Minos/CVBpy/QmlMinos

This example program is located in your CVB installation under %CVB%Tutorial/Minos/CVBpy/QmlMinos.

main.py:

# @brief Example for OCR reading with Minos in QML environment.
 
import os, sys
 
import cvb
import cvb.ui
from minos_result_model import MinosResultModel
from minos_search import MinosSearch
 
if sys.version_info >= (3, 11):
    from PySide6.QtCore import QObject, QUrl, QAbstractListModel, Qt, QModelIndex
    from PySide6.QtQml import QQmlApplicationEngine, qmlRegisterType
    from PySide6.QtGui import QGuiApplication
    from PySide6.QtGui import QIcon
else:
    from PySide2.QtCore import QObject, QUrl, QAbstractListModel, Qt, QModelIndex
    from PySide2.QtQml import QQmlApplicationEngine, qmlRegisterType
    from PySide2.QtGui import QGuiApplication
    from PySide2.QtGui import QIcon
 
 
if __name__ == "__main__":
 
    app = QGuiApplication([])
    app.setOrganizationName('STEMMER IMAGING')
    app.setOrganizationDomain('https://www.stemmer-imaging.com/')
    app.setApplicationName('Minos Python tutorial')
    
    # tell Windows the correct AppUserModelID for this process (shows icon in the taskbar)
    if sys.platform == 'win32':
        import ctypes
        myappid = u'stemmerimaging.commonvisionblox.pyminos.0'
        ctypes.windll.shell32.SetCurrentProcessExplicitAppUserModelID(myappid)
  
    app.setWindowIcon(QIcon('Tutorial-Python_32x32.png'))
 
    # open device
    with cvb.DeviceFactory.open(
            os.path.join(cvb.install_path(), "drivers", "CVMock.vin"),
            cvb.AcquisitionStack.Vin) as device:
        
        # setup QML interface objects
        image_controller = cvb.ui.ImageController()
        minos_result_model = MinosResultModel(image_controller)
 
        # main search object
        minos_search = MinosSearch(device.stream(), minos_result_model)
 
        # register QML components for an image display
        cvb.ui.ImageViewItem.register()
        cvb.ui.ImageLabelItem.register()
 
        engine = QQmlApplicationEngine()
        context = engine.rootContext()
 
        # create a controller object to communicate with QML
        context.setContextProperty("mainImage", image_controller)
        # create a Minos result model to communicate with QML for overlays
        context.setContextProperty("minosResultModel", minos_result_model)
        # create a Minos search object to communicate with QML (grab, snap)
        context.setContextProperty("minosSearch", minos_search)
 
        # load main QML file
        engine.load(os.path.join(os.path.dirname(os.path.abspath(__file__)), "main.qml"))
 
        # do a first snap at startup
        minos_search.snap()
 
        app.exec_()

minos_result_model.py:

import os, sys
 
import cvb
import cvb.ui
 
if sys.version_info >= (3, 11):
    from PySide6.QtCore import QObject, QAbstractListModel, Qt, QModelIndex, Property, Signal, Slot
else:
    from PySide2.QtCore import QObject, QAbstractListModel, Qt, QModelIndex, Property, Signal, Slot
 
class MinosResultData(object):
    
    def __init__(self):
        self.image = None
        self.list = []
 
class MinosResultModel(QAbstractListModel):
 
    LineText = Qt.UserRole
    StartPosition = Qt.UserRole + 1
 
    def __init__(self, image_controller , parent=None):
        super(MinosResultModel, self).__init__(parent)
        self._image_controller = image_controller
        self._result_queue = [None] * 0
        self._current_result = MinosResultData()
        self._text = ""
        self._processing_time = 0.0
 
        self.notify_refresh.connect(self.refresh)
 
    def roleNames(self):
        roles = dict()
        roles[MinosResultModel.LineText] = b"lineText"
        roles[MinosResultModel.StartPosition] = b"startPosition"
        return roles
 
    def rowCount(self, parent = QModelIndex()):
        return len(self._current_result.list)
 
    def data(self, index, role = Qt.DisplayRole):
        if not index.isValid():
            return None
        position, text = self._current_result.list[index.row()]
        if role == MinosResultModel.LineText:
            return text
        elif role == MinosResultModel.StartPosition:
            return cvb.ui.cvb_to_qt_point(position)
        else:
            return None
 
    def push_data(self, data):
        self._result_queue.append(data)
        self.notify_refresh.emit()
 
    @Slot()
    def refresh(self):
        if (len(self._result_queue) == 0):
            return None
 
        
        self._current_result = self._result_queue.pop()
        self._result_queue.clear()
        # refresh the model
 
        self.layoutChanged.emit()
        # refresh the text area
        stream = ""
        for i in range(len(self._current_result.list)):
            stream += (self._current_result.list[i][1] + '\n')
        self.set_search_result_text(stream)
        # refresh the image
        self._image_controller.refresh(self._current_result.image)
 
    def set_search_result_text(self, text):
        if (self._text != text):
            self._text = text
            self.notify_search_result_text.emit()
 
    def set_processing_time(self, processing_time):
        if (self._processing_time != processing_time):
            self._processing_time = processing_time
            self.notify_processing_time.emit()
 
    def search_result_text(self):
        return self._text
 
    def processing_time(self):
        return self._processing_time
 
    notify_search_result_text = Signal()
    notify_processing_time = Signal()
    notify_refresh = Signal()
 
    searchResultText = Property(str, search_result_text, notify=notify_search_result_text)
    processingTime = Property(float, processing_time, notify=notify_processing_time)

minos_search.py:

import os, sys
 
import cvb
import cvb.minos
 
if sys.version_info >= (3, 11):
    from PySide6 import QtCore
    from PySide6.QtCore import QObject, Qt, Slot
else:
    from PySide2 import QtCore
    from PySide2.QtCore import QObject, Qt, Slot
 
from minos_result_model import MinosResultModel, MinosResultData
 
# Global params
FIRST_AREA_RADIUS = 8.0
LINE_RADIUS = 4
WORD_HEIGHT = 4
WORD_WIDTH = 4
LINE_STEP_RADIUS = 8
 
LINE_STEP = cvb.Point2D(0.0, 16.0)
OCR_AOI = cvb.Area2D(cvb.Point2D(-3.0, -3.0), cvb.Point2D(3.0, -3.0), cvb.Point2D(-3.0, 3.0))
 
 
class MinosSearch(QtCore.QObject):
 
    def __init__(self, stream, minos_result_model):
        super().__init__()
        self._stream = stream
        self._model = minos_result_model
        self._classifier = cvb.minos.Classifier(os.path.join(cvb.install_path(), "tutorial", "Minos", "Images", "OCR", "Training Set", "Numbers.clf"))
 
 
 
    @Slot()
    def snap(self):
        image, wait_status = self._stream.get_timed_snapshot(1000)
        if (wait_status != cvb.WaitStatus.Ok):
            return None
 
        # Minos search in snap
        result = MinosResultData()
        result.image = image
 
        stop_watch = cvb.StopWatch()
        stop_watch.start()
        result.list = self.search(image.planes[0])
        self._model.set_processing_time(stop_watch.time_span)
 
        self._model.push_data(result)
 
    # The Minos search
    def search(self, plane):
        width = plane.map().width
        height = plane.map().height
 
        # STEP 1 finding the top most line
        # create an area centered in horizontal direction
        # size of the area is FIRST_AREA_WIDTH = ImageHeight
        # the direction of the area is set to the bottom
        # create an area from these values
        search_aoi = cvb.Area2D(cvb.Point2D(width / 2 - FIRST_AREA_RADIUS, 0.0), cvb.Point2D(width / 2 + FIRST_AREA_RADIUS, 0.0), cvb.Point2D(width / 2 - FIRST_AREA_RADIUS, height - 1.0))
 
        # search for the first object
        search_result = self._classifier.search(plane, cvb.minos.SearchMode.FindFirst, search_aoi)
 
        
        if (search_result.quality < 0.1):
            return []
 
        # now we found the first line of the text
 
        # STEP 2 finding the beginning of a line
        # try to find the first char in the line
        # create an area centered around the first line
        # beginning at the left side of the image
        # size of the area is LINE_HEIGHT * dXPos
        # the direction of the area is set to the right
        # create an area from these values
        position = search_result.position
 
        line_aoi = cvb.Area2D(cvb.Point2D(0.0, position.y - LINE_RADIUS), cvb.Point2D(0.0, position.y + LINE_RADIUS),cvb.Point2D(position.x, position.y - LINE_RADIUS))
 
        # do a line by line search 
        result_list = [None] * 0
        self.search_line(plane, line_aoi, result_list)
        return result_list
 
 
    # Minos search per line
    def search_line(self, plane, line_aoi, result_list):
 
        # search for the first object
        search_result = self._classifier.search(plane, cvb.minos.SearchMode.FindFirst, line_aoi)
 
        if (search_result.quality == 0.0):
            return None
 
        # now we found the first char in the first line
 
        # STEP 3 read the string
        # save the x and y position of the line
        # try to read the first word
        # create an area left aligned the first char
        # size of WORD_WIDTH * WORD_HEIGHT
        # create an area from these values
        line_start = search_result.position
 
        read_aoi = cvb.Area2D(cvb.Point2D(line_start.x, line_start.y - WORD_HEIGHT / 2), cvb.Point2D(line_start.x + WORD_WIDTH / 2,line_start.y - WORD_HEIGHT / 2), cvb.Point2D(line_start.x, line_start.y + WORD_HEIGHT / 2))
 
        # read the word
        search_result = self._classifier.read(plane, read_aoi, OCR_AOI)
 
        # save the result
        stream = ""
        for i in range(len(search_result)):
            stream += search_result[i].name
 
        result = (line_start, stream)
        result_list.append(result)
 
        line_start += LINE_STEP
 
        # STEP 4 jump to the next line
        # try to read the first word
        # create an area left aligned the first char
        # size of WORD_WIDTH * WORD_HEIGHT
        # create an area from these values
        next_line_aoi = cvb.Area2D(cvb.Point2D(line_start.x - LINE_STEP_RADIUS, line_start.y - LINE_STEP_RADIUS), cvb.Point2D(line_start.x + LINE_STEP_RADIUS, line_start.y - LINE_STEP_RADIUS), cvb.Point2D(line_start.x - LINE_STEP_RADIUS, line_start.y + LINE_STEP_RADIUS))
 
        # next line (recursion should - no to deep)
        self.search_line(plane, next_line_aoi, result_list)