CN110517287A - Method, device, equipment and storage medium for obtaining motion trajectory of robotic fish - Google Patents

Abstract

The invention discloses a method, a device, a device and a storage medium for obtaining the motion track of a robotic fish, which are used to solve the problem in the related art that the position of the robotic fish is lost when tracking the motion track of the underwater robotic fish. The method for acquiring the motion trajectory of the robotic fish includes: acquiring an image of the robotic fish swimming in water; segmenting the image to obtain a plurality of sub-images; identifying the robotic fish in the plurality of sub-images to determine the robotic fish The position of the fish in the image; when the robotic fish cannot be identified in the plurality of sub-images, after controlling the robotic fish to perform a preset action, acquire the image of the robotic fish again and determine the The position of the robotic fish in the re-acquired images; the trajectory of the robotic fish is determined according to the positions of the robotic fish in the images acquired at least twice. The invention can effectively track the position of the underwater robot fish.

Description

Method, device, equipment and storage medium for obtaining motion trajectory of robotic fish

technical field

The present invention relates to the technical field of trajectory tracking, in particular to a method, device, equipment and storage medium for obtaining the trajectory of a robotic fish.

Background technique

In recent years, with the continuous advancement of bionics technology, the research on underwater robots (robotic fish) that imitates fish underwater propulsion technology has attracted increasing attention and has become one of the hot spots in the field of underwater robots. Robotic fish can not only play a role in underwater operations, ocean monitoring, and reconnaissance in complex environments, but also provide a new way of thinking for the development of new underwater vehicles.

At present, domestic and foreign research mainly focuses on individual robotic fish, but in practical applications, due to the complexity, uncertainty, and concurrency of tasks, it is necessary to use multiple robotic fish to cooperate to complete the task. Since the robot fish itself has no positioning and telemetry capabilities, the visual system is its only "organ" to perceive the environment. Extract effective information as the basis for decision-making and control. Only by quickly and accurately tracking the position and direction of movement of the robot fish and the moving target, can the decision-making control module make corresponding decisions quickly to ensure the completion of multi-robot fish collaborative tasks. One of the key technologies to complete the multi-robot fish collaboration task is multi-robot fish real-time tracking, that is, to find multiple robot fish in the video image, and display the respective position sequences after one-to-one correspondence between the robot fish in different frames. However, when tracking the trajectory of the underwater robotic fish, the position of the robotic fish is often lost and the position of the robotic fish cannot be obtained.

Contents of the invention

In view of this, the object of the present invention is to propose a method, device, equipment and storage medium for acquiring the motion track of the robotic fish, which can effectively track the position of the underwater robotic fish.

According to the first aspect of the present invention, there is provided a method for obtaining the motion track of a robotic fish, comprising: obtaining an image of a robotic fish swimming in water; segmenting the image to obtain multiple sub-images; Identify the robotic fish in three sub-images to determine the position of the robotic fish in the image; when the robotic fish cannot be identified in the multiple sub-images, control the robotic fish to perform preset actions Afterwards, acquire the image of the robotic fish again and determine the position of the robotic fish in the image acquired again; determine the motion of the robotic fish according to the position of the robotic fish obtained at least twice in the image track.

Optionally, segmenting the image includes: using the color vector of each pixel in the image to be segmented as an input vector of an input neuron, and using the color vectors of eight pixels adjacent to each pixel as radial Each eigenvector of the function RBF; determine the seed neuron in the image to be segmented, wherein, when the maximum Manhattan distance from a pixel point to an adjacent pixel point is less than the first threshold, the pixel point is a seed pixel point, and the seed pixel point The corresponding neuron is a seed neuron, and the seed neuron in the image to be segmented constitutes a seed region; the seed region is grown by a preset growth rule to obtain a plurality of grouped regions; the average feature vector of each grouped region is calculated; The calculated average eigenvector replaces the eigenvectors contained in all neurons in the grouping area to which it belongs; if there is a neuron that is not connected to any grouping area, and the distance between the neuron and its adjacent grouping area is less than the second threshold , then connect the neuron to the grouping area closest to it; merge the adjacent grouping areas to obtain multiple areas to be segmented; segment the image to be segmented according to the multiple areas to obtain The plurality of sub-images.

Optionally, merging adjacent grouped regions includes: when the area of the region to be merged is smaller than a preset area, and the color distance of the region to be merged is smaller than a preset color distance threshold, combining the region to be merged merged into adjacent regions.

Optionally, determining the position of the robotic fish includes: calculating the target region in the image and the probability of pixel feature values in the candidate region based on a Meanshift target tracking algorithm to obtain a target model description and a candidate model description; using a similarity function to measure the The similarity of the target model and the candidate model of the current frame; select the candidate model that maximizes the similarity function and obtain the Meanshift vector of the target model; iteratively calculate the Meanshift vector, and obtain the position of the robotic fish through convergence.

Optionally, controlling the robotic fish to perform a preset action includes: controlling the robotic fish to turn according to a preset angle and a preset direction.

According to the second aspect of the present invention, there is provided a device for acquiring the motion trajectory of a robotic fish, comprising: a first acquiring module, configured to acquire an image of a robotic fish swimming in water; a segmentation module, configured to process the image Carry out segmentation to obtain a plurality of sub-images; an identification module is used to identify the robotic fish in the plurality of sub-images to determine the position of the robotic fish in the image; the second acquisition module, when in the plurality of sub-images When the robotic fish cannot be identified in any of the sub-images, after controlling the robotic fish to perform a preset action, acquire the image of the robotic fish again and determine the position of the robotic fish in the image obtained again; the determination module , for determining the motion trajectory of the robotic fish according to the position of the robotic fish in the image obtained at least twice.

Optionally, the segmentation module includes: a setting unit, configured to use the color vector of each pixel in the image to be segmented as an input vector of an input neuron, and set the color vectors of eight adjacent pixels to each pixel As each eigenvector of the radial function RBF; a determination unit, configured to determine the seed neuron in the image to be segmented, wherein, when the maximum Manhattan distance from a pixel point to an adjacent pixel point is less than a first threshold, the pixel point is A seed pixel point, the neuron corresponding to the seed pixel point is a seed neuron, and the seed neuron in the image to be segmented constitutes a seed region; a generation unit is used to grow the seed region by a preset growth rule to obtain multiple The grouping area; the first calculation unit is used to calculate the average eigenvector of each grouping area; the replacement unit is used to use the calculated average eigenvector to replace the eigenvectors contained in all neurons of the grouping area to which it belongs; the connection unit, If there is a neuron that is not connected to any grouping area, and the distance between the neuron and its adjacent grouping area is less than the second threshold, then the neuron is connected to the nearest grouping area; the merging unit is used Combining the adjacent grouped regions to obtain a plurality of regions to be segmented; a segmenting unit configured to segment the image to be segmented according to the plurality of regions to obtain the plurality of sub-images.

Optionally, the merging unit is configured to merge the region to be merged into an adjacent region when the area of the region to be merged is smaller than a preset area and the color distance of the region to be merged is smaller than a preset color distance threshold .

According to a third aspect of the present invention, there is provided an electronic device, including a memory, a processor, and a computer program stored on the memory and operable on the processor, when the processor executes the program, the computer program according to the present invention is realized. The first aspect provides any method for obtaining the trajectory of the robot fish.

According to a fourth aspect of the present invention, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the first aspect of the present invention. One aspect provides any method for obtaining the motion trajectory of the robotic fish.

As can be seen from the above, the method for obtaining the trajectory of the robotic fish provided by the present invention, when tracking the trajectory of the robotic fish swimming in the water, segments the obtained image of the robotic fish underwater After that, identifying the robot fish in the segmented image can improve the recognition rate of the robot fish, so as to avoid the problem of losing the position of the robot fish in the process of tracking the trajectory of the robot fish as much as possible.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a flow chart of a method for obtaining a robotic fish trajectory according to an exemplary embodiment;

Fig. 2 is a schematic diagram of a multi-target real-time positioning and tracking system in water under full vision shown according to an exemplary embodiment;

Fig. 3 is a top view of a robotic fish according to an exemplary embodiment;

Fig. 4 is a bottom view of a robotic fish according to an exemplary embodiment;

Fig. 5 is a flow chart showing image segmentation according to an exemplary embodiment;

Fig. 6 is a flow chart of a method for obtaining the trajectory of a robotic fish according to an exemplary embodiment;

Fig. 7 is a block diagram of a device for acquiring a motion track of a robotic fish according to an exemplary embodiment.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

It should be noted that all expressions using "first", "second", "third" and "fourth" in the embodiments of the present invention are to distinguish between two entities with the same name but different parameters or different parameters , it can be seen that "first", "second", "third" and "fourth" are only for the convenience of expression, and should not be construed as limiting the embodiments of the present invention, which will not be described one by one in the subsequent embodiments.

Fig. 1 is a flow chart of a method for obtaining a robotic fish trajectory according to an exemplary embodiment. As shown in Fig. 1, the method includes:

Step 101: acquiring an image of a robotic fish swimming in water;

In step 101, an image of a robotic fish swimming in water may be acquired through a visual sensor.

Step 102: Segment the image to obtain multiple sub-images;

Step 103: identifying the robotic fish in the plurality of sub-images to determine the position of the robotic fish in the image;

In step 103, the existing image recognition algorithm can be used to identify the robotic fish in the sub-images obtained by segmentation in step 102.

Step 104: When the robotic fish cannot be identified in the plurality of sub-images, after controlling the robotic fish to perform a preset action, acquire the image of the robotic fish again and determine that the robotic fish is captured again position in the image of ;

In step 104, if the robotic fish cannot be recognized in multiple sub-images, it means that the robotic fish has lost its position. The swimming posture makes the image presented on the water surface too small. In this case, the robotic fish can be controlled to perform specified actions so that the robotic fish can change its posture in the water. For example, the robotic fish can be controlled to move left or right Carry out the maximum degree of turning, the maximum degree of turning is the maximum turning angle that the robotic fish can use due to its physical structure. After the robot fish performs preset actions, the image is acquired again and the robot fish is identified in the image, which can increase the probability of the robot fish being successfully identified.

It should be noted that in step 104, after the robot fish is controlled to perform preset actions, the image of the robot fish swimming in the water can be acquired again, and the image obtained again can be segmented to obtain multiple sub-images. The robotic fish is identified in multiple sub-images.

Step 105: Obtain the motion trajectory of the robotic fish according to the position of the robotic fish obtained at least twice in the image.

As can be seen from the above, the method for obtaining the trajectory of the robotic fish provided by the present invention, when tracking the trajectory of the robotic fish swimming in the water, segments the obtained image of the robotic fish underwater After that, identifying the robot fish in the segmented image can improve the recognition rate of the robot fish, so as to avoid the problem of losing the position of the robot fish in the process of tracking the trajectory of the robot fish as much as possible.

Fig. 2 is a schematic diagram of a multi-target real-time positioning and tracking system in water under full vision according to an exemplary embodiment, and the above-mentioned method can be applied to the system. As shown in Figure 2, the system includes: visual sensor (1), aluminum support frame and tank guardrail (2), tank (3) and robotic fish (4). Fig. 3 is the top view of robotic fish, and Fig. 4 is the bottom view of robotic fish, shown in conjunction with Fig. 3 and Fig. 4, robotic fish (4) comprises: pectoral fin (5), waterproof fish skin (6), antenna (7), Charging plug (8), aluminum frame (9), communication module (10), fish fin (11), battery (12), control system (13), first joint (14), second joint (15) and Third joint (16).

Among them, the end of the visual sensor (1) is directly connected to the top of the system frame so that it can capture all the images of the pool, and the water tank guardrail (2) is fixed around the bottom of the system frame, which can play a role in the surrounding of the water tank (3). protection and support;

The robotic fish (4) can swim anywhere inside the tank (3), and its posture information at any moment is captured by the visual sensor (1);

The communication module (10), the battery (12) are positioned at the front end of the aluminum frame (9) and rigidly connected, for balancing with the first joint (14), the second joint (15) and the third joint (16);

The waterproof fish skin (6) can completely wrap the antenna (7), the aluminum frame (9), the communication module (10), the battery (12), the control system (13), the first joint (14), the second joint (15 ) and the third joint (16), so that these instruments are isolated from the water.

Wherein, the first joint (14), the second joint (15) and the third joint (16) can be multi-level servos.

In a practicable manner, segmenting the image may include: taking the color vector of each pixel in the image to be segmented as an input vector of an input neuron, and using the color vectors of eight pixels adjacent to each pixel Vector as each feature vector of radial function RBF; Determine the seed neuron in the image to be segmented, wherein, when the maximum Manhattan distance from a pixel to an adjacent pixel is less than the first threshold, the pixel is a seed pixel, The neuron corresponding to the seed pixel is a seed neuron, and the seed neuron in the image to be segmented constitutes a seed region; the seed region is grown by a preset growth rule to obtain a plurality of grouped regions; the average value of each grouped region is calculated Eigenvector; use the calculated average eigenvector to replace the eigenvectors contained in all neurons in the grouping area to which it belongs; if there is a neuron that is not connected to any grouping area, and the distance between the neuron and its adjacent grouping area If it is less than the second threshold, the neuron is connected to the nearest grouping area; the adjacent grouping areas are merged to obtain multiple areas to be segmented; the image to be segmented is processed according to the multiple areas performing segmentation to obtain the multiple sub-images.

In the above image segmentation process, for any pixel s that satisfies the formula (1), it can be used as a seed pixel.

μ _s < θ _μ

Among them, s is the pixel label, j is the subscript of adjacent pixels, ||x ^s -c ^j || is the Manhattan distance, the expression is shown in formula (2), θ _μ is the preset threshold (that is, the first Threshold), μ _s is the maximum distance between pixel s and its 8 adjacent pixels.

In the above image segmentation process, a pixel point will be connected to its 8 adjacent pixel points. In the color space, according to the above formula (1), the maximum Manhattan distance between a pixel point and an adjacent pixel point is less than the preset threshold (ie When the above-mentioned first threshold value), the pixel point is the seed pixel point. It can be seen that the seed pixel and its neighbor pixels have similar characteristics. In the M-PCNN image segmentation algorithm, the seed pixel and its neighbor pixels are connected to form a seed region, and the seed region captures the same feature pixels and continues to expand. It should be noted that the selection of seed pixels and the expansion of seed regions are carried out in parallel. This process is the initial stage of region merging in image segmentation, and repeated segmentation is used to avoid losing important details of the image. In the above formula (1), the threshold value of θ _μ is set to a small value. According to the color quantization theory, each RGB can be divided into 26 quantization levels to distinguish 17,576 colors. This color resolution is high enough for the human visual perception of 17,000 colors. According to these data, θ _μ is equal to the radius of the quantization interval in RGB space, and its normalized value is set to 1/52. The image segmentation process will be further described below in conjunction with FIG. 5 . As shown in Figure 5, the image segmentation process may include:

Step 501: After inputting the image to be segmented, use the color vector x of each pixel of the image as the input vector μ of an input neuron, and the color vectors of 8 adjacent pixels adjacent to the current pixel as the radial basis function RBF For each eigenvector c of each eigenvector c, use the seed selection condition to determine the initial seed point; use the growth rule to grow the seed area, use the selected seed point as the starting point, use formula (1) as the judgment condition, and use the formula (1) as the judgment condition to determine the initial seed point. 8 pixels are grown until the pixels no longer satisfy the formula (1), and the grouping number of the area is obtained;

Step 502: Calculate the average feature vector σ ^g of each grouping area, such as formula (6), σR ^g , σB ^g , are the average values of the red, green, and blue components respectively, and M is the number of pixels in the grouping area numbered g . Replace the resulting average eigenvector with the eigenvectors contained in all neurons in the region;

Step 503: Determine whether there are unconnected neurons, if there are unconnected neurons, and the difference between the unconnected neuron and the adjacent area (that is, the distance between the two) is less than the threshold _θi , execute step 504 The θ _i is a given threshold: use formula (7) to connect it to the nearest adjacent area, otherwise, perform step 505;

Among them, x ^μ is the feature vector of connecting pixel points μ, σ ^gj is the average feature vector of the adjacent area numbered gj. Perform the connection operation on all unconnected neurons at the same time, and update the threshold θ _i+1 = θ _i +Δθ _i (θ _i+1 is the new threshold, Δθ _i is the threshold increment), and repeat the above step 502;

Step 505: Calculate the area R ^s and the color distance R ^d ;

Step 506: Determine whether the area R ^s of the region and the color distance R ^d satisfy R ^s <θ ^s and R ^d <θ ^d , if so, execute step 507, otherwise, the process ends;

Step 507: Merge the obtained adjacent regions, and merge all spatial regions in parallel;

Merging rule: When the area R ^s and color distance R ^d of a region satisfy R ^s <θ ^s and R ^d <θ ^d , θ ^s and θ ^d are the preset area size threshold and color distance threshold respectively, and the region is merged into adjacent areas. If the color distance of multiple neighborhoods is less than the set threshold, each step has and only any one area is merged;

Step 507 is repeated until the region merging stop condition is satisfied, and the color image segmentation is completed.

In a practicable manner, merging adjacent grouped regions may include: when the area of the region to be merged is smaller than a preset area, and the color distance of the region to be merged is smaller than a preset color distance threshold, combining the The regions to be merged are merged into adjacent regions.

In a practicable manner, determining the position of the robotic fish may include: calculating the target region in the image and the pixel feature value probability in the candidate region based on the Meanshift target tracking algorithm, and obtaining the target model description and the candidate model description; Utilize similarity function to measure the similarity of the candidate model of described target model and current frame; Select the candidate model that makes described similarity function maximum and obtain the Meanshift (mean shift algorithm) vector of target model; Iterative calculation Meanshift vector, obtains by convergence Describe the position of the robotic fish.

In an implementable manner, controlling the robotic fish to perform a preset action may include: controlling the robotic fish to turn according to a preset angle and a preset direction. Wherein, the preset angle is, for example, the maximum steering angle that the robotic fish can use due to its physical structure. The preset direction can be left or right of the current moving direction of the robotic fish.

Fig. 6 is a flow chart of a method for obtaining a motion trajectory of a robotic fish according to an exemplary embodiment. The method for obtaining a motion trajectory of a robotic fish according to the present invention will be exemplarily described below in conjunction with Fig. 6 . As shown in Figure 6, the method includes:

Step 601: Obtain an image (also called an original image) collected by a visual sensor;

Step 602: Segment the image obtained in step 601 to obtain multiple sub-images;

Step 603: Binarize the multiple sub-images obtained in step 602;

Step 604: Carry out image segmentation processing on each sub-image processed by binarization;

Step 605: Recognize the robotic fish in the processed image, and obtain the recognition result;

Step 606: Execute the control program to control the movement of the robotic fish;

Step 607: Determine whether the position of the robotic fish is lost;

Step 608: If the position of the robotic fish is lost, determine that the robotic fish stops swimming;

Step 609: Control the robotic fish to turn left to the maximum angle;

Step 610: re-acquire the image collected by the vision sensor, and return to step 605.

Fig. 7 is a block diagram of a device for obtaining the motion track of a robotic fish according to an exemplary embodiment. As shown in Fig. 7, the device 70 includes:

The first acquiring module 71 is used to acquire the image of the robotic fish swimming in the water;

A segmentation module 72, configured to segment the image to obtain a plurality of sub-images;

An identification module 73, configured to identify the robotic fish in the plurality of sub-images, to determine the position of the robotic fish in the image;

The second acquisition module 74, when the robotic fish cannot be identified in the plurality of sub-images, after controlling the robotic fish to perform a preset action, acquire the image of the robotic fish again and determine that the robotic fish is in the The position in the image obtained again;

Determining module 75, is used for determining the motion track of described robotic fish according to the position of described robotic fish in described image that obtains at least twice.

In a practicable manner, the segmentation module includes: a setting unit, configured to use the color vector of each pixel in the image to be segmented as an input vector of an input neuron, and use the eight neurons adjacent to each pixel The color vector of the pixel is used as each eigenvector of the radial function RBF; the determination unit is used to determine the seed neuron in the image to be segmented, wherein, when the maximum Manhattan distance from a pixel point to an adjacent pixel point is less than the first threshold, The pixel point is a seed pixel point, and the neuron corresponding to the seed pixel point is a seed neuron, and the seed neuron in the image to be segmented constitutes a seed region; a generation unit is used to grow the seed region through a preset growth rule , to obtain multiple grouping areas; the first calculation unit is used to calculate the average eigenvector of each grouping area; the replacement unit is used to use the calculated average eigenvector to replace the eigenvectors contained in all neurons of the grouping area to which it belongs ; A connection unit, configured to connect the neuron to the nearest grouping area if there is a neuron not connected to any grouping area, and the distance from the neuron to its adjacent grouping area is less than a second threshold; The merging unit is configured to merge adjacent grouped regions to obtain multiple regions to be divided; the segmentation unit is configured to segment the image to be divided according to the multiple regions to obtain the plurality of sub-images.

In an implementable manner, the merging unit may be configured to: merge the regions to be merged when the area of the region to be merged is smaller than a preset area and the color distance of the region to be merged is smaller than a preset color distance threshold to the adjacent area.

Optionally, determining the position of the robotic fish may include: calculating the target region in the image and the probability of pixel feature values in the candidate region based on the Meanshift target tracking algorithm to obtain a target model description and a candidate model description; using a similarity function to measure The similarity of the target model and the candidate model of the current frame; select the candidate model that maximizes the similarity function and obtain the Meanshift vector of the target model; iteratively calculate the Meanshift vector, and obtain the position of the robotic fish through convergence.

Optionally, controlling the robotic fish to perform a preset action may include: controlling the robotic fish to turn according to a preset angle and a preset direction.

The devices in the above embodiments are used to implement the corresponding methods in the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

Based on the same inventive concept, an embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the program, the above-mentioned A method for obtaining the motion trajectory of a robotic fish described in an embodiment.

Based on the same inventive concept, an embodiment of the present invention also provides a non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the above-mentioned A method for obtaining the motion trajectory of a robotic fish described in an embodiment.

Those of ordinary skill in the art should understand that: the discussion of any of the above embodiments is exemplary only, and is not intended to imply that the scope of the present disclosure (including claims) is limited to these examples; under the idea of the present invention, the above embodiments or Combinations between technical features in different embodiments are also possible, steps may be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not presented in detail for the sake of brevity.

In addition, well-known power/ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided figures, for simplicity of illustration and discussion, and so as not to obscure the present invention. . Furthermore, devices may be shown in block diagram form in order to avoid obscuring the invention, and this also takes into account the fact that details regarding the implementation of these block diagram devices are highly dependent on the platform on which the invention is to be implemented (i.e. , these details should be well within the understanding of those skilled in the art). Where specific details (eg, circuits) have been set forth to describe example embodiments of the invention, it will be apparent to those skilled in the art that other embodiments may be implemented without or with variations from these specific details. Implement the present invention down. Accordingly, these descriptions should be regarded as illustrative rather than restrictive.

Although the invention has been described in conjunction with specific embodiments of the invention, many alternatives, modifications and variations of those embodiments will be apparent to those of ordinary skill in the art from the foregoing description. For example, other memory architectures such as dynamic RAM (DRAM) may use the discussed embodiments.

Embodiments of the present invention are intended to embrace all such alterations, modifications and variations that fall within the broad scope of the appended claims. Therefore, within the spirit and principle of the present invention, any omission, modification, equivalent replacement, improvement, etc. should be included in the protection scope of the present invention.

Claims (10)

1. A method for obtaining the motion track of a robotic fish, comprising: Get images of robotic fish swimming in water; Segmenting the image to obtain multiple sub-images; identifying the robotic fish in the plurality of sub-images to determine a position of the robotic fish in the images; When the robotic fish cannot be identified in the plurality of sub-images, after the robotic fish is controlled to perform a preset action, the image of the robotic fish is acquired again and it is determined that the robotic fish is in the acquired image again s position; The motion track of the robotic fish is determined according to the position of the robotic fish in the image acquired at least twice. 2. The method according to claim 1, wherein segmenting the image comprises: The color vector of each pixel in the image to be segmented is used as an input vector of an input neuron, and the color vectors of eight pixels adjacent to each pixel are used as each feature vector of the radial function RBF; Determine the seed neuron in the image to be segmented, wherein, when the maximum Manhattan distance from a pixel point to an adjacent pixel point is less than the first threshold, the pixel point is a seed pixel point, and the neuron corresponding to the seed pixel point is a seed neuron unit, the seed neuron in the image to be segmented constitutes a seed region; The seed area is grown by preset growth rules to obtain multiple grouping areas; Calculate the average eigenvector of each grouped area; Use the calculated average eigenvector to replace the eigenvectors contained in all neurons in the grouping area to which it belongs; If there is a neuron not connected to any grouping area, and the distance from the neuron to its adjacent grouping area is less than a second threshold, then the neuron is connected to the nearest grouping area; Merging the adjacent grouping regions to obtain multiple regions to be divided; The image to be segmented is segmented according to the multiple regions to obtain the multiple sub-images. 3. The method according to claim 2, wherein merging adjacent grouping regions comprises: When the area of the region to be merged is smaller than a preset area and the color distance of the region to be merged is smaller than a preset color distance threshold, the region to be merged is merged into an adjacent region. 4. The method according to claim 1, wherein determining the position of the robotic fish comprises: The target tracking algorithm based on Meanshift calculates the target region and the pixel feature value probability in the candidate region in the image, and obtains a target model description and a candidate model description; Using a similarity function to measure the similarity between the target model and the candidate model of the current frame; Select the candidate model that makes the similarity function maximum and obtain the Meanshift vector of the target model; Calculate the Meanshift vector iteratively, and obtain the position of the robotic fish through convergence. 5. The method according to any one of claims 1 to 4, wherein controlling the robotic fish to perform preset actions includes: The robotic fish is controlled to turn according to a preset angle and a preset direction. 6. A device for obtaining the trajectory of a robotic fish, comprising: The first acquisition module is used to acquire the image of the robotic fish swimming in water; A segmentation module, configured to segment the image to obtain multiple sub-images; An identification module, configured to identify the robotic fish in the plurality of sub-images, so as to determine the position of the robotic fish in the images; The second acquisition module, when the robotic fish cannot be identified in the plurality of sub-images, controls the robotic fish to perform a preset action, acquires the image of the robotic fish again and determines that the robotic fish is re-identified The position in the obtained image; A determining module, configured to determine the motion trajectory of the robotic fish according to the position of the robotic fish obtained at least twice in the image. 7. The device according to claim 6, wherein the segmentation module comprises: A setting unit is used to use the color vector of each pixel in the image to be segmented as an input vector of an input neuron, and use the color vectors of eight pixels adjacent to each pixel as each feature vector of the radial function RBF; A determining unit, configured to determine a seed neuron in the image to be segmented, wherein, when the maximum Manhattan distance from a pixel to an adjacent pixel is less than a first threshold, the pixel is a seed pixel, and the seed pixel corresponds to The neuron is a seed neuron, and the seed neuron in the image to be segmented constitutes a seed region; A generation unit is used to grow the seed region through preset growth rules to obtain multiple grouped regions; The first calculation unit is used to calculate the average feature vector of each grouping area; A replacement unit, configured to use the calculated average eigenvector to replace the eigenvectors contained in all neurons in the grouping area to which it belongs; A connection unit configured to connect the neuron to the nearest grouping area if there is a neuron not connected to any grouping area, and the distance from the neuron to its adjacent grouping area is less than a second threshold; a merging unit, configured to merge adjacent grouped regions to obtain multiple regions to be divided; A segmentation unit, configured to segment the image to be segmented according to the multiple regions to obtain the multiple sub-images. 8. The device according to claim 7, wherein the merging unit is used for: When the area of the region to be merged is smaller than a preset area and the color distance of the region to be merged is smaller than a preset color distance threshold, the region to be merged is merged into an adjacent region. 9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, wherein the processor implements any of claims 1 to 5 when executing the program. The method for obtaining the motion trajectory of the robotic fish described in the item. 10. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute any one of claims 1 to 5 The method for obtaining the motion trajectory of the robot fish.