本文主要是讲目标跟踪sort中的卡尔曼滤波在上述匈牙利匹配结果的基础上,怎样实现如下过程:
1、怎样predict ?
2、怎样update ?
3、track.hit_streak怎样实现连击 >= min_hits时,赋予该track一个id ?
4、track.time_since_update > max_age时,删除该轨迹track的实现方式?
具体的实现过程讲解:
python使用卡尔曼滤波的几个包:
1、从opencv导入
2、从filterpy导入
3、自己写一个 class Kalman()
代码中用到的卡尔曼滤波器类from filterpy.kalman import KalmanFilter:
查看代码详细注释,应该能搞懂。
# -*- coding: utf-8 -*-
"""
Time : 2022/5/13 11:22
Author : cong
"""
from filterpy.kalman import KalmanFilter
import numpy as np
def convert_bbox_to_z(bbox):
"""
Takes a bounding box in the form [x1,y1,x2,y2] and returns z in the form
[x,y,s,r] where x,y is the centre of the box and s is the scale/area and r is
the aspect ratio
"""
w = bbox[2] - bbox[0]
h = bbox[3] - bbox[1]
x = bbox[0] + w / 2.
y = bbox[1] + h / 2.
s = w * h # scale is just area
r = w / float(h)
return np.array([x, y, s, r]).reshape((4, 1))
def convert_x_to_bbox(x, score=None):
"""
Takes a bounding box in the centre form [x,y,s,r] and returns it in the form
[x1,y1,x2,y2] where x1,y1 is the top left and x2,y2 is the bottom right
"""
w = np.sqrt(x[2] * x[3])
h = x[2] / w
if score is None:
return np.array([x[0] - w / 2., x[1] - h / 2., x[0] + w / 2., x[1] + h / 2.]).reshape((1, 4))
else:
return np.array([x[0] - w / 2., x[1] - h / 2., x[0] + w / 2., x[1] + h / 2., score]).reshape((1, 5))
class KalmanBoxTracker(object):
"""
This class represents the internal state of individual tracked objects observed as bbox.
"""
count = 0
def __init__(self, bbox):
"""
Initialises a tracker using initial bounding box.
"""
# define constant velocity model
# x= [x,y,s,r,vx,vy,vs], z=[x,y,s,r]
# 初始化卡尔曼滤波器参数,7个状态变量和4个观测输入,运动形式和转换矩阵的确定都是基于匀速运动模型
self.kf = KalmanFilter(dim_x=7, dim_z=4)
# 状态转移矩阵F
self.kf.F = np.array(
[[1, 0, 0, 0, 1, 0, 0], [0, 1, 0, 0, 0, 1, 0], [0, 0, 1, 0, 0, 0, 1], [0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0], [0, 0, 0, 0, 0, 1, 0], [0, 0, 0, 0, 0, 0, 1]])
# 观测矩阵
self.kf.H = np.array(
[[1, 0, 0, 0, 0, 0, 0], [0, 1, 0, 0, 0, 0, 0], [0, 0, 1, 0, 0, 0, 0], [0, 0, 0, 1, 0, 0, 0]])
# 观测噪声协方差矩阵
self.kf.R[2:, 2:] *= 10.
# 先验估计协方差
self.kf.P[4:, 4:] *= 1000. # give high uncertainty to the unobservable initial velocities
self.kf.P *= 10.
# 状态噪声协方差
self.kf.Q[-1, -1] *= 0.01
self.kf.Q[4:, 4:] *= 0.01
self.kf.x[:4] = convert_bbox_to_z(bbox)
self.time_since_update = 0
self.id = KalmanBoxTracker.count
KalmanBoxTracker.count += 1
self.history = []
self.hits = 0
self.hit_streak = 0
self.age = 0
def update(self, bbox):
"""
Updates the state vector with observed bbox.
"""
self.time_since_update = 0
self.history = []
self.hits += 1
self.hit_streak += 1
self.kf.update(convert_bbox_to_z(bbox))
def predict(self):
"""
Advances the state vector and returns the predicted bounding box estimate.
"""
if (self.kf.x[6] + self.kf.x[2]) <= 0:
self.kf.x[6] *= 0.0
self.kf.predict()
self.age += 1
if self.time_since_update > 0:
self.hit_streak = 0
self.time_since_update += 1
self.history.append(convert_x_to_bbox(self.kf.x))
return self.history[-1]
def get_state(self):
"""
Returns the current bounding box estimate.
"""
return convert_x_to_bbox(self.kf.x)
参考:
https://www.likecs.com/show-331696.html
https://blog.csdn.net/zimiao552147572/article/details/105985289
文章出处登录后可见!
已经登录?立即刷新