Parallax ranging method and system suitable for moving target and electronic equipment
By generating a parallax probability distribution sequence and calculating the connection score, the inter-frame jitter problem of autonomous driving vehicles during distance measurement of moving targets is solved, the ranging accuracy and smoothness are improved, and the vehicle control experience is improved.
Patent Information
- Application Number
- CN202411564212.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2025-01-17
AI Technical Summary
In the prior art, the autonomous driving vehicle has inter-frame jitter when measuring distances of moving targets, especially when the target is blocked, resulting in inaccurate distance measurement and affecting the smoothness of vehicle control.
By generating a disparity probability distribution sequence and introducing the concept of ‘path’, the connection scores between each probability unit are calculated, and the disparity value corresponding to the maximum connection score is selected as the target disparity, optimizing the path cost to improve ranging accuracy and smoothness.
It effectively solves the problem of inter-frame jitter, improves the ranging accuracy and smoothness of moving targets in the parallax map, and improves the control experience of autonomous driving vehicles.
Smart Images
Figure CN119321752A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of binocular vision, and in particular to a parallax ranging method suitable for moving targets. Background Art
[0002] Binocular vision technology is being widely used in the fields of autonomous driving and mobile robots due to its unique technical advantages. Usually, the target frame identified in the reference image (such as the right image) is mapped to the disparity map, and then the disparity value in the target frame is calculated, such as the mean, median, and probability distribution. The corresponding calculation result is then selected as the disparity corresponding to the target frame to obtain the distance between each target identified in the reference image and the binocular camera.
[0003] In the prior art, taking autonomous driving as an example, the target distance obtained by using this single-frame distance measurement method will have jitter between frames for the same target, especially when the target is partially blocked during movement. This jitter will lead to inaccurate target distance measurement, affecting the adaptive cruise control, AEBS alarm and other functions of the autonomous driving vehicle, and causing uneven vehicle control. Therefore, it is necessary to improve the accuracy and smoothness of parallax distance measurement of moving targets in time sequence. Summary of the invention
[0004] The purpose of this application is to improve the ranging accuracy and ranging smoothness of moving targets in disparity maps.
[0005] The technical solution of the first aspect of the present application is: a disparity ranging method suitable for moving targets is provided, the method comprising: performing moving target recognition on a reference image, obtaining a target frame, and mapping the target frame to a disparity map; generating a disparity probability distribution sequence at the current moment based on the disparity value within the target frame range in the disparity map; calculating a connection score between each probability unit in the disparity probability distribution sequence at the previous moment and any probability unit in the disparity probability distribution sequence at the current moment; selecting the disparity value corresponding to the probability unit with the highest connection score, recording it as the target disparity of the target frame, and calculating the distance of the target frame to the target based on the target disparity.
[0006] In any of the above technical solutions, further, the process of calculating the connection score specifically includes: selecting any probability unit in the disparity probability distribution series at the current moment, recorded as the first target unit, obtaining the target disparity probability of the first target unit in the disparity distribution series, and the first connection score of the first target unit at the previous moment; using a traversal method, sequentially obtaining a probability unit in the disparity probability distribution series at the previous moment, recorded as the second target unit, and calculating the path cost between the second target unit and the first target unit, wherein the path cost is determined by the disparity difference between the first target unit and the second target unit; based on the target disparity probability, the first connection score and the path cost, calculating the second connection score between the second target unit and the first target unit, and selecting the maximum value of the second connection score as the connection score.
[0007] In any of the above technical solutions, further, the process of calculating the connection score also includes: optimizing the path cost based on a preset attenuation coefficient, wherein the optimized path cost is used to calculate the second connection score.
[0008] In any of the above technical solutions, further, the calculation formula for the connection score corresponding to the i-th probability unit at time T is:
[0009]
[0010] In the formula, is the connection score corresponding to the ith probability unit at time T, max[] is the maximum value formula, is the connection score corresponding to the i-th probability unit at time T-1, path(j→i) is the path cost between the j-th probability unit at time T-1 and the i-th probability unit at time T, is the target disparity probability of the ith probability unit at time T, and α and β are preset attenuation coefficients.
[0011] In any of the above technical solutions, further, a disparity probability distribution sequence at the current moment is generated, specifically including: counting the number of disparities of disparity points in the target frame range of the disparity map that have the same disparity value as the preset statistical disparity point, wherein the preset statistical disparity point is determined by a preset disparity statistical range and a preset disparity step; based on the counted number of disparities at each preset statistical disparity point, calculating the proportion probability of each preset statistical disparity point within the target frame; based on the proportion probability, forming a disparity probability distribution sequence in the order of the disparity values of each preset statistical disparity point from small to large.
[0012] The technical solution of the second aspect of the present application is: a parallax ranging system suitable for moving targets is provided, the system comprising: an image processing unit, the image processing unit is configured to perform moving target recognition on a reference image, obtain a target frame, and map the target frame to a disparity map; a sequence generation unit, the sequence generation unit is configured to generate a disparity probability distribution sequence at a current moment based on the disparity value within the target frame range in the disparity map; a score calculation unit, the score calculation unit is configured to calculate the connection score between each probability unit in the disparity probability distribution series at a previous moment and any probability unit in the disparity probability distribution series at a current moment; a distance calculation unit, the distance calculation unit is configured to select the disparity value corresponding to the probability unit with the highest connection score, record it as the target disparity of the target frame, and calculate the distance of the target frame to the target based on the target disparity.
[0013] In any of the above technical solutions, further, the score calculation unit is specifically configured as follows: select any probability unit in the disparity probability distribution series at the current moment, record it as the first target unit, obtain the target disparity probability of the first target unit in the disparity distribution series, and the first connection score of the first target unit at the previous moment; adopt a traversal method to sequentially obtain a probability unit in the disparity probability distribution series at the previous moment, record it as the second target unit, calculate the path cost between the second target unit and the first target unit, wherein the path cost is determined by the disparity difference between the first target unit and the second target unit; calculate the second connection score between the second target unit and the first target unit based on the target disparity probability, the first connection score and the path cost, and select the maximum value of the second connection scores as the connection score.
[0014] In any of the above technical solutions, further, the score calculation unit is also configured to: optimize the path cost based on a preset attenuation coefficient, wherein the optimized path cost is used to calculate the second connection score.
[0015] In any of the above technical solutions, further, the sequence generation unit is specifically configured to: count the number of disparities of disparity points in the target frame range that have the same disparity value as the preset statistical disparity point in the statistical disparity map, wherein the preset statistical disparity point is determined by a preset disparity statistical range and a preset disparity step; calculate the proportion probability of each preset statistical disparity point within the target frame range based on the counted number of disparities at each preset statistical disparity point; and compose a disparity probability distribution series based on the proportion probability, in the order of the disparity values of each preset statistical disparity point from small to large.
[0016] The technical solution of the third aspect of the present application is: an electronic device is provided, comprising: a memory, the memory storing execution instructions; and a processor, the processor executing the execution instructions stored in the memory, so that the processor executes any one of the parallax ranging methods suitable for moving targets described in the technical solution of the first aspect.
[0017] The beneficial effects of this application are:
[0018] The technical solution in the present application is different from the existing single-frame ranging method. It first generates a disparity probability distribution sequence at the current moment through the disparity value, and the sequence is composed of multiple probability units. Then, by introducing the concept of "path", the connection score between each probability unit at the current moment and the previous moment is calculated, and the disparity value corresponding to the maximum connection score is selected as the target disparity of the target. This effectively solves the problem of inter-frame jitter in the ranging process, ensures the smoothness of disparity ranging for moving targets, improves the ranging accuracy of moving targets in the disparity map, and thus helps to improve the experience of vehicle autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] The advantages of the above and / or additional aspects of the present application will become apparent and easily understood in the description of the embodiments in conjunction with the following drawings, in which:
[0020] Figure 1 is a schematic flow chart of a parallax ranging method applicable to a moving target according to an embodiment of the present application;
[0021] Figure 2 is a schematic diagram of an occlusion target and a disparity probability distribution according to an embodiment of the present application;
[0022] Figure 3 It is a schematic block diagram of a parallax ranging system suitable for moving targets according to an embodiment of the present application. DETAILED DESCRIPTION
[0023] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0024] In the following description, many specific details are elaborated to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present application is not limited to the specific embodiments disclosed below.
[0025] Embodiment 1:
[0026] Now let's take autonomous driving as an example. Existing distance measurement solutions usually estimate the distance of a single target based on a single frame of disparity data, and then perform smoothing in the time domain using methods such as Kalman filtering.
[0027] However, this ranging method will cause inter-frame jitter, especially for moving targets with occlusion, such as Figure 1As shown in the figure, the light pole as the foreground image will interfere with the distance measurement of the vehicle target, and the distance measurement value will deviate from the true value, causing large fluctuations in the vehicle's automatic driving control. Therefore, it is necessary to improve the distance measurement accuracy and smoothness of such moving targets in the disparity map.
[0028] like Figure 1 As shown, this embodiment provides a parallax ranging method applicable to a moving target, the method comprising:
[0029] Step 1: Identify the moving target in the reference image, obtain the target frame, and map the target frame to the disparity map;
[0030] Specifically, for a vehicle-mounted binocular camera, the left and right images can be obtained by installing two cameras on a beam. Either of the left and right images is selected as a reference image, and then the corresponding disparity map can be generated based on the reference image through the disparity calculation principle. Figure 1 One to one correspondence.
[0031] The autonomous driving system uses a deep learning model to identify the reference image (such as the right image) and obtains the target frame (including left vertex coordinates, width, height, etc.), label and other information of each target in the reference image. Figure 1 Therefore, the target frame in the reference image can be mapped to the corresponding position in the disparity map. Then, by calculating the disparity value (such as mean, median) of the disparity point within the target frame, combined with the corresponding distance calculation principle, the distance between the target and the vehicle can be obtained.
[0032] Step 2: Generate the disparity probability distribution sequence at the current moment based on the disparity value within the target frame in the disparity map;
[0033] Step 3: Calculate the connection score between each probability unit in the disparity probability distribution sequence at the previous moment and any probability unit in the disparity probability distribution sequence at the current moment; wherein the connection score is determined by at least the probability dist of the probability unit, the difference in disparity between the two probability units, and the connection score of the probability unit at the previous moment;
[0034] Step 4: Select the disparity value corresponding to the probability unit with the highest connection score, record it as the target disparity of the target box, and calculate the distance between the target box and the target based on the target disparity.
[0035] Specifically, based on the set disparity value distribution interval, such as [0-300], the disparity values of the disparity points within the target frame range of the disparity map are used as samples, and the number of disparity points at each disparity value in the sample interval is counted to obtain the corresponding probability dist, and then the disparity probability distribution sequence at the current moment is formed. In this sequence, each disparity value in the interval corresponds to a unit, and the elements in the unit are the corresponding probabilities. The disparity probability distribution sequence is stored in chronological order, and the resulting sequence table is shown in Table 1.
[0036] Table 1
[0037]
[0038]
[0039] Considering that if only the disparity value corresponding to the maximum probability in each unit is selected as the disparity of the target frame at the current moment, there may be frame jumps, especially for occluded moving targets, such as Figure 2 shown.
[0040] Therefore, in order to improve the accuracy of the disparity value corresponding to the target frame and avoid inter-frame jitter, and considering that the movement process of the target in the real environment should be smooth, the present embodiment introduces the concept of "path".
[0041] When determining the disparity (target disparity) corresponding to the target frame at the current moment, the connection score between each unit (probability unit) at the previous moment and any unit at the current moment is first calculated, wherein the connection score can be calculated according to the set rules / formula based on the corresponding probability dist, the difference in disparity between the two units, and the connection score of the unit at the previous moment. Therefore, each unit at the current moment will have multiple (initial) connection scores.
[0042] Afterwards, the maximum value is selected from these connection scores as the final connection score of the unit and retained. Then, the disparity value corresponding to the unit with the largest connection score among the units at the current moment is selected as the disparity value of the target frame at the current moment, which is recorded as the target disparity.
[0043] In this way, not only the calculation result of the current frame disparity is taken into account, but also the corresponding historical data. By introducing the calculated "connectivity score" and selecting its maximum value, it helps to improve the accuracy of target disparity selection. At the same time, by introducing the difference of disparity (i.e., distance difference), it also helps to ensure the smoothness of target disparity selection.
[0044] It should be noted that after obtaining the above distance measurement results, the corresponding speed measurement results can also be obtained according to the corresponding time, and the specific process will not be repeated here.
[0045] In any of the above embodiments, further, considering that the value range of the effective disparity in the target frame should be smaller than the value range of the disparity of the entire disparity map, in order to reduce the amount of calculation and eliminate some disparities with low accuracy, step 2 specifically includes:
[0046] Step 21: Counting the number of disparity points in the target frame range of the disparity map that have the same disparity value as the preset statistical disparity point, wherein the preset statistical disparity point is determined by a preset disparity statistical range and a preset disparity step size;
[0047] Step 22: based on the counted disparity numbers at each preset statistical disparity point, calculate the proportion probability of each preset statistical disparity point within the target frame;
[0048] Step 23: Based on the proportion probability, a disparity probability distribution series is formed according to the disparity values of the preset statistical disparity points in order from small to large.
[0049] Specifically, combined with the reliability of the calculated disparity value and the corresponding ranging range, the preset disparity statistical range can be set to [50-200], and the preset disparity step size can be set to 2, that is, the number of disparity points at disparity values of 50, 52, 54, ..., 198, and 200 within the target frame range is counted in sequence. After that, the corresponding proportion probability can be calculated, and the disparity values are arranged in order from small to large (from 50 to 200) to obtain the disparity probability distribution series, where the cumulative sum of the proportion probabilities should be 1.
[0050] It should be noted that the preset disparity statistical range and the preset disparity step size can be set according to actual needs, wherein the minimum value of the preset disparity step size can be 1.
[0051] In any of the above embodiments, further, step 3 specifically includes:
[0052] Step 31: Select any probability unit in the disparity probability distribution sequence at the current moment, record it as the first target unit, and obtain the target disparity probability of the first target unit in the disparity distribution sequence and the first connection score of the first target unit at the previous moment
[0053] Step 32: Using a traversal method, obtain one probability unit in the disparity probability distribution sequence at the previous moment in turn, record it as the second target unit, and calculate the path cost path (j→i) between the second target unit and the first target unit.
[0054] Wherein, the path cost is determined by the disparity difference between the first target unit and the second target unit;
[0055] Step 33: Based on the target disparity probability First connection score and the path cost path(j→i), calculate the second connection score between the second target unit and the first target unit, and select the maximum value of the second connection scores as the connection score.
[0056] Specifically, it is assumed that time T0 is the first frame and time T1 is the second frame, and the corresponding disparity probability distribution sequence is shown in Table 2.
[0057] Table 2
[0058]
[0059] Since time T0 is the first frame, the connection score score of the probability unit corresponding to each disparity value at this moment is the same as the value of the target disparity probability dist.
[0060] At time T1, the i-th probability unit is used as an example for explanation. When calculating the (second) connection score of the i-th probability unit, the corresponding target disparity probability can be obtained by querying First connection score Select a probability unit in [0,300] in turn, record it as the second target unit, and calculate the path cost path(j→i) between the second target unit and the first target unit.
[0061] Considering that the connection score and the target disparity probability are both decimals, and the path cost should be a negative value, the path cost is set to -1 / 10 of the disparity difference between the two target units. For example, when probability unit i+2 is selected as the second target unit, that is, j=i+2, the corresponding path cost value is -|i-(i+2)| / 10=-0.2.
[0062] Therefore, the second connection score between probability unit i+2 (T0 time) and probability unit i (T1 time) can be the target disparity probability First connection score The path cost path(j→i)=-0.2 is the sum of the three.
[0063] Therefore, each probability unit at time T0 will have a second connection score corresponding to the i-th probability unit at time T1, and the maximum value of these second connection scores is selected and saved as the connection score of the i-th probability unit. In this way, each probability unit at time T1 will correspond to a unique connection score.
[0064] In any of the above embodiments, further, considering that the corresponding score should be attenuated during the path propagation process and the difference in the connection scores of different paths is increased, an attenuation coefficient is introduced when calculating the connection score. Therefore, step 3 also includes:
[0065] The path cost is optimized based on a preset attenuation coefficient, wherein the optimized path cost is used to calculate the second connection score.
[0066] In any of the above embodiments, further, the calculation formula for the connection score corresponding to the i-th probability unit at time T is:
[0067]
[0068] In the formula, is the connection score corresponding to the ith probability unit at time T, max[] is the maximum value formula, is the connection score corresponding to the i-th probability unit at time T-1, path(j→i) is the path cost between the j-th probability unit at time T-1 and the i-th probability unit at time T, is the target disparity probability of the ith probability unit at time T, α and β are preset attenuation coefficients, and their values can be set according to actual needs, such as α=-0.1, β=1.
[0069] Embodiment 2:
[0070] like Figure 3 As shown, the present embodiment also provides a parallax ranging system 100 suitable for moving targets, the system 100 includes: an image processing unit 101, the image processing unit 101 is configured to perform moving target recognition on a reference image, obtain a target frame, and map the target frame to a disparity map; a sequence generation unit 102, the sequence generation unit 102 is configured to generate a disparity probability distribution sequence at a current moment based on the disparity value within the target frame range in the disparity map; a score calculation unit 103, the score calculation unit 103 is configured to calculate the connection score between each probability unit in the disparity probability distribution series at a previous moment and any probability unit in the disparity probability distribution series at a current moment; a distance calculation unit 104, the distance calculation unit 104 is configured to select the disparity value corresponding to the probability unit with the highest connection score, record it as the target disparity of the target frame, and calculate the distance of the target frame to the target based on the target disparity.
[0071] In any of the above embodiments, further, the score calculation unit 103 is specifically configured to: select any probability unit in the disparity probability distribution series at the current moment, record it as the first target unit, obtain the target disparity probability of the first target unit in the disparity distribution series, and the first connection score of the first target unit at the previous moment; adopt a traversal method to sequentially obtain a probability unit in the disparity probability distribution series at the previous moment, record it as the second target unit, calculate the path cost between the second target unit and the first target unit, wherein the path cost is determined by the disparity difference between the first target unit and the second target unit; based on the target disparity probability, the first connection score and the path cost, calculate the second connection score between the second target unit and the first target unit, and select the maximum value of the second connection score as the connection score.
[0072] In any of the above embodiments, further, the score calculation unit 103 is further configured to: optimize the path cost based on a preset attenuation coefficient, wherein the optimized path cost is used to calculate the second connection score. The calculation formula of the (second) connection score corresponding to the i-th probability unit at time T is:
[0073]
[0074] In the formula, is the connection score corresponding to the ith probability unit at time T, max[] is the maximum value formula, is the connection score corresponding to the i-th probability unit at time T-1, path(j→i) is the path cost between the j-th probability unit at time T-1 and the i-th probability unit at time T, is the target disparity probability of the ith probability unit at time T, and α and β are preset attenuation coefficients.
[0075] In any of the above embodiments, further, the sequence generation unit 102 is specifically configured to: count the number of disparities of disparity points within the target frame range of the disparity map that have the same disparity value as a preset statistical disparity point, wherein the preset statistical disparity point is determined by a preset disparity statistical range and a preset disparity step; calculate the proportion probability of each preset statistical disparity point within the target frame range based on the counted number of disparities at each preset statistical disparity point; and form a disparity probability distribution series based on the proportion probability, in the order of the disparity values of each preset statistical disparity point from small to large.
[0076] Embodiment 3:
[0077] This embodiment further provides an electronic device, including: a memory, the memory storing execution instructions; and a processor or other hardware module, the processor or other hardware module executes the execution instructions stored in the memory, so that the processor or other hardware module executes the parallax ranging method suitable for moving targets in the above embodiment.
[0078] So far, various embodiments of the present application have been described in detail. In order to avoid obscuring the concept of the present application, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed herein.
[0079] Although some specific embodiments of the present application have been described in detail through examples, those skilled in the art should understand that the above examples are only for illustration rather than for limiting the scope of the present application.
[0080] The steps in this application can be adjusted in order, combined, and deleted according to actual needs.
[0081] Although the present application is disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely exemplary and are not intended to limit the application of the present application. The scope of protection of the present application is defined by the appended claims and may include various modifications, alterations and equivalents made to the invention without departing from the scope and spirit of the present application.
Claims
1. A parallax ranging method suitable for moving targets, characterized in that: The method comprises: Performing moving target recognition on the reference image, obtaining a target frame, and mapping the target frame to a disparity map; Based on the disparity values within the target frame range in the disparity map, generating a disparity probability distribution sequence at a current moment; Calculate the connection score between each probability unit in the disparity probability distribution sequence at the previous moment and any probability unit in the disparity probability distribution sequence at the current moment; The disparity value corresponding to the probability unit with the highest connection score is selected as the target disparity of the target frame, and the distance between the target frame and the target is calculated based on the target disparity.
2. The parallax ranging method for moving targets according to claim 1, characterized in that: The process of calculating the connection score specifically includes: Select any probability unit in the disparity probability distribution sequence at the current moment, record it as a first target unit, obtain the target disparity probability of the first target unit in the disparity distribution sequence, and the first connection score of the first target unit at the previous moment; By traversing, a probability unit in the disparity probability distribution sequence at the previous moment is obtained in turn, recorded as the second target unit, and the path cost between the second target unit and the first target unit is calculated. Wherein, the path cost is determined by a disparity difference between the first target unit and the second target unit; Based on the target disparity probability, the first connection score and the path cost, a second connection score between the second target unit and the first target unit is calculated, and a maximum value of the second connection scores is selected as the connection score.
3. The parallax ranging method for moving targets as claimed in claim 2, characterized in that: The process of calculating the connection score further includes: The path cost is optimized based on a preset attenuation coefficient, wherein the optimized path cost is used to calculate the second connection score.
4. The parallax ranging method for moving targets as claimed in claim 3, characterized in that: The calculation formula of the connection score corresponding to the i-th probability unit at time T is: In the formula, is the connection score corresponding to the ith probability unit at time T, max[ ] is the maximum value formula, is the connection score corresponding to the i-th probability unit at time T-1, path(j→i) is the path cost between the j-th probability unit at time T-1 and the i-th probability unit at time T, is the target disparity probability of the ith probability unit at time T, and α and β are the preset attenuation coefficients.
5. The parallax ranging method applicable to a moving target according to any one of claims 1 to 4, characterized in that: The generating of the disparity probability distribution sequence at the current moment specifically includes: Counting the number of disparity points within the target frame range in the disparity map that have the same disparity value as the preset statistical disparity point, The preset statistical disparity point is determined by a preset disparity statistical range and a preset disparity step size; Based on the counted number of disparities at each of the preset statistical disparity points, calculating the proportion probability of each of the preset statistical disparity points within the target frame; Based on the proportion probability, the disparity probability distribution series is composed according to the disparity values of the preset statistical disparity points in order from small to large.
6. A parallax ranging system suitable for moving targets, characterized in that: The system comprises: An image processing unit, wherein the image processing unit is configured to perform moving target recognition on the reference image, obtain a target frame, and map the target frame to a disparity map; A sequence generating unit, wherein the sequence generating unit is configured to generate a disparity probability distribution sequence at a current moment based on the disparity value within the target frame range in the disparity map; A score calculation unit, wherein the score calculation unit is configured to calculate a connection score between each probability unit in the disparity probability distribution sequence at the previous moment and any probability unit in the disparity probability distribution sequence at the current moment; A distance calculation unit is configured to select a disparity value corresponding to the probability unit with the highest connection score, record it as the target disparity of the target frame, and calculate the distance between the target frame and the target based on the target disparity.
7. The parallax ranging system for moving targets as claimed in claim 6, characterized in that: The score calculation unit is specifically configured as follows: Select any probability unit in the disparity probability distribution sequence at the current moment, record it as a first target unit, obtain the target disparity probability of the first target unit in the disparity distribution sequence, and the first connection score of the first target unit at the previous moment; By traversing, a probability unit in the disparity probability distribution sequence at the previous moment is obtained in turn, recorded as the second target unit, and the path cost between the second target unit and the first target unit is calculated. Wherein, the path cost is determined by a disparity difference between the first target unit and the second target unit; Based on the target disparity probability, the first connection score and the path cost, a second connection score between the second target unit and the first target unit is calculated, and a maximum value of the second connection scores is selected as the connection score.
8. The parallax ranging system for moving targets according to claim 6 or 7, characterized in that: The score calculation unit is further configured to: The path cost is optimized based on a preset attenuation coefficient, wherein the optimized path cost is used to calculate the second connection score.
9. The parallax ranging system suitable for moving targets according to any one of claims 6 to 8, characterized in that: The sequence generation unit is specifically configured as follows: Counting the number of disparity points within the target frame range in the disparity map that have the same disparity value as the preset statistical disparity point, The preset statistical disparity point is determined by a preset disparity statistical range and a preset disparity step size; Based on the counted number of disparities at each of the preset statistical disparity points, calculating the proportion probability of each of the preset statistical disparity points within the target frame; Based on the proportion probability, the disparity probability distribution series is composed according to the disparity values of the preset statistical disparity points in order from small to large.
10. An electronic device, characterized in that: include: A memory storing execution instructions; as well as A processor, wherein the processor executes the execution instructions stored in the memory, so that the processor executes the parallax ranging method applicable to a moving target according to any one of claims 1 to 5.