-
data input
- orientation diff between previous record and current.
Vector3d pre_ornt = ornt_data.row(i-1); Vector3d current_ornt = ornt_data.row(i); Vector3d ornt_diff = current_ornt - pre_ornt;- normalized accelerometer in b frame and n frame.
Vector4d q = quaternions.GetQFromEuler(current_ornt); Matrix3d dcm = quaternions.GetDCMFromQ(q); Vector3d acc_v = acc_data; Vector3d acc_n = dcm * acc_v; Vector3d acc_v_norm = Normalise(acc_v); Vector3d acc_n_norm = Normalise(acc_n);- normalized gravity in b frame and n frame.
Vector3d g_v = g_data; Vector3d g_n = dcm * g_v; Vector3d g_v_norm = Normalise(g_v); Vector3d g_n_norm = Normalise(g_n);- diff between accelerometer and gravity in n frame.
Vector3d a_diff = acc_n - g_n;- normalized gyroscope.
Vector3d gyro_v = gyro_data; Vector3d gyro_v_norm = Normalise(gyro_v);- normalized magnetic diff between previous record and current in n frame.
Vector3d pre_mag_v = mag_data.row(i-1); Vector3d current_mag_v = mag_data.row(i); Vector3d mag_v_diff = (dcm * current_mag_v) - (dcm * pre_mag_v); Vector3d mag_v_norm = accelerometer.Normalise(mag_v_diff);- examples:
| acc_b_norm_x | acc_b_norm_y | acc_b_norm_z | acc_n_norm_x | acc_n_norm_y | acc_n_norm_z | g_b_norm_x | g_b_norm_y | g_b_norm_z | g_n_norm_x | g_n_norm_y | g_n_norm_z | gyro_norm_x | gyro_norm_y | gyro_norm_z | mag_n_norm_x | mag_n_norm_y | mag_n_norm_z | mag_n_diff_norm_x | mag_n_diff_norm_y | mag_n_diff_norm_z | acc_g_n_diff_x | acc_g_n_diff_y | acc_g_n_diff_z | ornt_diff_x | ornt_diff_y | ornt_diff_z | label |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.129066189 | 0.489784426 | 0.862237285 | -0.504075138 | 0.568416979 | 0.65023872 | 0.161449964 | 0.394004726 | 0.904817211 | -0.406554964 | 0.597145075 | 0.691470043 | 0.457245866 | -0.07097188 | -0.886503925 | 0.482868564 | 0.285974018 | 0.827681588 | 3.382281287 | 2.003121847 | 5.797544418 | -1.374394617 | 0.189218618 | 0.134366798 | -2.456686 | 2.903637 | -2.6312 | 1 |
| 0.098092756 | 0.441567958 | 0.891849511 | -0.417273415 | 0.538611451 | 0.731970356 | 0.115516734 | 0.376209022 | 0.919305529 | -0.350114553 | 0.550039758 | 0.758205819 | -0.52091589 | 0.693465674 | -0.497746918 | 0.485898133 | -0.436930527 | -0.756964146 | 1.580799524 | -1.421490477 | -2.462673718 | -0.715750162 | -0.038637689 | -0.157519063 | 2.954984 | -1.104863 | -1.016671 | 1 |
| 0.128268463 | 0.385579444 | 0.913715324 | -0.340256272 | 0.575695299 | 0.74350561 | 0.11413304 | 0.384831439 | 0.915903058 | -0.34341474 | 0.563926869 | 0.751034489 | 0.785250568 | -0.150309687 | -0.60065676 | 0.722509897 | 0.322986569 | 0.611276635 | 0.872183616 | 0.389895827 | 0.737907491 | 0.210543669 | -0.188356055 | -0.466219017 | 0.05889 | 0.534217 | -1.278368 | 1 |
| 0.090377072 | 0.506492751 | 0.857494652 | -0.430280858 | 0.540274825 | 0.723160768 | 0.0714945 | 0.373371378 | 0.924922889 | -0.305098583 | 0.513380471 | 0.80209435 | -0.40415878 | 0.907416845 | -0.115110168 | -0.018786556 | -0.478580999 | -0.877842408 | -0.056784486 | -1.446565096 | -2.653377777 | -1.56801271 | 0.690716987 | -0.202952674 | 2.683109 | -0.709583 | 0.002222 | 1 |
| 0.003769136 | 0.413187081 | 0.910638364 | -0.273759419 | 0.441200565 | 0.854633162 | -0.006953619 | 0.349818259 | 0.936791777 | -0.214272881 | 0.418783958 | 0.882443839 | -0.299333594 | 0.767442181 | -0.566949644 | 0.452037207 | -0.074923127 | -0.888846943 | 1.474141036 | -0.244332222 | -2.898623682 | -0.583401164 | 0.219646233 | -0.273327666 | 4.845336 | -1.44749 | -2.451917 | 1 |
| -0.012789521 | 0.414001052 | 0.910186551 | -0.238513744 | 0.437943501 | 0.866785258 | -0.033002042 | 0.360327767 | 0.932241796 | -0.192046192 | 0.40733654 | 0.892857885 | 0.761586392 | 0.040498834 | -0.646796732 | 0.696550054 | -0.020242884 | -0.717222593 | 1.845812816 | -0.053642339 | -1.900593713 | -0.3771996 | 0.155847649 | -0.54158446 | 1.602177 | 0.644123 | -1.650386 | 1 |
| -0.08225435 | -0.306475501 | 0.948317979 | -0.278243257 | 0.417762745 | 0.864901716 | -0.059733105 | -0.286562656 | 0.956197574 | -0.263618426 | 0.394924855 | 0.880079363 | -0.425352445 | -0.817550551 | -0.388183453 | 0.75555205 | -0.627310411 | -0.188739898 | 0.464302059 | -0.385494971 | -0.115984495 | -0.155815903 | 0.242576808 | -0.110514225 | -0.023439 | 0.017182 | -0.004636 | 0 |
| -0.072043 | -0.295740867 | 0.952547713 | -0.269786878 | 0.407713857 | 0.872344227 | -0.05961394 | -0.287154194 | 0.956027535 | -0.264061745 | 0.395357894 | 0.879751971 | -0.463747259 | -0.805428147 | -0.369085328 | -0.402699867 | 0.40874646 | -0.81899887 | -0.229401034 | 0.232845521 | -0.466548918 | -0.066589347 | 0.136967852 | -0.038958311 | -0.006482 | -0.035383 | 0.009744 | 0 |
| -0.06279196 | -0.294665 | 0.953535373 | -0.270584795 | 0.399787014 | 0.87575922 | -0.059551967 | -0.287931685 | 0.955797525 | -0.264696252 | 0.396037988 | 0.879255256 | -0.259987511 | -0.746441091 | -0.61256199 | 0.52469785 | -0.850812152 | 0.028475391 | 0.171805847 | -0.278587957 | 0.009323916 | -0.044139517 | 0.016644083 | -0.078400482 | -0.002848 | -0.046509 | 0.009805 | 0 |
| -0.060826806 | -0.287624644 | 0.95580969 | -0.26442447 | 0.397496032 | 0.878678897 | -0.059906299 | -0.288264403 | 0.955675086 | -0.265247511 | 0.396729437 | 0.878777282 | 0.260738265 | -0.506384611 | -0.821942932 | 0.162231636 | -0.868545219 | 0.468305562 | 0.066946532 | -0.358414006 | 0.193251046 | 0.034322951 | -0.031938411 | -0.088189996 | 0.02162 | -0.019908 | 0.002586 | 0 |
| -0.069062529 | -0.288161093 | 0.955088243 | -0.263882463 | 0.405076711 | 0.87537358 | -0.060623123 | -0.288152188 | 0.955663724 | -0.265685523 | 0.397368956 | 0.878355916 | 0.06959041 | -0.568321684 | -0.819858304 | 0.812688923 | -0.011961465 | 0.582575006 | 0.219561499 | -0.00323159 | 0.157392377 | 0.059482981 | 0.011454942 | -0.16790262 | 0.042848 | 0.006716 | -0.002917 | 0 |
- training
import xgboost as xgb
train_x, train_y, test_x, test_y = LoadData()
x, y = LoadVaildData()
D_train = xgb.DMatrix(train_x, label=train_y)
D_test = xgb.DMatrix(test_x, label=test_y)
D_validation = xgb.DMatrix(x, label=y)
param = {
'objective':'multi:softprob',
'eta': 0.3,
'max_depth': 10,
'num_class': 2
}
start_time = time.time()
model = xgb.train(param, D_train, 100)
end_time = time.time()
test_preds = model.predict(D_test)
test_best_preds = np.asarray([np.argmax(line) for line in test_preds])
vali_preds = model.predict(D_validation)
vali_best_preds = np.asarray([np.argmax(line) for line in vali_preds])
# dump model
model.dump_model('./raw_model.txt', dump_format='txt')
print("Cost time: " + str(int(end_time - start_time)))
print(classification_report(test_y, test_best_preds, target_names=["停车", "运动"]))
print(classification_report(y, vali_best_preds, target_names=["停车", "运动"]))
- rewrite the prediction in c++
#include "utils/Tools.h"
#include "models/XgboostDetector.h"
std::string model_path = "raw_model.txt";
XgboostDetector xgboostDetector = XgboostDetector(model_path);
StopDetection &stopDetection = xgboostDetector;
bool is_stop = stopDetection.IsStopping(input_data);