我正在尝试基于从2D坐标的两个点的像素偏移来计算帧之间的实际速度。
以co_ordinates[frame_number][pixel_index][pixel_value_x][pixel_value_y]的方式存储坐标>
例如，这是frame_number=1时显示pixel_index=1和co_ordinates[0][0]的方式>
array([[113. , 231. ], [112.83630458, 230.73530376]])这里是尝试
import math def calc(image_dim , real_Dim , Point): return (Point[0] * real_Dim[0]/image_dim[0], Point[1] * real_Dim[1]/image_dim[1]) image_dim = (320,240) # dimension of image (width x height) real_Dim = (1,0.9) # real dimension of image in mm (width x height) number_of_frames = 120 # Total number of frames number_of_points =58 # Number of co-ordinates where readings are taken in an image video_speed = [] for frame_number in range(0,number_of_frames): # loop over all frames frame_speed = [] for point_number in range(0, number_of_points): # calculate speed per pixel_index per frame Point_one = calc(image_dim, real_Dim, (co_ordinates[frame_number][point_number][0][0], co_ordinates[frame_number][point_number][1][0])) Point_two = calc(image_dim, real_Dim, (co_ordinates[frame_number][point_number][0][1], co_ordinates[frame_number][point_number][1][1])) distance = math.sqrt((Point_one[0]-Point_two[0])**2 + (Point_one[1]-Point_two[1])**2) time = 0.05 # Time between frames frame_speed.append(distance / time) video_speed.append(np.mean(frame_speed)) # calculate mean speed of frame如何正确计算
mm/s中各帧之间的实际速度？如果正确，是否有进一步优化的方法？
EXTRAS：
出于测试目的，这是来自坐标的6帧的输出
[array([[[ 50. , 50. ], [ 49.89107906, 50.01721625]], [[150. , 50. ], [149.98240056, 50.01998935]], [[250. , 50. ], [249.64290327, 51.01493382]], [[ 50. , 150. ], [ 49.92744978, 150.02948231]], [[150. , 150. ], [149.64272794, 149.82678588]], [[250. , 150. ], [250.00543924, 150.0026971 ]]]), array([[[ 50. , 50. ], [ 50.03984383, 49.97097593]], [[150. , 50. ], [150.01179785, 49.99240989]], [[250. , 50. ], [250.23686095, 48.90773654]], [[ 50. , 150. ], [ 50.10918073, 151.85269225]], [[150. , 150. ], [150.19248188, 149.56399652]], [[250. , 150. ], [249.99657853, 150.00377174]]]), array([[[ 50. , 50. ], [ 50.02103496, 50.0230453 ]], [[150. , 50. ], [149.9971778 , 50.00171494]], [[250. , 50. ], [249.97102118, 49.13421589]], [[ 50. , 150. ], [ 50.00297619, 148.87509906]], [[150. , 150. ], [150.31536585, 148.69849575]], [[250. , 150. ], [249.99691853, 149.99868659]]]), array([[[ 50. , 50. ], [ 50.00037101, 50.00599535]], [[150. , 50. ], [149.99339349, 50.00473171]], [[250. , 50. ], [249.67023763, 50.2115393 ]], [[ 50. , 150. ], [ 49.95433414, 150.75595504]], [[150. , 150. ], [149.54204383, 146.9134376 ]], [[250. , 150. ], [250.00658676, 150.00301765]]]), array([[[ 50. , 50. ], [ 50.00842951, 49.97589724]], [[150. , 50. ], [149.9824532 , 50.00823016]], [[250. , 50. ], [250.59972095, 50.823488 ]], [[ 50. , 150. ], [ 49.94134316, 148.85299742]], [[150. , 150. ], [150.69609386, 155.63426733]], [[250. , 150. ], [249.99170064, 150.00060952]]])]我正在尝试基于从2D坐标的两个点的像素偏移来计算帧之间的实际速度。以以下方式存储坐标：co_ordinates [frame_number] [...
您的代码应按原样工作。您可能需要首先计算帧点云重心的坐标，然后从that
推断帧速度。这样更快，并具有稍微减少错误传播的优点：