layer of the preset image recognition model according to the key information features through the dynamic adjustment layer of the preset image recognition model;

[0023] The first object recognition information of the target animal object is output by using the fully connected layer of the preset image recognition model after adjusting the weight parameters of the specific network layer.

[0024] As an optional embodiment, the method further includes:

[0025] Determine a target convolutional neural network, wherein the target convolutional neural network includes at least an input layer, a convolutional layer, an activation layer, a pooling layer, a fully connected layer, and a dynamic adjustment layer;

[0026] Acquire a target training set, wherein the target training set includes images of animal objects of different animal species and different colors under different experimental environment conditions;

[0027] After the target training set is input into the target convolutional neural network, key features are extracted from the image of the target training set through the convolution layer and the pooling layer;

[0028] Training the target convolutional neural network according to the key features to obtain a first convolutional neural network;

[0029] Specific features are determined from the key features through the dynamic adjustment layer to optimize the network parameters of the first convolutional neural network through the specific features, thereby obtaining a second convolutional neural network to obtain the preset image recognition model.

[0030] As an optional embodiment, determining a specific feature from the key features through the dynamic adjustment layer, optimizing the network parameters of the first convolutional neural network through the specific feature to obtain a second convolutional neural network, and obtaining the preset image recognition model, includes:

[0031] Determine the illumination change feature, the color feature and the breed feature of the animal object from the key features through the attention mechanism of the dynamic adjustment layer;

[0032] Adjusting the weight of the target network layer of the first convolutional neural network according to the illumination change characteristics, the color characteristics and the breed characteristics of the animal object;

[0033] The network parameters of the first convolutional neural network are optimized according to the weights of the target network layer of the first convolutional neural network to obtain a second convolutional neural network, and the second convolutional neural network is used as a preset image recognition model.

[0034] As an optional embodiment, the method further includes:

[0035] Determining whether the operating robot arm completes the experimental operation on the target animal object;

[0036] According to the operation result of the experimental operation, rewarding the experimental operation strategy of the operating robot arm according to a preset reward function;

[0037] Based on the reward, the experimental operation strategy of the operating robot arm is updated through a preset reinforcement learning algorithm.

[0038] In a second aspect, an embodiment of the present invention provides an automatic control system for an animal laboratory table based on image processing, the system comprising an animal laboratory table, an image acquisition device, an image processing device, an operation control device and an operation mechanical arm, wherein:

[0039] The animal experimental table is used to place a target animal object so as to perform experimental operations on the target animal object through the operating robot arm;

[0040] The image acquisition device is installed on the animal experimental platform, and is used to acquire image data of the target animal object on the animal experimental platform; and is also used to acquire second real-time image data of the animal experimental platform in real time during the movement of the operating mechanical arm, and the second real-time image data includes the motion image data of the operating mechanical arm and the current image data of the target animal object;

[0041] an image processing device, connected to the image acquisition device and the operation control device respectively, provided with a preset image recognition model, for receiving image data from the image acquisition device, processing the image data by the preset image recognition model to obtain first object recognition information, and inputting the first object recognition information into the operation control device, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, wherein the first positioning information is position information of the target animal object on the animal experimental table; and further used for adjusting the real-time motion trajectory of the operation robot arm according to the second real-time image data and the planned motion trajectory;

[0042] The operation control device is connected to the image processing device and the operation mechanical arm respectively, and is used to receive the first object recognition information from the image processing device, and determine the planned motion trajectory of the operation mechanical arm according to the first object recognition information, so as to control the operation mechanical arm to move according to the planned motion trajectory, and perform the experimental operation on the target animal object;

[0043] The operating robot arm is connected to the operating control device and is used to move according to the planned motion trajectory of the operating control device to perform experimental operations on the target animal object.

[0044] In some possible embodiments, adjusting the real-time motion trajectory of the operating robotic arm according to the second real-time image data and the planned motion trajectory includes: determining a first real-time motion trajectory of the operating robotic arm according to the second real-time image data; determining a deviation value between the first real-time motion trajectory and the planned motion trajectory; adjusting the first real-time motion trajectory of the operating robotic arm according to the deviation value to obtain a second real-time motion trajectory, so as to control the operating robotic arm to move according to the second real-time motion trajectory to perform experimental operations on the target animal object.

[0045] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned automatic control method of an animal laboratory bench based on image processing.

[0046] In a fourth aspect, an embodiment of the invention provides an electronic device, comprising:

[0047] one or more processors;

[0048] A storage device for storing one or more programs;

[0049] When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned automatic control method of an animal laboratory bench based on image processing.

[0050] The above technical solution has the following beneficial effects: during the experiment, real-time experimental images are collected, image recognition is performed through image processing technology, and the operating robot arm is controlled to perform experimental operations based on the recognition results, thereby realizing the automation of experimental operations on the animal experimental table and improving the efficiency of animal experiments. The operating robot arm is controlled by image processing technology throughout the process, and the experimental operations are completed by the operating robot arm, which reduces the dependence of the experimental results on the operator's skills and improves the accuracy and repeatability of animal experiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 This is one of the flow charts of an automatic control method of an animal laboratory table based on image processing provided by an embodiment of the present invention;

[0053] Figure 2This is a second flow chart of an automatic control method for an animal laboratory table based on image processing provided by an embodiment of the present invention;

[0054] Figure 3 This is a third flow chart of an automatic control method for an animal laboratory table based on image processing provided by an embodiment of the present invention;

[0055] Figure 4 This is a fourth flow chart of an automatic control method of an animal laboratory table based on image processing provided by an embodiment of the present invention;

[0056] Figure 5 It is a structural block diagram of an automatic control system for an animal laboratory table based on image processing provided by an embodiment of the present invention;

[0057] Figure 6 is a schematic diagram of a structural block of a computer-readable storage medium provided by an embodiment of the present invention;

[0058] Figure 7 It is a structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] During experiments on animals, experimenters are usually required to complete the entire process, such as animal capture, injection or surgery. These manual experimental operations are not only inefficient, but may also produce large errors. Especially during injection or surgical operations, inconsistent experience and skills of experimenters may affect the experimental results.

[0061] In view of this, the present invention provides an automatic control method and system for an animal laboratory table based on image processing, so as to improve experimental efficiency, reduce experimental errors, and thus ensure the repeatability and accuracy of the experiment.

[0062] Figure 1 FIG. 1 is one of the flow charts of an automatic control method of an animal laboratory table based on image processing provided by an embodiment of the present invention. Figure 1 As shown, an automatic control method for an animal laboratory table based on image processing may include the following steps:

[0063] Step 110 : When performing an animal experiment, first real-time image data of the animal test bench is collected.

[0064] Those skilled in the art can understand that, during the animal experiment, the target animal object can be placed on the animal laboratory table, for example, the target animal object can be controlled in a fixed position by some animal control devices. In the embodiment of the present invention, for example, a high-resolution camera can be used to capture the image (i.e., the first real-time image data) on the animal laboratory table in real time, and the image at least includes the image of the target animal object (i.e., the animal to be subjected to the experimental operation), the image of the table surface of the animal laboratory table, etc.

[0065] The high-resolution camera used in the embodiment of the present invention is, for example, Sony IMX series camera of Sony Corporation of Japan, FLIR industrial camera of FLIR Systems of the United States, Basler ace series camera of Basler AG of Germany, etc., so as to clearly obtain images of the animal test bench in various experimental environments and images of the target animal object. The image of the target animal object at least includes visual sensory information such as its position, posture, color, etc. on the animal test bench, so as to be used for the subsequent identification and positioning of the target animal object.

[0066] Step 120: input the first real-time image data into a preset image recognition model to obtain first object recognition information.

[0067] In an embodiment of the present invention, the preset image recognition model is a convolutional neural network, which collects a large number of animal laboratory table images including animal objects under different environmental conditions (for example, different lighting environments, light and dark), and then annotates these images, such as annotating the type of animal objects (for example, mice, rabbits, etc.), color, posture data, and the coordinate position of the animal objects on the animal laboratory table, etc., and then uses these data to train and learn the convolutional neural network, continuously optimize its network parameters, and finally form a model that can identify the position and posture of the target animal objects.

[0068] As an optional embodiment, the preset image recognition model may include all network layers of the convolutional neural network plus a dynamic adjustment layer. Based on this, this step can be specifically implemented as follows: extract the key information features of the first real-time image data through the previous layer of the dynamic adjustment layer of the preset image recognition model, wherein the key information features include the visual feature information of the target animal object and the environmental feature information of the current environment of the target animal object, and then adjust the weight parameters of the specific network layer of the preset image recognition model according to the key information features through the dynamic adjustment layer of the preset image recognition model, and use the fully connected layer of the preset image recognition model after adjusting the weight parameters of the specific network layer to output the first object recognition information of the target animal object.

[0069] It should be further explained that the first object recognition information includes the first positioning information and the first posture information of the target animal object, and the first positioning information is the position information of the target animal object on the animal test bench. The visual feature information of the target animal object is, for example, the color, type, and posture represented by the image of the target animal object, and the environmental feature information of the current environment of the target animal object is, for example, the brightness and darkness of the current environment. The dynamic adjustment layer can be implemented by an auxiliary network (such as a parameter generation network, including but not limited to a conditional batch normalization network (Conditional Batch Normalization), a feature generation network (Feature-wise Linear Modulation, FiLM), an adaptive instance normalization (Adaptive Instance Normalization, AdaIN) and an attention gating network (Attention Gating Networks), etc.), which dynamically generates parameters for adjusting the weights of certain layers in the main network according to the input features, and the previous layers of the dynamic adjustment layer are network layers such as convolutional layers and pooling layers of the convolutional neural network for extracting key feature information.

[0070] The first positioning information can represent the position information of the target animal object on the animal laboratory table, specifically, for example, relative coordinate position information, that is, position information relative to the animal laboratory table. The first posture information can represent the current posture state of the target animal object, which is used to determine the target point for experimental operation. Based on the positioning information, a large range (the position of the target animal object on the animal laboratory table) is determined, and then a smaller range (the position or part of the target animal object to be subjected to the experimental operation) is determined based on the posture information, so as to more accurately determine the target position for the experimental operation and improve the accuracy of the experiment.

[0071] Step 130 , determining a planned motion trajectory of the operating robot arm according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform an experimental operation on the target animal object.

[0072] The planned motion trajectory in the embodiment of the present invention includes control points for the movement of the manipulator, such as a starting point, an end point, and some necessary intermediate points, and a path for the movement of the manipulator is determined by these control points, i.e., the planned motion trajectory. The starting point is, for example, the initial position of the manipulator before the movement, and the end point is, for example, the position of the target animal object on the animal test bench. Some intermediate points can be determined between the starting point and the end point, and then the planned motion trajectory is generated using, for example, a spline function (including but not limited to Bezier curves, B-splines, or Catmull-Rom splines).

[0073] Specifically, for example, the step of generating a planned motion trajectory of the operating robot arm using a spline function may include:

[0074] Step 1: Determine the control points, which may include the starting point (the initial position before the manipulator moves), the end point (the position of the target animal object on the experimental table), and some necessary intermediate points (based on the laboratory environment and experimental equipment layout, select path points or intermediate points that can avoid collisions);

[0075] Step 2: Select a spline function. Select a suitable spline function according to the number of control points and the complexity of the path. In the embodiment of the present invention, Bezier curve, B-spline or Catmull-Rom spline (Catmull-RomSpline) is preferred;

[0076] Step 3: Calculate the spline function parameters. Calculate the specific parameters of the spline curve (such as the weight of the control point, the node vector, etc.) through the control points and the selected spline function type.

[0077] Step 4: Generate a path. Use the spline function to generate a smooth curve path through the control points. This path will be used as the planned motion trajectory of the manipulator.

[0078] Specifically, taking the Bezier curve as an example, a spline function formula is:

[0079]

[0080] Among them, B(t) is a point on the curve, t is a parameter in the range [0, 1], indicating the position on the curve, and b i is the i-th control point, n is the total number of control points minus one, It is the number of combinations, which represents the number of ways to select i elements from n elements.

[0081] Among them, the middle point can be determined according to obstacles in the experimental environment, such as experimental equipment, other robot parts, etc., and some points that can avoid collision with these parts are selected as middle points, so that the generation of the planned motion trajectory is safe and efficient. For example, if only the starting point and the end point are used as control points to determine the motion trajectory of the operating robot, for example, they can be directly connected into a straight line, so as to efficiently complete the experimental operation; however, since there are some other robot parts in the experimental environment, if a collision occurs, it may cause damage to the equipment and affect the experimental operation. Therefore, for the location of these robot parts, their coordinate points are determined, and the connection between the starting point and the end point needs to avoid the coordinate points of these robot parts to form another short and safe motion trajectory.

[0082] exist Figure 1Based on the embodiment shown, the embodiment of the present invention also provides Figure 2 The embodiment shown is used to optimize and adjust the motion trajectory of the robot arm in real time during its motion process to ensure that it reaches the target position accurately.

[0083] Figure 2 FIG2 is a flow chart of an automatic control method of an animal laboratory table based on image processing provided by an embodiment of the present invention. Figure 2 As shown, an automatic control method for an animal laboratory table based on image processing may include the following steps:

[0084] Step 210: When performing an animal experiment, first real-time image data of the animal test bench is collected.

[0085] Step 220: input the first real-time image data into a preset image recognition model to obtain first object recognition information.

[0086] Step 230, determining a planned motion trajectory of the operating robot arm according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform an experimental operation on the target animal object.

[0087] Among them, steps 210 to 230 can refer to Figure 1 For the sake of brevity, steps 110 to 130 are not described in detail here.

[0088] Step 240 : While operating the robot arm, collect second real-time image data of the animal experimental table in real time.

[0089] The embodiment of the present invention uses a high-resolution camera to acquire images of the animal laboratory table in real time, i.e., second real-time image data. The image may include motion image data of the operating robotic arm, such as the current position image and status information of the operating robotic arm, and may also include current image data of the target animal object, such as its posture information.

[0090] Step 250: adjusting the real-time motion trajectory of the manipulator according to the second real-time image data and the planned motion trajectory.

[0091] In order to ensure that the operating robot arm can accurately reach the target position and avoid environmental factors interfering with its movement so that it deviates from the originally planned motion trajectory, after obtaining the second real-time image data, the real-time motion trajectory of the operating robot arm is adjusted according to the second real-time image data and the planned motion trajectory.

[0092] Specifically, first, the first real-time motion trajectory of the operating mechanical arm is determined according to the second real-time image data. For example, the second real-time image data can be input into a preset image recognition model to obtain the second object recognition information. The second object recognition information can include the motion information of the operating mechanical arm, the second positioning information and the second posture information of the target animal object, etc., wherein the motion information of the operating mechanical arm includes the trajectory control point and the motion speed of the operating mechanical arm, and then the first real-time motion trajectory of the operating mechanical arm is determined according to the trajectory control point and the motion speed. As an optional embodiment, the real-time motion trajectory of the operating mechanical arm can be generated again using a spline function (including but not limited to Bezier curve, B-spline or Catmull-Rom spline) according to the second object recognition information. Then, the deviation value between the first real-time motion trajectory and the planned motion trajectory is calculated. For example, the deviation value can be determined by calculating the cosine similarity between the first real-time motion trajectory and the planned motion trajectory, or by determining the relative deviation or absolute deviation of each control point on the first real-time motion trajectory and the planned motion trajectory.

[0093] For example, the specific algorithm for calculating the deviation may include:

[0094] Method 1: Cosine similarity.

[0095] Formula: Similarity = (A·B) / (||A||||B||);

[0096] Among them, A and B represent two vectors (for example, A represents the vector of the first real-time motion trajectory, and B represents the vector of the planned motion trajectory), ||A|| and ||B|| are the moduli of vectors A and B respectively, and the cosine similarity is the angular relationship between the two vectors A and B. The closer the value is to 1, the greater the similarity.

[0097] Method 2: Relative or absolute deviation of control points.

[0098] formula:

[0099] Among them, p i represents the i-th control point on the first real-time motion trajectory, q i represents the i-th control point on the planned motion trajectory, and n represents the total number of control points. This formula calculates the Euclidean distance between corresponding control points on two trajectories, which is used to quantify the actual deviation of the two trajectories in space.

[0100] It should be noted that the above method of determining the deviation value is only an example and does not constitute a limitation to the embodiments of the present invention. An algorithm for determining the deviation value between other vectors or points can also be used to determine the deviation value between the first real-time motion trajectory and the planned motion trajectory. The embodiments of the present invention will not be listed one by one here.

[0101] After determining the deviation value between the two track trajectories, the path only needs to be adjusted to achieve the purpose of motion correction of the operating robot arm. Specifically, the first real-time motion trajectory of the operating robot arm can be adjusted based on the deviation value to obtain a second real-time motion trajectory. For example, the trajectory control points and the motion speed of the first real-time motion trajectory are adjusted according to the deviation value to obtain the second real-time motion trajectory, thereby controlling the operating robot arm to move according to the second real-time motion trajectory to perform experimental operations on the target animal object.

[0102] The following is an example of the steps of using the deviation value to correct the path of the operating robot arm, for example, including:

[0103] Step 1: Calculate the deviation between the first real-time motion trajectory and the planned motion trajectory using the above methods (e.g., cosine similarity, absolute / relative deviation, etc.).

[0104] Step 2: Determine the adjustment amount according to the size and direction of the deviation value, and adjust the position of each control point of the first real-time motion trajectory, that is, determine how much the coordinate position of each control point needs to be adjusted.

[0105] Step 3: Use the spline function to regenerate the motion trajectory containing the adjusted control points, that is, the second real-time motion trajectory, so that the operating robot arm moves according to the second real-time motion trajectory.

[0106] Specifically, a specific example of deviation correction is as follows:

[0107] Assuming that the control points of the original planned trajectory are (10,10) and (20,20), and the control points of the first real-time trajectory are (12,11) and (22,21), the calculated deviation value is:

[0108]

[0109] Adjust the trajectory according to the deviation value, for example, adjust the position of the first control point down by 2 units and the second one down by 1 unit.

[0110] In summary, during the experiment, real-time experimental images are collected, image recognition is performed through image processing technology, and the operating robot arm is controlled to perform experimental operations based on the recognition results, so as to realize the automation of experimental operations of the animal experimental platform and improve the efficiency of animal experiments. The operating robot arm is controlled by image processing technology throughout the process, and the experimental operation is completed by the operating robot arm, so that the experimental results are less dependent on the operator's skills and the accuracy and repeatability of animal experiments are improved. In addition, by adopting path planning technology such as spline functions, it is ensured that the operating robot arm can reach the target position smoothly and accurately. In addition, the image captured in real time by the camera (second real-time image data) is used to correct the motion trajectory of the operating robot arm in real time, reduce errors in the operation process, achieve high-precision positioning, and improve the performance and reliability of automatic control of the animal experimental platform based on image processing.

[0111] The present invention also provides another embodiment of an automatic control method for an animal laboratory table based on image processing, and an embodiment of training a preset image recognition model. It is understandable that after completing the training of the aforementioned preset image recognition model, the following steps are performed. Figure 1 and Figure 2 The steps of the embodiment shown are not repeated here for the sake of brevity. Figure 1 and Figure 2 Relevant steps of the illustrated embodiment.

[0112] like Figure 3 As shown, an automatic control method for an animal laboratory table based on image processing may include the following steps:

[0113] Step 310, determine the target convolutional neural network.

[0114] Step 320, obtaining a target training set.

[0115] Step 330, after the target training set is input into the target convolutional neural network, key features are extracted from the images of the target training set through the convolution layer and the pooling layer.

[0116] Step 340: Train the target convolutional neural network according to the key features to obtain a first convolutional neural network.

[0117] Step 350, determining specific features from the key features through the dynamic adjustment layer, optimizing the network parameters of the first convolutional neural network through the specific features, obtaining a second convolutional neural network, and using the second convolutional neural network as a preset image recognition model.

[0118] In an embodiment of the present invention, in addition to the input layer, convolution layer, activation layer, pooling layer, fully connected layer and other network layers, the target convolutional neural network (CNN) also adds a dynamic adjustment layer. Among them, the dynamic adjustment layer plays a key role in the target convolutional neural network, which enables the network to adaptively adjust parameters according to the specific features of the input image to improve the recognition and positioning accuracy of animal images. So that the embodiment of the present invention can be applied to process images in a variable experimental environment, such as animal experiments under different lighting conditions, or animals of various colors and types that need to be recognized and operated. As an optional embodiment, the dynamic adjustment layer can be implemented by an auxiliary network (such as a parameter generation network), which dynamically generates parameters for adjusting the weights of certain layers in the main network according to the input features.

[0119] The parameter generation network of the embodiment of the present invention may be, for example:

[0120] Conditional Batch Normalization: This technique allows the model to adjust the parameters of its batch normalization layer based on additional conditional information (such as different tasks or environmental variables). This can make the network more flexible to different input conditions.

[0121] Feature-wise Linear Modulation (FiLM): The FiLM layer changes the behavior of the network by applying a linear transformation (scale and offset) to the intermediate features. This technique is very effective in visual question answering and other tasks that require the model to adapt to different input information.

[0122] Adaptive Instance Normalization (AdaIN): AdaIN technology adapts to different styles or conditions by adjusting the mean and variance of features.

[0123] Dynamic Convolution: Dynamic convolution enables the network to dynamically adjust its convolutional layer based on the input by generating a specific convolution kernel for each input.

[0124] With the dynamic adjustment layer, the target training set of the embodiment of the present invention can have more complex experimental environment image data, for example, it can include images of animal objects of different animal species and different colors under different experimental environment conditions. In some embodiments, for the images of animal objects captured by a high-resolution camera, these animal object images can provide visual information of the animal objects, including but not limited to position, posture, type and color, etc. In order to ensure the quality of the image, the images in the target training set can be further preprocessed, including but not limited to denoising, standardization and other preprocessing, to improve the processing efficiency and accuracy of the image.

[0125] Further, the training phase of the target convolutional neural network is entered: key features are extracted from the image of the target training set through the convolution layer and the pooling layer, and the key features include but are not limited to key visual information such as the shape, size, texture, and environmental visual information of the animal object, and the target convolutional neural network is trained according to these key features to obtain a first convolutional neural network, and then the specific features are determined from the key features through the dynamic adjustment layer to optimize the network parameters of the first convolutional neural network through the specific features to obtain a second convolutional neural network, and the second convolutional neural network is used as a preset image recognition model. Specifically, the attention mechanism of the dynamic adjustment layer can be used to determine the illumination change features, the color features and the breed features of the animal object from the key features, and the weight of the target network layer of the first convolutional neural network is adjusted according to the illumination change features, the color features and the breed features of the animal object, and the network parameters of the first convolutional neural network are optimized according to the weight of the target network layer of the first convolutional neural network to obtain a second convolutional neural network, and the second convolutional neural network is used as a preset image recognition model, and finally a preset image recognition model is obtained. In an embodiment of the present invention, the attention mechanism enables the target convolutional neural network to focus on certain parts of the image (i.e., the aforementioned specific features) and dynamically adjust the weights according to the features of these parts. In addition, during the network training phase, the dynamic adjustment layer can also be trained together through the back propagation algorithm (BP algorithm) to ensure that optimizing the network parameters of the first convolutional neural network can effectively improve the model performance of the preset image recognition model.

[0126] The following is a specific example:

[0127] Suppose in an animal experiment, the experimental conditions include observing white and black mice under different lighting conditions. In this case, the preset image recognition model needs to accurately identify and locate these mice despite their different colors and lighting conditions:

[0128] First, input the image: the image received by the preset image recognition model may be a white mouse under strong light or a black mouse under dim light.

[0129] Then, feature extraction is performed: the convolutional layer of the preset image recognition model is used to extract features from the image, such as the shape, size, and color of the mouse.

[0130] Then, the dynamic adjustment layer comes into play:

[0131] If the image is of a white mouse taken under bright lighting, the dynamic adjustment layer will increase the weights of features that are sensitive to brightness and color contrast to better distinguish between the background and the target animal;

[0132] If the image is of a black mouse taken in dim light, a dynamic adjustment layer would adjust the weights to emphasize shape and outline features, helping to identify animals that are harder to see in dim environments.

[0133] Finally, the recognition result is output: the network adjusted by the dynamic adjustment layer can more accurately identify and locate animals under different lighting and color conditions, output precise position and posture information, and provide a basis for subsequent experimental operations.

[0134] In this way, the dynamic adjustment layer automatically adjusts the weights according to the specific features of the image, so that the preset image recognition model can maintain efficient and accurate performance under changing experimental conditions, especially in complex environments that need to deal with different lighting conditions and multiple animal characteristics. The preset image recognition model based on the convolutional neural network can automatically adjust its processing flow according to different environmental conditions and object characteristics to achieve the best image recognition results, ensuring the accuracy and efficiency of experimental operations.

[0135] exist Figures 1 to 3 Based on the embodiment shown, the embodiment of the present invention also provides another embodiment, whereby each time the robot arm completes an operation, it adjusts its strategy according to the result of the operation (success or failure), in order to achieve a higher success rate in future operations. It should be noted that the relevant steps of this embodiment are after step 130 or step 250, and for the sake of brevity, they are not repeated here. Figures 1 to 3 The relevant steps of the embodiments in are described repeatedly.

[0136] Figure 4 FIG4 is a flow chart of an automatic control method of an animal laboratory table based on image processing provided by an embodiment of the present invention. Figure 4 As shown, an automatic control method for an animal laboratory table based on image processing may also include the following steps:

[0137] Step 410, determining whether the operating robot arm has completed the experimental operation on the target animal object.

[0138] Step 420 , based on the operation result of the experimental operation, reward the experimental operation strategy of operating the robotic arm according to a preset reward function.

[0139] Step 430 , updating the experimental operation strategy of the operating robot arm through a preset reinforcement learning algorithm based on the reward.

[0140] In an embodiment of the present invention, a reward function can be designed to guide the learning process, which takes into account factors such as the accuracy of the operation, the time required, and the stress response of the animal to calculate the reward value to reward the experimental operation strategy of the operating robot arm. Specifically, the operating robot arm completes an experimental operation. If the operation is successful (such as accurately completing the injection), a positive reward is given according to the reward function, and if the operation fails or is inefficient, a negative reward or a lower positive reward is given according to the reward function. In some embodiments, the reward function also considers the side effects of the operation, such as the stress response of the target animal object, to ensure that the experimental operation is as gentle as possible. Further, according to the rewards received and state changes, the experimental operation strategy of the operating robot arm is updated by a preset reinforcement learning algorithm (for example: Q-Learning, deep Q network (deep Q network, DQN), policy gradient method, etc.) to maximize the future rewards obtained. For the above process, it can be repeated continuously to gradually learn and optimize the experimental operation strategy of the operating robot arm. Through continuous learning and optimization, the accuracy and efficiency of the operation of the operating robot arm are continuously improved.

[0141] Through this intelligent learning mechanism based on reinforcement learning algorithm, it is possible to continuously learn from actual operations and results, automatically adjust and optimize experimental operation strategies, and ultimately achieve automated, efficient and precise experimental operations while minimizing adverse effects on animal subjects.

[0142] Correspondingly, the embodiment of the present invention further provides a system embodiment corresponding to the aforementioned method embodiment, Figure 5 Schematic diagram of the structure of an automatic control system for an animal laboratory table based on image processing provided by an embodiment of the present invention. Figure 5 As shown, an automatic control system 500 for an animal laboratory table based on image processing includes an animal laboratory table 510, an image acquisition device 520, an image processing device 530, an operation control device 540 and an operation mechanical arm 550, wherein:

[0143] An animal experimental table 510, used to place a target animal object so as to perform an experimental operation on the target animal object through the operating robot arm;

[0144] An image acquisition device 520, installed on the animal experimental platform, for acquiring image data of a target animal object on the animal experimental platform;

[0145] An image processing device 530 is connected to the image acquisition device and the operation control device, respectively, and is provided with a preset image recognition model, for receiving image data from the image acquisition device, processing the image data by the preset image recognition model to obtain first object recognition information, and inputting the first object recognition information into the operation control device, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, and the first positioning information is the position information of the target animal object on the animal laboratory table;

[0146] an operation control device 540, connected to the image processing device and the operation manipulator, respectively, for receiving first object recognition information from the image processing device, and determining a planned motion trajectory of the operation manipulator according to the first object recognition information, so as to control the operation manipulator to move according to the planned motion trajectory, and perform an experimental operation on the target animal object;

[0147] An operating robot 550, connected to the operating control device, and configured to move according to a planned motion trajectory of the operating control device to perform an experimental operation on the target animal object;

[0148] In addition, during the movement of the operating mechanical arm 550, the image acquisition device 520 can also be used to acquire second real-time image data of the animal experimental table in real time, and the second real-time image data includes the motion image data of the operating mechanical arm and the current image data of the target animal object;

[0149] The image processing device 530 can also be used to adjust the real-time motion trajectory of the operating mechanical arm according to the second real-time image data and the planned motion trajectory, so that the operation control device 540 controls the operating mechanical arm 550 to perform experimental operations on the target animal object according to the adjusted real-time motion trajectory.

[0150] It should be noted that the image acquisition device 520, the image processing device 530 and the operation control device 540 can be integrated into a software system to complete related operations such as controlling the operation robot arm to move to a specific position, perform a specific operation (such as injection), etc.

[0151] In summary, during the experiment, real-time experimental images are collected, image recognition is performed through image processing technology, and the operating robot arm is controlled based on the recognition results to perform experimental operations, thereby realizing the automation of experimental operations on the animal experimental table and improving the efficiency of animal experiments. The entire process is controlled by image processing technology to operate the robot arm, and the experimental operations are completed by the operating robot arm, which reduces the dependence of the experimental results on the operator's skills and improves the accuracy and repeatability of animal experiments.

[0152] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0153] The embodiment of the present invention further provides a computer readable storage medium 600, such as Figure 6 As shown, a computer program is stored in the computer-readable storage medium 600, and when the computer program is executed by the processor, each step of the above-mentioned method for automatically controlling an animal laboratory table based on image processing is implemented. For example, when the computer program is executed by the processor, the following steps are implemented:

[0154] When performing an animal experiment, collecting first real-time image data of the animal experimental table, wherein the first real-time image data includes an image of a target animal object;

[0155] Inputting the first real-time image data into a preset image recognition model to obtain first object recognition information, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, wherein the first positioning information is position information of the target animal object on the animal laboratory table;

[0156] The planned motion trajectory of the operating robot arm is determined according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform experimental operation on the target animal object.

[0157] In some embodiments, the storage medium is further configured to store program code 610 for performing the following steps:

[0158] When performing an animal experiment, collecting first real-time image data of the animal experimental table, wherein the first real-time image data includes an image of a target animal object;

[0159] Inputting the first real-time image data into a preset image recognition model to obtain first object recognition information, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, wherein the first positioning information is position information of the target animal object on the animal laboratory table;

[0160] The planned motion trajectory of the operating robot arm is determined according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform experimental operation on the target animal object.

[0161] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. Of course, there are other ways of readable storage media, such as quantum memory, graphene memory, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0162] The embodiment of the present invention further provides an electronic device 700, such as Figure 7 As shown, it includes one or more processors 710 , a communication interface 720 , a memory 730 and a communication bus 740 , wherein the processor 710 , the communication interface 720 , and the memory 730 communicate with each other via the communication bus 740 .

[0163] Memory 730, for storing computer programs;

[0164] The processor 710 is used to implement the steps of the above-mentioned method for automatically controlling an animal laboratory table based on image processing when executing the program stored in the memory 730. For example, when the processor 710 executes the program stored in the memory 730, the following steps are implemented:

[0165] When performing an animal experiment, collecting first real-time image data of the animal experimental table, wherein the first real-time image data includes an image of a target animal object;

[0166] Inputting the first real-time image data into a preset image recognition model to obtain first object recognition information, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, wherein the first positioning information is position information of the target animal object on the animal laboratory table;

[0167] The planned motion trajectory of the operating robot arm is determined according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform experimental operation on the target animal object.

[0168] Processor 710 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0169] The memory 730 may include a large capacity memory for data or instructions. For example, but not limitation, the memory 730 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 730 may include a removable or non-removable (or fixed) medium. In a specific embodiment, the memory 730 is a non-volatile solid-state memory. In a specific embodiment, the memory 730 includes a read-only memory (ROM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM) or a flash memory or a combination of two or more of these.

[0170] Communication bus 740 includes hardware, software or both, for coupling the above-mentioned components to each other.For example, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations.In suitable cases, bus may include one or more buses.Although the embodiment of the present invention describes and shows a specific bus, the present invention considers any suitable bus or interconnection.

[0171] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0172] Although the present application provides method operation steps such as embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment).

[0173] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0174] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0175] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0176] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An automatic control method for an animal laboratory bench based on image processing, characterized in that: The method comprises: When performing an animal experiment, collecting first real-time image data of the animal experimental table, wherein the first real-time image data includes an image of a target animal object; Inputting the first real-time image data into a preset image recognition model to obtain first object recognition information, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, wherein the first positioning information is position information of the target animal object on the animal laboratory table; Determining a planned motion trajectory of an operating robot arm according to the first object recognition information, so as to control the operating robot arm to move according to the planned motion trajectory and perform an experimental operation on the target animal object; During the movement of the operating mechanical arm, second real-time image data of the animal experimental table is collected in real time, wherein the second real-time image data includes the motion image data of the operating mechanical arm and the current image data of the target animal object; adjusting the real-time motion trajectory of the operating robot arm according to the second real-time image data and the planned motion trajectory; The method further comprises: Determine a target convolutional neural network, wherein the target convolutional neural network includes at least an input layer, a convolutional layer, an activation layer, a pooling layer, a fully connected layer, and a dynamic adjustment layer; the dynamic adjustment layer is implemented by a parameter generation network; Acquire a target training set, wherein the target training set includes images of animal objects of different animal species and different colors under different experimental environment conditions; After the target training set is input into the target convolutional neural network, key features are extracted from the image of the target training set through the convolution layer and the pooling layer; Training the target convolutional neural network according to the key features to obtain a first convolutional neural network; Determining specific features from the key features through the dynamic adjustment layer, optimizing network parameters of the first convolutional neural network through the specific features, obtaining a second convolutional neural network, and obtaining the preset image recognition model; The method of determining specific features from the key features through the dynamic adjustment layer to optimize the network parameters of the first convolutional neural network through the specific features to obtain a second convolutional neural network to obtain the preset image recognition model includes: Determine the illumination change feature, the color feature and the breed feature of the animal object from the key features through the attention mechanism of the dynamic adjustment layer; Adjusting the weight of the target network layer of the first convolutional neural network according to the illumination change characteristics, the color characteristics and the breed characteristics of the animal object; Optimizing the network parameters of the first convolutional neural network according to the weights of the target network layer of the first convolutional neural network to obtain a second convolutional neural network, and using the second convolutional neural network as a preset image recognition model; The method further comprises: Determining whether the operating robot arm completes the experimental operation on the target animal object; According to the operation result of the experimental operation, rewarding the experimental operation strategy of the operating robot arm according to a preset reward function; Based on the reward, the experimental operation strategy of the operating robot arm is updated through a preset reinforcement learning algorithm.

2. The method according to claim 1, characterized in that The step of adjusting the real-time motion trajectory of the operating robot arm according to the second real-time image data and the planned motion trajectory comprises: Determine a first real-time motion trajectory of the operating robot arm according to the second real-time image data; Determining a deviation value between the first real-time motion trajectory and the planned motion trajectory; The first real-time motion trajectory of the operating robot arm is adjusted according to the deviation value to obtain a second real-time motion trajectory, so as to control the operating robot arm to move according to the second real-time motion trajectory to perform experimental operations on the target animal object.

3. The method according to claim 2, characterized in that The determining a first real-time motion trajectory of the operating robot arm according to the second real-time image data comprises: Inputting the second real-time image data into the preset image recognition model to obtain second object recognition information, wherein the second object recognition information includes motion information of the operating robot arm, second positioning information and second posture information of the target animal object, wherein the motion information of the operating robot arm includes trajectory control points and motion speed of the operating robot arm; A first real-time motion trajectory of the operating robot arm is determined according to the trajectory control points and the motion speed.

4. The method according to claim 3, characterized in that The step of adjusting the first real-time motion trajectory of the operating robot arm according to the deviation value to obtain a second real-time motion trajectory includes: The trajectory control points and the motion speed of the first real-time motion trajectory are adjusted according to the deviation value to obtain a second real-time motion trajectory.

5. The method according to claim 1, characterized in that The step of inputting the first real-time image data into a preset image recognition model to obtain first object recognition information of the target animal object includes: Extracting key information features of the first real-time image data through a previous layer of the dynamic adjustment layer of the preset image recognition model, wherein the key information features include visual feature information of the target animal object and environmental feature information of the environment where the target animal object is currently located; Adjusting the weight parameters of a specific network layer of the preset image recognition model according to the key information features through the dynamic adjustment layer of the preset image recognition model; The first object recognition information of the target animal object is output by using the fully connected layer of the preset image recognition model after adjusting the weight parameters of the specific network layer.

6. An automatic control system for an animal laboratory bench based on image processing, characterized in that: The system includes an animal experimental table, an image acquisition device, an image processing device, an operation control device and an operation mechanical arm, wherein: The animal experimental table is used to place a target animal object so as to perform an experimental operation on the target animal object through the operating robot arm; The image acquisition device is installed on the animal experimental platform, and is used to acquire image data of the target animal object on the animal experimental platform; and is also used to acquire second real-time image data of the animal experimental platform in real time during the movement of the operating mechanical arm, and the second real-time image data includes the motion image data of the operating mechanical arm and the current image data of the target animal object; The image processing device is connected to the image acquisition device and the operation control device respectively, and is provided with a preset image recognition model, for receiving image data from the image acquisition device, processing the image data by the preset image recognition model to obtain first object recognition information, and inputting the first object recognition information into the operation control device, wherein the first object recognition information includes first positioning information and first posture information of the target animal object, and the first positioning information is the position information of the target animal object on the animal experimental table; and is also used to adjust the real-time motion trajectory of the operating robot arm according to the second real-time image data and the planned motion trajectory; The operation control device is connected to the image processing device and the operation mechanical arm respectively, and is used to receive the first object recognition information from the image processing device, and determine the planned motion trajectory of the operation mechanical arm according to the first object recognition information, so as to control the operation mechanical arm to move according to the planned motion trajectory, and perform the experimental operation on the target animal object; The operating robot arm is connected to the operating control device and is used to move according to the planned motion trajectory of the operating control device to perform experimental operations on the target animal object; The image processing device is further used for: Determine a target convolutional neural network, wherein the target convolutional neural network includes at least an input layer, a convolutional layer, an activation layer, a pooling layer, a fully connected layer, and a dynamic adjustment layer; the dynamic adjustment layer is implemented by a parameter generation network; Acquire a target training set, wherein the target training set includes images of animal objects of different animal species and different colors under different experimental environment conditions; After the target training set is input into the target convolutional neural network, key features are extracted from the image of the target training set through the convolution layer and the pooling layer; Training the target convolutional neural network according to the key features to obtain a first convolutional neural network; Determining specific features from the key features through the dynamic adjustment layer, optimizing network parameters of the first convolutional neural network through the specific features, obtaining a second convolutional neural network, and obtaining the preset image recognition model; The method of determining a specific feature from the key features through the dynamic adjustment layer, optimizing the network parameters of the first convolutional neural network through the specific feature to obtain a second convolutional neural network, and obtaining the preset image recognition model, includes: Determine the illumination change feature, the color feature and the breed feature of the animal object from the key features through the attention mechanism of the dynamic adjustment layer; Adjusting the weight of the target network layer of the first convolutional neural network according to the illumination change characteristics, the color characteristics and the breed characteristics of the animal object; Optimizing the network parameters of the first convolutional neural network according to the weights of the target network layer of the first convolutional neural network to obtain a second convolutional neural network, and using the second convolutional neural network as a preset image recognition model; The image processing device is further used for: Determining whether the operating robot arm completes the experimental operation on the target animal object; According to the operation result of the experimental operation, rewarding the experimental operation strategy of the operating robot arm according to a preset reward function; Based on the reward, the experimental operation strategy of the operating robot arm is updated through a preset reinforcement learning algorithm.

7. The automatic control system for animal laboratory bench based on image processing according to claim 6, characterized in that: The method of adjusting the real-time motion trajectory of the operating robot arm according to the second real-time image data and the planned motion trajectory includes: determining a first real-time motion trajectory of the operating robot arm according to the second real-time image data; determining a deviation value between the first real-time motion trajectory and the planned motion trajectory; adjusting the first real-time motion trajectory of the operating robot arm according to the deviation value to obtain a second real-time motion trajectory, so as to control the operating robot arm to move according to the second real-time motion trajectory to perform experimental operations on the target animal object.

Images