Files
RFG_stm32_ADC_receiver_GUI/rfg_adc_plotter/processing/pipeline.py
2026-02-25 20:20:40 +03:00

416 lines
18 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

"""Явный pipeline предобработки свипов перед помещением в RingBuffer."""
from __future__ import annotations
from dataclasses import dataclass
import os
from typing import Optional, Tuple
import numpy as np
from rfg_adc_plotter.io.capture_reference_loader import (
CaptureParseSummary,
aggregate_capture_reference,
detect_reference_file_format,
load_capture_sweeps,
)
from rfg_adc_plotter.processing.normalizer import (
build_calib_envelopes,
normalize_by_calib,
normalize_by_envelope,
)
DEFAULT_CALIB_ENVELOPE_PATH = "calib_envelope.npy"
DEFAULT_BACKGROUND_PATH = "background.npy"
def _normalize_path(path: str) -> str:
return str(path).strip()
def _normalize_save_npy_path(path: str) -> str:
p = _normalize_path(path)
if not p:
return p
_root, ext = os.path.splitext(p)
if ext:
return p
return f"{p}.npy"
def _summary_for_npy(path: str) -> CaptureParseSummary:
return CaptureParseSummary(
path=path,
format="npy",
sweeps_total=0,
sweeps_valid=0,
channels_seen=tuple(),
dominant_width=None,
dominant_n_valid=None,
aggregation="median",
warnings=tuple(),
)
@dataclass(frozen=True)
class SweepProcessingResult:
"""Результат предобработки одного свипа."""
processed_sweep: np.ndarray
normalized_sweep: Optional[np.ndarray]
calibration_applied: bool
background_applied: bool
calibration_source: str # off|live|npy|capture
background_source: str # off|npy|capture(raw)|capture(raw->calib)
is_calibration_reference: bool
stage_trace: Tuple[str, ...]
class SweepPreprocessor:
"""Управляет калибровкой/фоном и применяет их к входному свипу."""
def __init__(
self,
norm_type: str = "projector",
calib_envelope_path: str = DEFAULT_CALIB_ENVELOPE_PATH,
background_path: str = DEFAULT_BACKGROUND_PATH,
auto_save_live_calib_envelope: bool = True,
):
self.norm_type = str(norm_type).strip().lower() or "projector"
self.calib_enabled = False
self.calib_mode = "live" # live | file
self.background_enabled = False
self.auto_save_live_calib_envelope = bool(auto_save_live_calib_envelope)
self.calib_envelope_path = _normalize_path(calib_envelope_path)
self.background_path = _normalize_path(background_path)
self.last_calib_sweep: Optional[np.ndarray] = None
self.calib_file_envelope: Optional[np.ndarray] = None
# background — в текущем домене вычитания (raw или normalized), UI использует для preview/state
self.background: Optional[np.ndarray] = None
# raw background loaded from capture file; преобразуется на лету при активной калибровке
self.background_raw_capture: Optional[np.ndarray] = None
# Источники и метаданные загрузки
self.calib_external_source_type: str = "none" # none|npy|capture
self.background_source_type: str = "none" # none|npy_processed|capture_raw
self.calib_reference_summary: Optional[CaptureParseSummary] = None
self.background_reference_summary: Optional[CaptureParseSummary] = None
self.last_reference_error: str = ""
# Параметры офлайн-парсинга capture (должны совпадать с live parser по настройке UI)
self.capture_fancy: bool = False
self.capture_logscale: bool = False
self.reference_aggregation_method: str = "median"
# ---- Конфигурация ----
def set_calib_mode(self, mode: str):
m = str(mode).strip().lower()
self.calib_mode = "file" if m == "file" else "live"
def set_calib_enabled(self, enabled: bool):
self.calib_enabled = bool(enabled)
def set_background_enabled(self, enabled: bool):
self.background_enabled = bool(enabled)
def set_capture_parse_options(self, *, fancy: Optional[bool] = None, logscale: Optional[bool] = None):
if fancy is not None:
self.capture_fancy = bool(fancy)
if logscale is not None:
self.capture_logscale = bool(logscale)
def set_calib_envelope_path(self, path: str):
p = _normalize_path(path)
if p:
if p != self.calib_envelope_path:
self.calib_file_envelope = None
if self.calib_external_source_type in ("npy", "capture"):
self.calib_external_source_type = "none"
self.calib_reference_summary = None
self.calib_envelope_path = p
def set_background_path(self, path: str):
p = _normalize_path(path)
if p:
if p != self.background_path:
self.background = None
self.background_raw_capture = None
self.background_source_type = "none"
self.background_reference_summary = None
self.background_path = p
def has_calib_envelope_file(self) -> bool:
return bool(self.calib_envelope_path) and os.path.isfile(self.calib_envelope_path)
def has_background_file(self) -> bool:
return bool(self.background_path) and os.path.isfile(self.background_path)
# ---- Загрузка/сохранение .npy ----
def _save_array(self, arr: np.ndarray, current_path: str, path: Optional[str]) -> str:
target = _normalize_save_npy_path(path if path is not None else current_path)
if not target:
raise ValueError("Пустой путь сохранения")
np.save(target, arr)
return target
def save_calib_envelope(self, path: Optional[str] = None) -> bool:
"""Сохранить огибающую из последнего live-калибровочного свипа (экспорт .npy)."""
if self.last_calib_sweep is None:
return False
try:
_lower, upper = build_calib_envelopes(self.last_calib_sweep)
self.calib_envelope_path = self._save_array(upper, self.calib_envelope_path, path)
self.last_reference_error = ""
return True
except Exception as exc:
self.last_reference_error = f"save calib envelope failed: {exc}"
return False
def save_background(self, sweep_for_ring: Optional[np.ndarray], path: Optional[str] = None) -> bool:
"""Сохранить текущий свип (в текущем домене обработки) как .npy-фон."""
if sweep_for_ring is None:
return False
try:
bg = np.asarray(sweep_for_ring, dtype=np.float32).copy()
self.background_path = self._save_array(bg, self.background_path, path)
self.background = bg
self.background_raw_capture = None
self.background_source_type = "npy_processed"
self.background_reference_summary = _summary_for_npy(self.background_path)
self.last_reference_error = ""
return True
except Exception as exc:
self.last_reference_error = f"save background failed: {exc}"
return False
# ---- Загрузка эталонов (.npy или capture) ----
def _detect_source_kind(self, path: str, source_kind: str) -> Optional[str]:
sk = str(source_kind).strip().lower() or "auto"
if sk == "auto":
return detect_reference_file_format(path)
if sk in ("npy", "bin_capture", "capture"):
return "bin_capture" if sk == "capture" else sk
return None
def _load_npy_vector(self, path: str) -> np.ndarray:
arr = np.load(path)
return np.asarray(arr, dtype=np.float32).reshape(-1)
def load_calib_reference(
self,
path: Optional[str] = None,
*,
source_kind: str = "auto",
method: str = "median",
) -> bool:
"""Загрузить калибровку из .npy (огибающая) или raw capture файла."""
if path is not None:
self.set_calib_envelope_path(path)
p = self.calib_envelope_path
if not p or not os.path.isfile(p):
self.last_reference_error = f"Файл калибровки не найден: {p}"
return False
fmt = self._detect_source_kind(p, source_kind)
if fmt is None:
self.last_reference_error = f"Неизвестный формат файла калибровки: {p}"
return False
try:
if fmt == "npy":
env = self._load_npy_vector(p)
self.calib_file_envelope = env
self.calib_external_source_type = "npy"
self.calib_reference_summary = _summary_for_npy(p)
self.last_reference_error = ""
return True
sweeps = load_capture_sweeps(p, fancy=self.capture_fancy, logscale=self.capture_logscale)
vec, summary = aggregate_capture_reference(
sweeps,
channel=0,
method=method or self.reference_aggregation_method,
path=p,
)
_lower, upper = build_calib_envelopes(vec)
self.calib_file_envelope = np.asarray(upper, dtype=np.float32)
self.calib_external_source_type = "capture"
self.calib_reference_summary = summary
self.last_reference_error = ""
return True
except Exception as exc:
self.last_reference_error = f"Ошибка загрузки калибровки: {exc}"
return False
def load_background_reference(
self,
path: Optional[str] = None,
*,
source_kind: str = "auto",
method: str = "median",
) -> bool:
"""Загрузить фон из .npy (готовый домен) или raw capture файла."""
if path is not None:
self.set_background_path(path)
p = self.background_path
if not p or not os.path.isfile(p):
self.last_reference_error = f"Файл фона не найден: {p}"
return False
fmt = self._detect_source_kind(p, source_kind)
if fmt is None:
self.last_reference_error = f"Неизвестный формат файла фона: {p}"
return False
try:
if fmt == "npy":
bg = self._load_npy_vector(p)
self.background = bg
self.background_raw_capture = None
self.background_source_type = "npy_processed"
self.background_reference_summary = _summary_for_npy(p)
self.last_reference_error = ""
return True
sweeps = load_capture_sweeps(p, fancy=self.capture_fancy, logscale=self.capture_logscale)
vec, summary = aggregate_capture_reference(
sweeps,
channel=0,
method=method or self.reference_aggregation_method,
path=p,
)
self.background_raw_capture = np.asarray(vec, dtype=np.float32)
# Для UI/preview текущий background отражает текущий домен (пока raw по умолчанию).
self.background = self.background_raw_capture
self.background_source_type = "capture_raw"
self.background_reference_summary = summary
self.last_reference_error = ""
return True
except Exception as exc:
self.last_reference_error = f"Ошибка загрузки фона: {exc}"
return False
# Совместимые обертки для старого API (строго .npy)
def load_calib_envelope(self, path: Optional[str] = None) -> bool:
target = path if path is not None else self.calib_envelope_path
return self.load_calib_reference(target, source_kind="npy")
def load_background(self, path: Optional[str] = None) -> bool:
target = path if path is not None else self.background_path
return self.load_background_reference(target, source_kind="npy")
# ---- Нормировка / фон ----
def _normalize_against_active_reference(self, raw: np.ndarray) -> Tuple[Optional[np.ndarray], bool, str]:
if not self.calib_enabled:
return None, False, "off"
if self.calib_mode == "file":
if self.calib_file_envelope is None:
return None, False, "off"
src = "capture" if self.calib_external_source_type == "capture" else "npy"
return normalize_by_envelope(raw, self.calib_file_envelope), True, src
if self.last_calib_sweep is None:
return None, False, "off"
return normalize_by_calib(raw, self.last_calib_sweep, self.norm_type), True, "live"
def _transform_raw_background_for_current_domain(self, calib_applied: bool) -> Optional[np.ndarray]:
if self.background_raw_capture is None:
return None
if not calib_applied:
return self.background_raw_capture
# Порядок pipeline фиксирован: raw -> calibration -> background -> IFFT.
# Поэтому raw-фон из capture нужно привести в тот же домен, что и текущий sweep_for_ring.
if self.calib_mode == "file" and self.calib_file_envelope is not None:
return normalize_by_envelope(self.background_raw_capture, self.calib_file_envelope)
if self.calib_mode == "live" and self.last_calib_sweep is not None:
return normalize_by_calib(self.background_raw_capture, self.last_calib_sweep, self.norm_type)
return None
def _effective_background(self, calib_applied: bool) -> Tuple[Optional[np.ndarray], str]:
if self.background_source_type == "capture_raw":
bg = self._transform_raw_background_for_current_domain(calib_applied)
if bg is None:
return None, "capture(raw->calib:missing-calib)"
self.background = np.asarray(bg, dtype=np.float32)
return self.background, ("capture(raw->calib)" if calib_applied else "capture(raw)")
if self.background_source_type == "npy_processed" and self.background is not None:
return self.background, "npy"
if self.background is not None:
return self.background, "unknown"
return None, "off"
def _subtract_background(self, sweep: np.ndarray, calib_applied: bool) -> Tuple[np.ndarray, bool, str]:
if not self.background_enabled:
return sweep, False, "off"
bg, bg_src = self._effective_background(calib_applied)
if bg is None:
return sweep, False, f"{bg_src}:missing"
out = np.asarray(sweep, dtype=np.float32).copy()
w = min(out.size, bg.size)
if w > 0:
out[:w] -= bg[:w]
return out, True, bg_src
def process(self, sweep: np.ndarray, channel: int, update_references: bool = True) -> SweepProcessingResult:
"""Применить к свипу калибровку/фон и вернуть явные этапы обработки."""
raw = np.asarray(sweep, dtype=np.float32)
ch = int(channel)
if ch == 0:
if update_references:
self.last_calib_sweep = raw
if self.auto_save_live_calib_envelope:
self.save_calib_envelope()
# ch0 всегда остаётся live-калибровочной ссылкой (raw), но при file-калибровке
# можем применять её и к ch0 для отображения/обработки независимо от канала.
calib_applied = False
calib_source = "off"
normalized: Optional[np.ndarray] = None
if self.calib_enabled and self.calib_mode == "file":
normalized, calib_applied, calib_source = self._normalize_against_active_reference(raw)
base = normalized if normalized is not None else raw
processed, bg_applied, bg_source = self._subtract_background(base, calib_applied=calib_applied)
stages = ["parsed_sweep", "raw_sweep", "ch0_live_calibration_reference"]
stages.append(f"calibration_{calib_source}" if calib_applied else "calibration_off")
stages.append(f"background_{bg_source}" if bg_applied else "background_off")
stages.extend(["ring_buffer", "ifft_db"])
return SweepProcessingResult(
processed_sweep=processed,
normalized_sweep=normalized,
calibration_applied=calib_applied,
background_applied=bg_applied,
calibration_source=calib_source if calib_applied else "off",
background_source=bg_source if bg_applied else "off",
is_calibration_reference=True,
stage_trace=tuple(stages),
)
normalized, calib_applied, calib_source = self._normalize_against_active_reference(raw)
base = normalized if normalized is not None else raw
processed, bg_applied, bg_source = self._subtract_background(base, calib_applied)
stages = ["parsed_sweep", "raw_sweep"]
stages.append(f"calibration_{calib_source}" if calib_applied else "calibration_off")
stages.append(f"background_{bg_source}" if bg_applied else "background_off")
stages.extend(["ring_buffer", "ifft_db"])
return SweepProcessingResult(
processed_sweep=processed,
normalized_sweep=normalized,
calibration_applied=calib_applied,
background_applied=bg_applied,
calibration_source=calib_source if calib_applied else "off",
background_source=bg_source if bg_applied else "off",
is_calibration_reference=False,
stage_trace=tuple(stages),
)