Position Prediction Method, Device, Electronic Device and Storage Medium

Abstract

The present invention provides a position prediction method, apparatus, electronic device and storage medium. The method includes: determining the coordinates of three trajectory points at equal time intervals in a historical trajectory; based on the coordinates of the three trajectory points, activating a first spiking neuron in a position prediction model to obtain pulses corresponding to the three trajectory points output by the activated first spiking neuron; based on the pulses corresponding to the three trajectory points respectively and the connection strengths between the first spiking neuron and a second spiking neuron in the position prediction model, activating the second spiking neuron to obtain a position prediction result output by the activated second spiking neuron. This method uses a position prediction model constructed by a spiking neural network. The model activates spiking neurons according to the coordinate information of the input three trajectory points and outputs a position prediction result, reducing the computational complexity, shortening the delay from input to output, improving the effect of the prediction result, and further improving the effect of real-time response.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to a position prediction method, device, electronic device, and storage medium. Background Art

[0002] At present, the position prediction of moving targets has important applications in fields such as robot control, autonomous driving, and security systems. For example, in an autonomous driving scenario, a vehicle detection system can predict the positions of surrounding pedestrians and other motor vehicles through a position prediction algorithm, so as to perform operations such as braking and steering in advance, reducing the probability of traffic accidents.

[0003] The existing position prediction of moving targets mainly uses a sequence-to-sequence prediction network built by a long short-term memory network to predict the position of a moving target based on multiple position coordinates in the historical path. However, the sequence-to-sequence prediction network built by the long short-term memory network has a high computational complexity, resulting in a long delay from input to output, thus leading to low efficiency of the prediction result in practical applications and an unsatisfactory fast response effect of position prediction. Summary of the Invention

[0004] The present invention provides a position prediction method, device, electronic device, and storage medium to solve the defect of low efficiency of the high computational complexity of the prediction network in the prior art.

[0005] The present invention provides a position prediction method, including:

[0006] Determine the coordinates of three trajectory points at equal time intervals in the historical trajectory;

[0007] Based on the coordinates of the three trajectory points, activate the first spiking neuron in the position prediction model to obtain the pulses corresponding to the three trajectory points output by the activated neurons in the first spiking neuron;

[0008] Based on the pulses corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activate the second spiking neuron to obtain the position prediction result output by the activated spiking neuron in the second spiking neuron.

[0009] According to a position prediction method provided by the present invention, the first spiking neuron includes three groups of spiking neurons;

[0010] The activating the first spiking neuron in the position prediction model based on the coordinates of the three trajectory points to obtain the pulses corresponding to the three trajectory points output by the activated neurons in the first spiking neuron includes:

[0011] Based on the index of any one of the three groups of pulse neurons and the index of the three trajectory points, determine the trajectory point corresponding to the any one group, and apply it to the coordinates of the trajectory point corresponding to the any one group to activate the pulse neurons in the any one group, and obtain the pulses output by the activated neurons in the any one group.

[0012] According to a position prediction method provided by the present invention, the any one group includes a horizontal coordinate pulse neuron and a vertical coordinate pulse neuron;

[0013] The applying it to the coordinates of the trajectory point corresponding to the any one group to activate the pulse neurons in the any one group and obtaining the pulses output by the activated neurons in the any one group includes:

[0014] Based on the horizontal coordinate and the vertical coordinate of the trajectory point corresponding to the any one group, respectively activate the horizontal coordinate pulse neuron and the vertical coordinate pulse neuron in the any one group, and obtain the pulses output by the activated horizontal coordinate pulse neuron and the activated vertical coordinate pulse neuron in the any one group respectively.

[0015] According to a position prediction method provided by the present invention, the second pulse neuron includes a second horizontal coordinate pulse neuron and a second vertical coordinate pulse neuron;

[0016] The activating the second pulse neurons in the position prediction model based on the pulses respectively corresponding to the three trajectory points and obtaining the position prediction result output by the activated second pulse neurons includes:

[0017] Based on the pulse output by the activated horizontal coordinate pulse neuron in the any one group, combine the connection strength between the activated horizontal coordinate pulse neuron in the any one group and the second horizontal coordinate pulse neuron to obtain the membrane potential of the second horizontal coordinate pulse neuron corresponding to the any one group; and based on the pulse output by the activated vertical coordinate pulse neuron in the any one group, combine the connection strength between the activated vertical coordinate pulse neuron in the any one group and the second vertical coordinate pulse neuron to obtain the membrane potential of the second vertical coordinate pulse neuron corresponding to the any one group;

[0018] Based on the membrane potentials of the second horizontal coordinate pulse neurons and the membrane potentials of the second vertical coordinate pulse neurons corresponding to each group in the three groups of pulse neurons, determine the total membrane potential of the second horizontal coordinate pulse neurons and the total membrane potential of the second vertical coordinate pulse neurons;

[0019] Based on the total membrane potential of the second abscissa pulse neuron and the total membrane potential of the second ordinate pulse neuron, a preset threshold is applied to activate the second abscissa pulse neuron and the second ordinate pulse neuron respectively, and the abscissa of the position prediction result output by the activated pulse neurons in the second abscissa pulse neuron and the ordinate of the position prediction result output by the activated pulse neurons in the second ordinate pulse neuron are obtained.

[0020] According to a position prediction method provided by the present invention, the connection strength between the activated abscissa pulse neurons in any group and the second abscissa pulse neuron is determined based on the intensity control parameter corresponding to the any group, the index of the activated abscissa pulse neurons in the any group, the index of the second abscissa pulse neuron, and the total number of the second abscissa pulse neurons; the connection strength between the activated ordinate pulse neurons in any group and the second ordinate pulse neuron is determined based on the intensity control parameter corresponding to the any group, the index of the activated ordinate pulse neurons in the any group, the index of the second ordinate pulse neuron, and the total number of the second ordinate pulse neurons.

[0021] According to a position prediction method provided by the present invention, the intensity control parameters corresponding to each group in the three pulse neuron groups are calculated based on the following formula:

[0022]

[0023]

[0024]

[0025] In the formula, k represents a prediction time adjustment parameter; ρ1 represents the intensity control parameter corresponding to the first group in the three pulse neuron groups; ρ2 represents the intensity control parameter corresponding to the second group in the three pulse neuron groups; ρ3 represents the intensity control parameter corresponding to the third group in the three pulse neuron groups; the trajectory point corresponding to the first group is the trajectory point at the latest moment among the three trajectory points; the trajectory point corresponding to the second group is the trajectory point at the middle moment among the three trajectory points; the trajectory point corresponding to the third group is the trajectory point at the earliest moment among the three trajectory points.

[0026] The present invention also provides a position prediction device, including:

[0027] A determination unit, configured to determine the coordinates of three trajectory points at equal time intervals in a historical trajectory;

[0028] An input unit, configured to activate a first spiking neuron in a position prediction model based on the coordinates of the three trajectory points, and obtain first spikes corresponding to the three trajectory points output by the activated first spiking neuron;

[0029] A prediction unit, configured to activate a second spiking neuron in the position prediction model based on the first spikes corresponding to the three trajectory points, and obtain a position prediction result output by the activated second spiking neuron.

[0030] The present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the program, the position prediction method as described in any one of the above is implemented.

[0031] The present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the position prediction method as described in any one of the above is implemented.

[0032] The present invention further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the position prediction method as described in any one of the above is implemented.

[0033] The position prediction, device, electronic device, and storage medium provided by the present invention activate the first spiking neuron in the position prediction model through three trajectory points with equal time intervals in the historical trajectory, and based on the pulses output by the activated neurons in the first spiking neuron, as well as the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activate the second spiking neuron, and obtain the position prediction result output by the activated neurons in the second spiking neuron, realizing the construction of a position prediction model with a spiking neural network. This model performs position prediction by activating spiking neurons, reducing the computational complexity, shortening the time delay from input to output, improving the effect of the prediction result, and thus improving the effect of real-time response. Description of the Drawings

[0034] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0035] Figure 1 is one of the schematic flowcharts of the position prediction method provided by the present invention;

[0036] Figure 2 is the second schematic flowchart of the position prediction method provided by the present invention;

[0037] Figure 3 It is a network structure diagram of the location prediction model provided by the present invention;

[0038] Figure 4 It is a structural schematic diagram of a position prediction device provided by the present invention;

[0039] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0041] The current method for predicting the position of moving targets uses a sequence-to-sequence prediction network built with a long short-term memory network. Due to the high computational complexity of the long short-term memory network and the long delay from input to output, the rapid response effect of applying it to position prediction is not ideal. Therefore, how to reduce the computational complexity of the predicted position and improve the efficiency of position prediction is a technical problem that needs to be solved urgently in this field.

[0042] In order to solve this technical problem, the present invention provides a location prediction method. Figure 1 FIG. 1 is one of the flow charts of the location prediction method provided by the present invention. Figure 1 As shown, the method includes:

[0043] Taking into account that the pulse neurons in the pulse neural network have their own time attributes and are suitable for processing timing-related input information, and the computational complexity of the pulse neural network is low, therefore, the embodiment of the present invention constructs a position prediction model through a pulse neural network, which can reduce the computational complexity and improve the efficiency of position prediction.

[0044] Step 110, determining the coordinates of three trajectory points with equal time intervals in the historical trajectory;

[0045] It should be noted that the target detection algorithm is used to obtain a historical trajectory of a moving target. The method for obtaining the moving target position information can be video, radar, etc., which is not limited in the embodiment of the present invention. The form of the target detection algorithm can be a convolutional neural network, a traditional computer vision algorithm, etc., which is not limited in the embodiment of the present invention. Three trajectory points with equal time intervals mean that the three trajectory points A, B and C are determined in time order, and the time interval between A and B is equal to the time interval between B and C.

[0046] Step 120: Based on the coordinates of the three trajectory points, activate the first spiking neuron in the position prediction model to obtain the spikes corresponding to the three trajectory points output by the activated neurons in the first spiking neuron.

[0047] Step 130: Based on the spikes corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activate the second spiking neuron in the position prediction model to obtain the position prediction result output by the activated spiking neuron in the second spiking neuron.

[0048] Specifically, the position prediction model includes a first spiking neuron and a second spiking neuron. According to the coordinates of the three trajectory points, activate the spiking neurons in the first spiking neuron corresponding to the coordinates of the three trajectory points to obtain the spikes corresponding to the three trajectory points output by the activated spiking neurons in the first spiking neuron. Then, according to the spikes corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activate the spiking neurons in the second spiking neuron to obtain the position prediction result output by the activated spiking neuron in the second spiking neuron.

[0049] It should be noted that the coordinate system in which the coordinates of the three trajectory points are located is determined when generating the historical trajectory. The first pulse neuron and the second pulse neuron in the position prediction model determine the number of pulse neurons and the correspondence between the indices of each pulse neuron in the first pulse neuron and the second pulse neuron and the specified coordinate points in the coordinate system according to the maximum future time information for predicting the position of the moving target. This correspondence can be the correspondence between the overall coordinates of the specified coordinate points in the coordinate system and the indices of each pulse neuron in the first pulse neuron and the second pulse neuron, or it can be the correspondence between the abscissa and ordinate of the specified coordinate points in the coordinate system and the indices of each pulse neuron in the first pulse neuron and the second pulse neuron. The embodiments of the present invention do not limit this. After the correspondence between the indices of each pulse neuron in the first pulse neuron and the second pulse neuron and the specified coordinate points in the coordinate system, any coordinate point in the coordinate system can determine a specified coordinate point through its coordinates, thereby realizing the correspondence between the indices of each pulse neuron in the first pulse neuron and the second pulse neuron and any coordinate point in the coordinate system. For example, any coordinate point in the coordinate system determines the specified coordinate point closest to this coordinate point by calculating the distances from each specified coordinate point, thereby realizing the correspondence between the indices of each pulse neuron in the first pulse neuron and the second pulse neuron and any coordinate point in the coordinate system. Among them, the specified coordinate points in the coordinate system can be determined based on certain rules. The rule can be that the specified coordinate points are all coordinate points with integer abscissas and ordinates within a certain range in the coordinate system. For example: if the range of the coordinate system is that the abscissa is from 0 to 100 and the ordinate is from 0 to 100, then the coordinates of the specified coordinate points are (i, j), where 0 ≤ i ≤ 100 and 0 ≤ j ≤ 100.

[0050] According to the correspondence between the indices of each pulse neuron in the first pulse neuron and any coordinate point in the coordinate system, the pulse neurons in the first pulse neuron corresponding to the coordinates of the three trajectory points can activate the pulse neurons in the first pulse neuron according to the correspondence between the overall coordinates of the three trajectory points and the first pulse neuron indices, or can also activate the pulse neurons in the first pulse neuron according to the correspondence between the abscissa and ordinate of the three trajectory points and the first pulse neuron indices respectively. At this time, the activated pulse neurons in the first pulse neuron include the pulse neurons activated according to the abscissa and the pulse neurons activated according to the ordinate. The embodiments of the present invention do not limit this. Among them, the correspondence between the indices of the pulse neurons in the first pulse neuron and any coordinate point in the coordinate system is determined according to the above-mentioned correspondence between the indices of each pulse neuron in the first pulse neuron and the specified coordinate points in the coordinate system.

[0051] In addition, the first spiking neurons in the position prediction model can be a group or three groups of spiking neurons. When the first spiking neurons are a group of spiking neurons, the spiking neurons in this group are directly activated based on the coordinates of three trajectory points by this group of spiking neurons. When the first spiking neurons are three groups of spiking neurons, it is necessary to determine the trajectory points corresponding to each group among the three groups of spiking neurons according to the indexes of each group among the three groups of spiking neurons and the indexes of the three trajectory points, and activate the spiking neurons in each group according to the trajectory points corresponding to each group. The embodiments of the present invention do not limit this.

[0052] According to the correspondence between the indexes of the second spiking neurons and any coordinate point in the coordinate system, determine the coordinate point in the coordinate system corresponding to the index of the activated spiking neurons in the second spiking neurons. This coordinate point is the position prediction result. Among them, the activated spiking neurons in the second spiking neurons can be obtained by activating the second spiking neurons according to the spikes corresponding to the three trajectory points respectively, and the connection strength between each spiking neuron in the first spiking neurons and each spiking neuron in the second spiking neurons. Among them, the connection strength between each spiking neuron in the first spiking neurons and each spiking neuron in the second spiking neurons can be obtained through training, or can be a preset connection strength mapping relationship, or can be dynamically obtained according to the intensity control parameter. The embodiments of the present invention do not limit this. Among them, the intensity control parameter is calculated according to the future time information of the predicted moving target position. The correspondence between the indexes of the spiking neurons in the second spiking neurons and any coordinate point in the coordinate system is determined according to the correspondence between the indexes of each spiking neuron in the second spiking neurons and the specified coordinate point in the coordinate system described above.

[0053] The position prediction method provided by the embodiments of the present invention activates the first spiking neurons in the position prediction model through three trajectory points with equal time intervals in the historical trajectory, and activates the second spiking neurons according to the spikes output by the activated neurons in the first spiking neurons and the connection strength between the first spiking neurons and the second spiking neurons in the position prediction model, and obtains the position prediction result output by the activated neurons in the second spiking neurons, realizing the construction of a position prediction model with a spiking neural network. This model uses the coordinate information of three positions to perform position prediction in the way of activating spiking neurons, reducing the computational complexity, shortening the time delay from input to output, improving the effect of the prediction result, and further improving the effect of real-time response.

[0054] Based on the above embodiments, the present invention further provides an embodiment. The first spiking neurons in the above embodiments include three groups of spiking neuron groups, and step 120 includes:

[0055] Based on the index of any one of the three groups of spiking neurons and the index of three trajectory points, determine the trajectory point corresponding to any group, and apply it to the coordinates of the trajectory point corresponding to the group, activate the spiking neurons in the group, and obtain the spikes output by the activated neurons in the group.

[0056] Considering using the coordinates of three trajectory points in the historical trajectory for prediction, if only one group of spiking neurons is used for position prediction, it is necessary to process the three trajectory points sequentially in a serial manner. That is, the first trajectory point among the three trajectory points activates the first spiking neuron. After obtaining the spike of this trajectory point output by the activated neurons of the first spiking neuron, then perform the operation on the second trajectory point, and after the second trajectory point is completed, perform the operation on the third trajectory point. This serial method requires waiting for the spiking neurons activated by the previous trajectory point to reset before subsequent trajectory points can be activated, resulting in low efficiency. Therefore, in the embodiments of the present invention, the first spiking neuron is divided into three groups of spiking neurons, and the indexes of the spiking neurons in each group of spiking neurons respectively form a corresponding relationship with the specified coordinate points in the coordinate system. In this way, the three trajectory points can be processed in parallel, that is, the spiking neurons in the three groups of spiking neurons can be activated respectively according to the three trajectory points in parallel, improving the execution efficiency.

[0057] Specifically, determine the indexes of the three trajectory points according to the time sequence of the three trajectory points, and the indexes of the three groups of spiking neurons are determined during the construction of the position prediction model. After determining the three trajectory points, on the condition that the indexes are the same, any one of the three groups of spiking neurons activates the spiking neurons in the group according to the coordinates of the trajectory point with the same index as the group among the three trajectory points, and obtains the spikes output by the activated spiking neurons in the group. For example: for the three trajectory points A, B, and C, according to the time sequence of A, B, and C, determine the index of A as 0, the index of B as 1, and the index of C as 2. Then, the group with index 0 among the three groups of spiking neurons activates the spiking neurons in the group according to the coordinates of A, and outputs the spikes output by the activated spiking neurons in the group. The activation operations of the group with index 1 and the group with index 2 are the same as the activation operation of the group with index 0, which will not be elaborated here.

[0058] It should be noted that the number of spiking neurons in the three groups of spiking neurons is the same, and the indexes of the spiking neurons in each of the three groups of spiking neurons have a corresponding relationship with the specified coordinate points in the coordinate system, and the corresponding relationship between the indexes of the spiking neurons in each group and any coordinate point in the coordinate system is determined according to this corresponding relationship.

[0059] Based on the above embodiments, the present invention further provides an embodiment. Any one of the three groups of pulse neurons in the above embodiment includes an abscissa pulse neuron and an ordinate pulse neuron; and the coordinates applied to the trajectory point corresponding to this group in step 120 are used to activate the pulse neurons in this group, and the pulses output by the activated neurons in this group are obtained, including:

[0060] Based on the abscissa and ordinate of the trajectory point corresponding to this group, the abscissa pulse neuron and the ordinate pulse neuron in this group are respectively activated, and the pulses output by the activated abscissa pulse neuron and the activated ordinate pulse neuron in this group are obtained.

[0061] Considering that if the indexes of the pulse neurons in each of the three groups of pulse neurons form a corresponding relationship with the overall coordinates of the specified coordinate point in the coordinate system, the number of pulse neurons in each group will be too large. For example, if the abscissa coordinate of the coordinate system ranges from 0 to 100 and the ordinate coordinate ranges from 0 to 100, then 10,000 pulse neurons are required to form a corresponding relationship with the specified coordinate point, which will cause the number of pulse neurons in each group to increase exponentially as the range of the abscissa and ordinate coordinates of the coordinate system increases, thereby increasing the computational complexity and reducing the prediction position efficiency. Therefore, in the embodiments of the present invention, the pulse neurons in each of the three groups of pulse neurons are divided into abscissa pulse neurons and ordinate pulse neurons. The index of the abscissa pulse neuron corresponds to the abscissa of the specified coordinate point in the coordinate system, and the index of the ordinate pulse neuron corresponds to the ordinate of the specified coordinate point in the coordinate system. Taking the coordinate system in the above example as an example, each of the three groups of pulse neurons in the embodiments of the present invention only needs 200 pulse neurons to complete the corresponding relationship with the abscissa and ordinate in the coordinate system, so that the number of pulse neurons in each of the three groups of pulse neurons is only the sum of the maximum coordinate values of the abscissa and ordinate of the coordinate system, realizing the corresponding relationship between the abscissa and ordinate of the specified coordinate point in the coordinate system with a small number of pulse neurons, reducing the computational complexity, and improving the prediction efficiency.

[0062] Specifically, according to the abscissa of the trajectory point corresponding to any one of the three groups of pulse neurons, the abscissa pulse neuron in this group is activated, and the pulse output by the activated abscissa pulse neuron in this group is obtained. According to the ordinate of the trajectory point corresponding to this group, the ordinate pulse neuron in this group is activated, and the pulse output by the activated ordinate pulse neuron in this group is obtained.

[0063] Based on the above embodiments, Figure 2 is the second schematic flow chart of the position prediction method provided by the present invention. As Figure 2 shown, step 130 includes:

[0064] Step 131: Based on the pulses output by the activated abscissa pulse neurons in any group, combined with the connection strength between the activated abscissa pulse neurons and the second abscissa pulse neurons in this group, obtain the membrane potential of the second abscissa pulse neurons corresponding to this group; and based on the pulses output by the activated ordinate pulse neurons in this group, combined with the connection strength between the activated ordinate pulse neurons and the second ordinate pulse neurons in this group, obtain the membrane potential of the second ordinate pulse neurons corresponding to this group.

[0065] Step 132: Based on the membrane potentials of the second abscissa pulse neurons and the second ordinate pulse neurons corresponding to each group in the three groups of pulse neurons, determine the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons.

[0066] Step 133: Based on the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons, apply a preset threshold to activate the second abscissa pulse neurons and the second ordinate pulse neurons respectively, and obtain the abscissa of the position prediction result output by the activated pulse neurons in the second abscissa pulse neurons, and the ordinate of the position prediction result output by the activated pulse neurons in the second ordinate pulse neurons.

[0067] Considering that the second pulse neurons are divided into second abscissa pulse elements for predicting the abscissa and second ordinate pulse elements for predicting the ordinate, a small number of pulse neurons can be used to predict the position, improving the efficiency of predicting the position.

[0068] Specifically, the second pulse neurons include second abscissa pulse neurons and second ordinate pulse neurons. According to the product of the pulses output by the activated abscissa pulse neurons in any group among the three groups of pulse neurons and the connection strength between the activated abscissa pulse neurons and each pulse neuron in the second abscissa pulse neurons, and the product of the pulses output by the activated ordinate pulse neurons in this group and the connection strength between the activated ordinate pulse neurons and each pulse neuron in the second abscissa pulse neurons, obtain the membrane potential of the second abscissa pulse neurons and the membrane potential of the second ordinate pulse neurons corresponding to this group respectively.

[0069] Sum up the membrane potentials of the second abscissa pulse neurons and the membrane potentials of the second ordinate pulse neurons corresponding to each group in the three groups of pulse neurons obtained in Step 131 to obtain the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons.

[0070] Compare the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons obtained in step 132 with a preset threshold respectively. If there are total membrane potentials greater than the preset threshold in the second abscissa pulse neurons and the second ordinate pulse neurons, activate the pulse neuron with the maximum total membrane potential in the second abscissa pulse neurons and the pulse neuron with the maximum total membrane potential in the second ordinate pulse neurons respectively, and obtain the abscissa and ordinate of the position prediction results output by the activated pulse neurons in the second abscissa pulse neurons and the second ordinate pulse neurons respectively.

[0071] It should be noted that the abscissa pulse neuron index of any one of the three pulse neuron groups is the same as the second abscissa pulse neuron index, and the ordinate pulse neuron index of this group is the same as the second ordinate pulse neuron index. The connection strengths between the activated abscissa pulse neurons and each pulse neuron in the second abscissa pulse neurons, and the connection strengths between the activated ordinate pulse neurons and each pulse neuron in the second ordinate pulse neurons in each of the three pulse neuron groups can be obtained through training, or can be a preset connection strength mapping relationship, or can be dynamically obtained according to the strength control parameter. The embodiments of the present invention do not limit this. Among them, the strength control parameter is calculated according to the future moment information of the predicted moving target position.

[0072] In addition, the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons are both calculated using the following formula:

[0073] P j =∑ i S i *W ij

[0074] In the formula, when predicting the abscissa of the position prediction result, i is the index of the activated abscissa pulse neuron in this group, j is the index of the j-th abscissa pulse neuron in the second abscissa pulse neurons, S i is the pulse output by the activated abscissa pulse neuron in this group, and W ij is the connection strength between the i-th abscissa pulse neuron in this group and the j-th abscissa pulse neuron in the second abscissa pulse neurons; when predicting the ordinate of the position prediction result, i is the index of the activated ordinate pulse neuron in this group, j is the index of the second ordinate pulse neuron, S i is the pulse output by the activated ordinate pulse neuron in this group, and W ij is the connection strength between the i-th ordinate pulse neuron in this group and the j-th abscissa pulse neuron in the second abscissa pulse neurons.

[0075] Based on the above embodiments, the present invention provides an embodiment of a method for obtaining connection strength, and the method includes:

[0076] The connection strength between the activated abscissa pulse neuron and the second abscissa pulse neuron in any group is determined based on the strength control parameter corresponding to the group, the index of the activated abscissa pulse neuron in the group, the index of the second abscissa pulse neuron, and the total number of the second abscissa pulse neurons; the connection strength between the activated ordinate pulse neuron and the second ordinate pulse neuron in the group is determined based on the strength control parameter corresponding to the group, the index of the activated ordinate pulse neuron in the group, the index of the second ordinate pulse neuron, and the total number of the second ordinate pulse neurons.

[0077] Specifically, the connection strength between the activated abscissa pulse neuron and the second abscissa pulse neuron in any one of the three groups of pulse neurons, and the connection strength between the activated ordinate pulse neuron and the second ordinate pulse neuron in the group are calculated by the following formula:

[0078]

[0079] In the formula, ρ is the strength control parameter of the group. When predicting the abscissa in the prediction result of the prediction position, i is the index of the activated abscissa pulse neuron in the group, j is the index of the j-th abscissa pulse neuron in the second abscissa pulse neurons, and m is the total number of abscissa pulse neurons in the group, which is also the total number of the second abscissa pulse neurons; when predicting the ordinate in the prediction result of the prediction position, i is the index of the activated ordinate pulse neuron in the group, j is the index of the j-th ordinate pulse neuron in the second ordinate pulse neurons, and m is the total number of ordinate pulse neurons in the group, which is also the total number of the second ordinate pulse neurons. Among them, 0 ≤ i < m, 0 ≤ j < m.

[0080] It should be noted that the strength control parameters ρ of each group in the three groups of pulse neurons are different. The strength control parameters ρ of each group can be preset positions, or can be dynamically determined according to the prediction time to adjust the parameters. The prediction time to adjust the parameters represents predicting the position of the moving target at the moment after the parameter number of equal time intervals. For example: assuming the time interval is Δt and the strength adjustment coefficient is 1, it means predicting the position of the moving target at the moment after 1 time interval Δt, and the strength adjustment coefficient is 2, which means predicting the position of the moving target at the moment after 2 time intervals Δt. The embodiments of the present invention do not limit this.

[0081] Based on the above embodiments, the present invention provides an embodiment of a method for obtaining the strength control parameter corresponding to three groups of pulse neurons, and the method includes:

[0082] The corresponding intensity control parameters in each of the three groups of pulse neurons are calculated based on the following formula:

[0083]

[0084]

[0085]

[0086] In the formula, k represents the prediction time adjustment parameter; ρ1 represents the intensity control parameter corresponding to the first group in the three groups of pulse neurons; ρ2 represents the intensity control parameter corresponding to the second group in the three groups of pulse neurons; ρ3 represents the intensity control parameter corresponding to the third group in the three groups of pulse neurons; the trajectory point corresponding to the first group is the trajectory point at the latest moment among the three trajectory points; the trajectory point corresponding to the second group is the trajectory point at the middle moment among the three trajectory points; the trajectory point corresponding to the third group is the trajectory point at the earliest moment among the three trajectory points.

[0087] It should be noted that the position at the predicted future moment can be determined according to the preset time interval △t, the current moment t, and the prediction time adjustment parameter k. When k is equal to 1, the predicted future moment is t + △t. When k is equal to 2, the predicted future moment is t + 2△t. That is, the formula for predicting the future moment is t + k△t.

[0088] Based on the above embodiments, the present invention provides a preferred embodiment. Figure 3 It is the network structure diagram of the position prediction model provided by the present invention. As Figure 3 shown in the figure, △t in the figure represents the time interval, t represents the current moment, k represents the prediction time adjustment parameter. The position prediction model includes an input layer and a prediction layer. Each pulse neuron in the input layer is fully connected to the prediction layer. Among them, the input layer includes three groups of pulse neurons. The three groups of pulse neurons respectively process three trajectory points. The three trajectory points are trajectory points with equal time intervals in the historical trajectory. The three trajectory points are respectively (x1, y1, t), (x2, y2, t - Δt), (x3, y3, t - 2Δt). The time to be predicted is t + kΔt, and (x, y, t + kΔt) is the position to be predicted.

[0089] Specifically, the input layer receives the input of the coordinate information of three trajectory points on the historical trajectory of the moving target. The time interval between the three trajectory points is Δt, and the corresponding moments of the three positions are t, t - Δt, and t - 2Δt respectively. According to the input trajectory point coordinates, the integrate-and-fire pulse neurons in the group corresponding to the trajectory point in the three pulse neuron groups in the input layer are activated, and pulses are generated and transmitted to the prediction layer. There is a full connection between the input layer and the prediction layer. The second pulse neurons in the prediction layer receive three abscissa pulses and three ordinates from the input layer, and accumulate the membrane potential according to the connection weights corresponding to the activated pulse neurons in the input layer. Finally, the index corresponding to the second abscissa pulse neuron with the maximum membrane potential in the prediction layer is the abscissa of the position coordinate predicted at the current moment, and the index corresponding to the second ordinate pulse neuron with the maximum membrane potential in the prediction layer is the ordinate of the position coordinate predicted at the current moment.

[0090] The position prediction device provided by the present invention will be described below. The position prediction device described below can be correspondingly referred to the position prediction method described above.

[0091] Figure 4 It is a schematic structural diagram of the position prediction device provided by the present invention. As Figure 4 shown, the device includes: a determination unit 410, an input unit 420, and a prediction unit 430.

[0092] Among them,

[0093] The determination unit 410 is configured to determine the coordinates of three trajectory points with equal time intervals in the historical trajectory;

[0094] The input unit 420 is configured to activate the first pulse neurons in the position prediction model based on the coordinates of the three trajectory points, and obtain the first pulses corresponding to the three trajectory points output by the activated first pulse neurons;

[0095] The prediction unit 430 is configured to activate the second pulse neurons in the position prediction model based on the first pulses corresponding to the three trajectory points, and obtain the position prediction result output by the activated second pulse neurons.

[0096] In an embodiment of the present invention, a determination unit is configured to determine the coordinates of three trajectory points at equal time intervals in a historical trajectory; an input unit is configured to activate a first spiking neuron in a position prediction model based on the coordinates of the three trajectory points, and obtain first spikes corresponding to the three trajectory points output by the activated first spiking neuron; a prediction unit is configured to activate a second spiking neuron in the position prediction model based on the first spikes corresponding to the three trajectory points, and obtain a position prediction result output by the activated second spiking neuron, thereby implementing a position prediction model constructed by a spiking neural network. The model predicts positions by activating spiking neurons, reducing computational complexity, shortening the latency from input to output, improving the effect of the prediction result, and further improving the effect of real-time response.

[0097] Based on any of the above embodiments, the first spiking neuron in the input unit 420 includes three groups of spiking neurons. The input unit 420 is specifically configured to determine the trajectory point corresponding to a group based on the index of any one of the three groups of spiking neurons and the index of the three trajectory points, and apply the coordinates of the trajectory point corresponding to the group to activate the spiking neurons in the group, and obtain the spikes output by the activated neurons in the group.

[0098] Based on any of the above embodiments, any one of the three groups of spiking neurons in the input unit 420 includes a horizontal coordinate spiking neuron and a vertical coordinate spiking neuron. The input unit 420 includes:

[0099] An activation subunit is configured to respectively activate the horizontal coordinate spiking neuron and the vertical coordinate spiking neuron of the group based on the horizontal coordinate and the vertical coordinate of the trajectory point corresponding to the group, and obtain the spikes output by the activated horizontal coordinate spiking neuron and the activated vertical coordinate spiking neuron in the group.

[0100] Based on any of the above embodiments, the second spiking neuron in the prediction unit 430 includes a second horizontal coordinate spiking neuron and a second vertical coordinate spiking neuron. The prediction unit 430 includes:

[0101] A membrane potential calculation subunit is configured to obtain the membrane potential of the second horizontal coordinate spiking neuron corresponding to the group based on the spike output by the activated horizontal coordinate spiking neuron in the group and in combination with the connection strength between the activated horizontal coordinate spiking neuron and the second horizontal coordinate spiking neuron in the group; and obtain the membrane potential of the second vertical coordinate spiking neuron corresponding to the group based on the spike output by the activated vertical coordinate spiking neuron in the group and in combination with the connection strength between the activated vertical coordinate spiking neuron and the second vertical coordinate spiking neuron in the group.

[0102] The total membrane potential determination subunit is configured to determine the total membrane potential of the second abscissa pulse neuron and the total membrane potential of the second ordinate pulse neuron based on the membrane potential of the second abscissa pulse neuron corresponding to each group and the membrane potential of the second ordinate pulse neuron in three groups of pulse neurons;

[0103] The prediction subunit is configured to, based on the total membrane potential of the second abscissa pulse neuron and the total membrane potential of the second ordinate pulse neuron, apply a preset threshold to activate the second abscissa pulse neuron and the second ordinate pulse neuron respectively, and obtain the abscissa of the position prediction result output by the activated pulse neurons in the second abscissa pulse neuron and the ordinate of the position prediction result output by the activated pulse neurons in the second ordinate pulse neuron.

[0104] Based on any of the above embodiments, the membrane potential calculation subunit includes:

[0105] The connection strength calculation subunit: The connection strength between the activated abscissa pulse neuron and the second abscissa pulse neuron in any group is determined based on the strength control parameter corresponding to the group, the index of the activated abscissa pulse neuron in the group, the index of the second abscissa pulse neuron, and the total number of the second abscissa pulse neurons; the connection strength between the activated ordinate pulse neuron and the second ordinate pulse neuron in the group is determined based on the strength control parameter corresponding to the group, the index of the activated ordinate pulse neuron in the group, the index of the second ordinate pulse neuron, and the total number of the second ordinate pulse neurons.

[0106] Based on any of the above embodiments, the connection strength calculation subunit includes:

[0107] The strength control parameter calculation subunit is configured to calculate the strength control parameters corresponding to each group in the three groups of pulse neurons based on the following formula:

[0108]

[0109]

[0110]

[0111] In the formula, k represents the prediction time adjustment parameter; ρ1 represents the strength control parameter corresponding to the first group in the three groups of pulse neurons; ρ2 represents the strength control parameter corresponding to the second group in the three groups of pulse neurons; ρ3 represents the strength control parameter corresponding to the third group in the three groups of pulse neurons; the trajectory point corresponding to the first group is the trajectory point at the latest moment among the three trajectory points; the trajectory point corresponding to the second group is the trajectory point at the middle moment among the three trajectory points; the trajectory point corresponding to the third group is the trajectory point at the earliest moment among the three trajectory points.

[0112] Figure 5 Illustrates a schematic diagram of the physical structure of an electronic device, as Figure 5 shown. The electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540. Among them, the processor 510, the communications interface 520, and the memory 530 complete communication with each other through the communication bus 540. The processor 510 may call the logical instructions in the memory 530 to execute a position prediction method, which includes: determining the coordinates of three trajectory points at equal time intervals in the historical trajectory; based on the coordinates of the three trajectory points, activating the first spiking neuron in the position prediction model to obtain the pulses corresponding to the three trajectory points output by the activated neurons in the first spiking neuron; based on the pulses corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activating the second spiking neuron to obtain the position prediction result output by the activated spiking neuron in the second spiking neuron.

[0113] In addition, when the logical instructions in the above-mentioned memory 530 can be implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0114] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the position prediction method provided by each of the above methods. The method includes: determining the coordinates of three trajectory points at equal time intervals in the historical trajectory; based on the coordinates of the three trajectory points, activating the first spiking neuron in the position prediction model to obtain the spikes corresponding to the three trajectory points output by the activated neurons in the first spiking neuron; based on the spikes corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activating the second spiking neuron to obtain the position prediction result output by the activated spiking neuron in the second spiking neuron.

[0115] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the position prediction method provided by each of the above methods. The method includes: determining the coordinates of three trajectory points at equal time intervals in the historical trajectory; based on the coordinates of the three trajectory points, activating the first spiking neuron in the position prediction model to obtain the spikes corresponding to the three trajectory points output by the activated neurons in the first spiking neuron; based on the spikes corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activating the second spiking neuron to obtain the position prediction result output by the activated spiking neuron in the second spiking neuron.

[0116] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0117] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A position prediction method, characterized in that, Including: Determine the coordinates of three trajectory points at equal time intervals in the historical trajectory; Based on the coordinates of the three trajectory points, activate the first spiking neuron in the position prediction model, and obtain the spikes corresponding to the three trajectory points output by the activated neurons in the first spiking neuron; Based on the spikes corresponding to the three trajectory points respectively, and the connection strength between the first spiking neuron and the second spiking neuron in the position prediction model, activate the second spiking neuron, and obtain the position prediction result output by the activated spiking neuron in the second spiking neuron; Wherein, the first spiking neuron includes three groups of spiking neurons; any group includes a horizontal coordinate spiking neuron and a vertical coordinate spiking neuron; The second spiking neuron includes a second horizontal coordinate spiking neuron and a second vertical coordinate spiking neuron; The activating the second spiking neuron in the position prediction model based on the spikes corresponding to the three trajectory points respectively, and obtaining the position prediction result output by the activated second spiking neuron includes: Based on the spike output by the activated horizontal coordinate spiking neuron in any group, and combining the connection strength between the activated horizontal coordinate spiking neuron in any group and the second horizontal coordinate spiking neuron, obtain the membrane potential of the second horizontal coordinate spiking neuron corresponding to any group; and based on the spike output by the activated vertical coordinate spiking neuron in any group, and combining the connection strength between the activated vertical coordinate spiking neuron in any group and the second vertical coordinate spiking neuron, obtain the membrane potential of the second vertical coordinate spiking neuron corresponding to any group; Based on the membrane potentials of the second horizontal coordinate spiking neurons and the membrane potentials of the second vertical coordinate spiking neurons corresponding to each group in the three groups of spiking neurons, determine the total membrane potential of the second horizontal coordinate spiking neuron and the total membrane potential of the second vertical coordinate spiking neuron; Based on the total membrane potential of the second horizontal coordinate spiking neuron and the total membrane potential of the second vertical coordinate spiking neuron, apply a preset threshold to activate the second horizontal coordinate spiking neuron and the second vertical coordinate spiking neuron respectively, and obtain the abscissa of the position prediction result output by the activated spiking neuron in the second horizontal coordinate spiking neuron, and the ordinate of the position prediction result output by the activated spiking neuron in the second vertical coordinate spiking neuron.

2. The position prediction method according to claim 1, characterized in that, The activating the first spiking neuron in the position prediction model based on the coordinates of the three trajectory points, and obtaining the spikes corresponding to the three trajectory points output by the activated neurons in the first spiking neuron includes: Based on the index of any group in the three groups of spiking neurons and the index of the three trajectory points, determine the trajectory point corresponding to any group, and apply the coordinates of the trajectory point corresponding to any group to activate the spiking neurons in any group, and obtain the spikes output by the activated neurons in any group.

3. The position prediction method according to claim 2, characterized in that, The coordinates applied to the trajectory points corresponding to any of the groups activate the pulsed neurons in any of the groups, and the pulses output by the activated neurons in any of the groups are obtained, including: Based on the abscissa and ordinate of the trajectory points corresponding to any of the groups, the abscissa pulsed neurons and ordinate pulsed neurons in any of the groups are respectively activated, and the pulses output by the activated abscissa pulsed neurons and activated ordinate pulsed neurons in any of the groups are obtained.

4. The position prediction method according to claim 1, characterized in that, The connection strength between the activated abscissa pulsed neurons in any of the groups and the second abscissa pulsed neurons is determined based on the strength control parameter corresponding to any of the groups, the index of the activated abscissa pulsed neurons in any of the groups, the index of the second abscissa pulsed neurons, and the total number of the second abscissa pulsed neurons; The connection strength between the activated ordinate pulsed neurons in any of the groups and the second ordinate pulsed neurons is determined based on the strength control parameter corresponding to any of the groups, the index of the activated ordinate pulsed neurons in any of the groups, the index of the second ordinate pulsed neurons, and the total number of the second ordinate pulsed neurons.

5. The position prediction method according to claim 4, characterized in that, The strength control parameters corresponding to each of the three pulsed neuron groups are calculated based on the following formula: In the formula, k represents the prediction time adjustment parameter; ρ1 represents the strength control parameter corresponding to the first group among the three pulsed neuron groups; ρ2 represents the strength control parameter corresponding to the second group among the three pulsed neuron groups; ρ3 represents the strength control parameter corresponding to the third group among the three pulsed neuron groups; the trajectory points corresponding to the first group are the trajectory points at the latest moment among the three trajectory points; the trajectory points corresponding to the second group are the trajectory points at the middle moment among the three trajectory points; the trajectory points corresponding to the third group are the trajectory points at the earliest moment among the three trajectory points.

6. A position prediction device, characterized in that, Including: A determination unit for determining the coordinates of three trajectory points at equal time intervals in the historical trajectory; An input unit for activating the first pulsed neurons in the position prediction model based on the coordinates of the three trajectory points, and obtaining the first pulses respectively corresponding to the three trajectory points output by the activated first pulsed neurons; A prediction unit for activating the second pulsed neurons in the position prediction model based on the first pulses respectively corresponding to the three trajectory points, and obtaining the position prediction result output by the activated second pulsed neurons; Wherein, the first pulsed neurons include three pulsed neuron groups; any of the groups includes abscissa pulsed neurons and ordinate pulsed neurons; The second pulsed neurons include second abscissa pulsed neurons and second ordinate pulsed neurons; The activating the second pulsed neurons in the position prediction model based on the pulses respectively corresponding to the three trajectory points, and obtaining the position prediction result output by the activated second pulsed neurons includes: Based on the pulses output by the activated abscissa pulse neurons in any of the groups, and in combination with the connection strength between the activated abscissa pulse neurons in any of the groups and the second abscissa pulse neurons, obtain the membrane potential of the second abscissa pulse neurons corresponding to any of the groups; and based on the pulses output by the activated ordinate pulse neurons in any of the groups, and in combination with the connection strength between the activated ordinate pulse neurons in any of the groups and the second ordinate pulse neurons, obtain the membrane potential of the second ordinate pulse neurons corresponding to any of the groups; Based on the membrane potential of the second abscissa pulse neurons and the membrane potential of the second ordinate pulse neurons corresponding to each of the three groups of pulse neurons, determine the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons; Based on the total membrane potential of the second abscissa pulse neurons and the total membrane potential of the second ordinate pulse neurons, apply a preset threshold to respectively activate the second abscissa pulse neurons and the second ordinate pulse neurons, and obtain the abscissa of the position prediction result output by the activated pulse neurons in the second abscissa pulse neurons, and the ordinate of the position prediction result output by the activated pulse neurons in the second ordinate pulse neurons.

7. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the position prediction method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by a processor, it implements the position prediction method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the position prediction method according to any one of claims 1 to 5.

Images