mirror of https://github.com/NaN-tic/nanscan.git
Show mouse position & findMatchingTemplateByText:
- Show current mouse position in template designer. - New 'findMatchingTemplateByText()' still not working properly but most code is already there including the action to test it from template designer.
This commit is contained in:
parent
c0e881db5d
commit
9cc4af299d
|
@ -24,6 +24,8 @@ from ocr import *
|
|||
from template import *
|
||||
from document import *
|
||||
from trigram import *
|
||||
from hamming import *
|
||||
from translator import *
|
||||
|
||||
import tempfile
|
||||
|
||||
|
@ -42,6 +44,7 @@ class Recognizer(QObject):
|
|||
QObject.__init__(self, parent)
|
||||
self.barcode = Barcode()
|
||||
self.ocr = Ocr()
|
||||
self.image = None
|
||||
|
||||
## @brief Returns the text of a given region of the image.
|
||||
def textInRegion(self, region, type=None):
|
||||
|
@ -145,7 +148,7 @@ class Recognizer(QObject):
|
|||
#
|
||||
# TODO: Using offsets to find the best template is easy but highly inefficient.
|
||||
# a smarter solution should be implemented.
|
||||
def findMatchingTemplate( self, templates, offset = 5 ):
|
||||
def findMatchingTemplateByOffset( self, templates, offset = 5 ):
|
||||
max = 0
|
||||
best = {
|
||||
'template': None,
|
||||
|
@ -165,7 +168,7 @@ class Recognizer(QObject):
|
|||
for documentBox in currentDocument.boxes:
|
||||
templateBox = documentBox.templateBox
|
||||
if documentBox.templateBox.type != 'matcher':
|
||||
print "Jumping %s due to type %s " % ( templateBox.name, templateBox.type )
|
||||
print "Jumping %s due to type %s" % ( templateBox.name, templateBox.type )
|
||||
continue
|
||||
matcherBoxes += 1
|
||||
similarity = Trigram.trigram( documentBox.text, templateBox.text )
|
||||
|
@ -182,6 +185,57 @@ class Recognizer(QObject):
|
|||
print "Template %s has score %s with offset (%s,%s)" % (template.name, score, xOffset, yOffset)
|
||||
return best
|
||||
|
||||
|
||||
## @brief Tries to find out the best template in 'templates' for the current
|
||||
# image.
|
||||
# This algorithm starts by looking for template boxes of type 'matching' in the
|
||||
# text and then looks if the relative positions of the new document and template
|
||||
# boxes are similar. This is intended to be faster than exhaustive algorithm used
|
||||
# in findMatchingTemplateByOffset().
|
||||
#
|
||||
# Note that the image must have been scanned (using scan() or startScan())
|
||||
# before using this function.
|
||||
#
|
||||
# TODO: Using offsets to find the best template is easy but highly inefficient.
|
||||
# a smarter solution should be implemented.
|
||||
def findMatchingTemplateByText( self, templates ):
|
||||
max = 0
|
||||
best = {
|
||||
'template': None,
|
||||
'document': Document(),
|
||||
'xOffset' : 0,
|
||||
'yOffset' : 0
|
||||
}
|
||||
for template in templates:
|
||||
if not template.boxes:
|
||||
continue
|
||||
# Find out template's offset
|
||||
offset = self.findTemplateOffset( template )
|
||||
if not offset:
|
||||
continue
|
||||
|
||||
score = 0
|
||||
matcherBoxes = 0
|
||||
# Apply template with offset found
|
||||
currentDocument = self.extractWithTemplate( template, offset.x(), offset.y() )
|
||||
for documentBox in currentDocument.boxes:
|
||||
templateBox = documentBox.templateBox
|
||||
if documentBox.templateBox.type != 'matcher':
|
||||
continue
|
||||
matcherBoxes += 1
|
||||
similarity = Trigram.trigram( documentBox.text, templateBox.text )
|
||||
score += similarity
|
||||
score = score / matcherBoxes
|
||||
if score > max:
|
||||
max = score
|
||||
best = {
|
||||
'template': template,
|
||||
'document': currentDocument,
|
||||
'xOffset' : offset.x(),
|
||||
'yOffset' : offset.y()
|
||||
}
|
||||
return best
|
||||
|
||||
## @brief Returns a QPoint with the offset that needs to be applied to the given
|
||||
# template to best fit the current image.
|
||||
def findTemplateOffset( self, template ):
|
||||
|
@ -190,12 +244,161 @@ class Recognizer(QObject):
|
|||
|
||||
lines = self.ocr.textLinesWithSpaces()
|
||||
|
||||
# Create a default translator only once
|
||||
translator = Translator()
|
||||
|
||||
# This list will keep a pointer to each template box of type 'matcher'
|
||||
matchers = []
|
||||
for templateBox in template.boxes:
|
||||
if templateBox.type != 'matcher':
|
||||
continue
|
||||
|
||||
for line in lines:
|
||||
templateBox.text
|
||||
templateBox.ranges = Range.extractAllRangesFromDocument(lines, len(templateBox.text))
|
||||
for ran in templateBox.ranges:
|
||||
text = ran.text()
|
||||
value = Hamming.hamming( text, templateBox.text, translator )
|
||||
ran.distance = value
|
||||
templateBox.ranges.sort( rangeDistanceComparison )
|
||||
|
||||
|
||||
if templateBox.ranges:
|
||||
bestRange = templateBox.ranges[0]
|
||||
print "The best match for template box '%s' is '%s'" % (templateBox.text, bestRange.text() )
|
||||
matchers.append( templateBox )
|
||||
|
||||
# Once we have all ranges sorted for each template box we search which
|
||||
# range combination matches the template.
|
||||
iterator = TemplateBoxRangeIterator( matchers )
|
||||
i = 0
|
||||
for ranges in iterator:
|
||||
documentBoxCenter = ranges[0].rect().center()
|
||||
templateBoxCenter = matchers[0].featureRect.center()
|
||||
diff = templateBoxCenter - documentBoxCenter
|
||||
print "Difference: ", diff
|
||||
found = True
|
||||
for pos in range(1,len(ranges)):
|
||||
documentBoxCenter = ranges[pos].rect().center()
|
||||
templateBoxCenter = matchers[pos].featureRect.center()
|
||||
d = templateBoxCenter - documentBoxCenter
|
||||
# If difference of relative positions of boxes between
|
||||
# template and document are bigger than 5mm we discard
|
||||
# the ranges
|
||||
print "Difference in loop: ", d
|
||||
if ( abs(d.x()) + 5.0 > abs(diff.x()) ):
|
||||
found = False
|
||||
break
|
||||
if ( abs(d.y()) + 5.0 > abs(diff.y()) ):
|
||||
found = False
|
||||
break
|
||||
if found:
|
||||
break
|
||||
i += 1
|
||||
if i > 10:
|
||||
break
|
||||
if found:
|
||||
return diff
|
||||
else:
|
||||
return None
|
||||
|
||||
|
||||
class TemplateBoxRangeIterator:
|
||||
def __init__(self, boxes):
|
||||
self.boxes = boxes
|
||||
self.pos = [0] * len(self.boxes)
|
||||
self.loopPos = [0] * len(self.boxes)
|
||||
self.added = None
|
||||
|
||||
def __iter__(self):
|
||||
return self
|
||||
|
||||
def next(self):
|
||||
result = []
|
||||
for x in range(len(self.boxes)):
|
||||
result.append( self.boxes[x].ranges[ self.pos[x] ] )
|
||||
|
||||
print '----'
|
||||
print (u', '.join( [x.text() for x in result] )).encode('ascii', 'ignore')
|
||||
print self.pos
|
||||
print self.loopPos
|
||||
if self.loopPos == self.pos:
|
||||
# Search next value to add
|
||||
value = float('infinity')
|
||||
pos = 0
|
||||
for x in range(len(self.pos)):
|
||||
if x >= len(self.boxes[x].ranges) - 1:
|
||||
continue
|
||||
if self.pos[x] < value:
|
||||
value = self.pos[x]
|
||||
self.added = x
|
||||
# If value is Infinity it means that we reached the end
|
||||
# of all possible iterations
|
||||
if value == float('infinity'):
|
||||
raise StopIteration
|
||||
|
||||
self.pos[self.added] += 1
|
||||
self.loopPos = [0] * len(self.boxes)
|
||||
self.loopPos[self.added] = self.pos[self.added]
|
||||
else:
|
||||
for x in range(len(self.loopPos)):
|
||||
if x == self.added:
|
||||
continue
|
||||
if self.loopPos[x] < self.pos[x]:
|
||||
self.loopPos[x] += 1
|
||||
break
|
||||
return result
|
||||
|
||||
def rangeDistanceComparison(x, y):
|
||||
if x.distance > y.distance:
|
||||
return 1
|
||||
elif x.distance < y.distance:
|
||||
return -1
|
||||
else:
|
||||
return 0
|
||||
|
||||
## @brief This class represents a group of characters in a document.
|
||||
class Range:
|
||||
def __init__(self):
|
||||
self.line = 0
|
||||
self.pos = 0
|
||||
self.length = 0
|
||||
self.document = None
|
||||
|
||||
## @brief Returns a unicode string with the text of the current range
|
||||
def text(self):
|
||||
line = self.document[self.line]
|
||||
chars = line[self.pos:self.pos + self.length]
|
||||
return u''.join( [x.character for x in chars] )
|
||||
|
||||
## @brief Returns the bounding rectangle of the text in the range
|
||||
def rect(self):
|
||||
line = self.document[self.line]
|
||||
chars = line[self.pos:self.pos + self.length]
|
||||
rect = QRectF()
|
||||
for c in chars:
|
||||
rect = rect.united( c.box )
|
||||
return rect
|
||||
|
||||
## @brief Returns a list with all possible ranges of size length of the
|
||||
# given document
|
||||
@staticmethod
|
||||
def extractAllRangesFromDocument(lines, length):
|
||||
if length <= 0:
|
||||
return []
|
||||
ranges = []
|
||||
for line in range(len(lines)):
|
||||
if length > len(lines[line]):
|
||||
ran = Range()
|
||||
ran.line = line
|
||||
ran.pos = 0
|
||||
ran.length = len(lines[line])
|
||||
ran.document = lines
|
||||
ranges.append( ran )
|
||||
continue
|
||||
for pos in range(len(lines[line]) - length):
|
||||
ran = Range()
|
||||
ran.line = line
|
||||
ran.pos = pos
|
||||
ran.length = length
|
||||
ran.document = lines
|
||||
ranges.append( ran )
|
||||
return ranges
|
||||
|
||||
|
|
|
@ -119,6 +119,7 @@ class TemplateBoxItem(QGraphicsRectItem):
|
|||
self.templateBox = None
|
||||
if featureRect:
|
||||
self.feature = QGraphicsRectItem(featureRect, self)
|
||||
self.feature.setZValue( -1 )
|
||||
|
||||
class DocumentScene(QGraphicsScene):
|
||||
|
||||
|
@ -192,6 +193,9 @@ class DocumentScene(QGraphicsScene):
|
|||
self.setTemplateBoxesVisible( self._templateBoxesVisible )
|
||||
|
||||
def mapRectFromRecognizer(self, rect):
|
||||
# When there's no image loaded yet, return the same rect
|
||||
if not self.dotsPerMillimeterX:
|
||||
return rect
|
||||
box = QRectF()
|
||||
box.setX( rect.x() * self.dotsPerMillimeterX )
|
||||
box.setY( rect.y() * self.dotsPerMillimeterY )
|
||||
|
@ -200,6 +204,9 @@ class DocumentScene(QGraphicsScene):
|
|||
return box
|
||||
|
||||
def mapRectToRecognizer(self, rect):
|
||||
# When there's no image loaded yet, return the same rect
|
||||
if not self.dotsPerMillimeterX:
|
||||
return rect
|
||||
box = QRectF()
|
||||
box.setX( rect.x() / self.dotsPerMillimeterX )
|
||||
box.setY( rect.y() / self.dotsPerMillimeterY )
|
||||
|
@ -208,12 +215,18 @@ class DocumentScene(QGraphicsScene):
|
|||
return box
|
||||
|
||||
def mapPointFromRecognizer(self, point):
|
||||
# When there's no image loaded yet, return the same point
|
||||
if not self.dotsPerMillimeterX:
|
||||
return point
|
||||
d = QPointF()
|
||||
d.setX( point.x() * self.dotsPerMillimeterX )
|
||||
d.setY( point.y() * self.dotsPerMillimeterY )
|
||||
return d
|
||||
|
||||
def mapPointToRecognizer(self, point):
|
||||
# When there's no image loaded yet, return the same point
|
||||
if not self.dotsPerMillimeterX:
|
||||
return point
|
||||
d = QPointF()
|
||||
d.setX( point.x() / self.dotsPerMillimeterX )
|
||||
d.setY( point.y() / self.dotsPerMillimeterY )
|
||||
|
@ -343,11 +356,14 @@ class DocumentScene(QGraphicsScene):
|
|||
if not self.isEnabled():
|
||||
return
|
||||
if self._mode == self.CreationMode:
|
||||
item = self.itemAt( event.scenePos() )
|
||||
if item and unicode(item.data( 0 ).toString()) == 'TemplateBox':
|
||||
self.setActiveItem( item )
|
||||
return
|
||||
# Search if there's a TemplateBox where the user clicked
|
||||
# If so, make the item active.
|
||||
for item in self.items( event.scenePos() ):
|
||||
if unicode(item.data(0).toString()) == 'TemplateBox':
|
||||
self.setActiveItem( item )
|
||||
return
|
||||
|
||||
# Otherwise, create a new template box
|
||||
rect = QRectF(event.scenePos().x(), event.scenePos().y(), 1, 1)
|
||||
self._selection = self.createTemplateBox( rect )
|
||||
elif self._mode == self.SelectionMode:
|
||||
|
@ -371,6 +387,7 @@ class DocumentScene(QGraphicsScene):
|
|||
self.emit( SIGNAL('newTemplateBox(QRectF)'), self.mapRectToRecognizer( rect ) )
|
||||
|
||||
def mouseMoveEvent(self, event):
|
||||
self.emit( SIGNAL('mouseMoved'), event.scenePos() )
|
||||
if not self.isEnabled():
|
||||
return
|
||||
if self._mode == self.CreationMode and self._selection:
|
||||
|
@ -414,6 +431,7 @@ class MainWindow(QMainWindow):
|
|||
|
||||
self.connect( self.scene, SIGNAL('newTemplateBox(QRectF)'), self.newTemplateBox )
|
||||
self.connect( self.scene, SIGNAL('currentTemplateBoxChanged(PyQt_PyObject,PyQt_PyObject)'), self.currentTemplateBoxChanged)
|
||||
self.connect( self.scene, SIGNAL('mouseMoved'), self.sceneMouseMoved )
|
||||
self.connect( self.actionExit, SIGNAL('triggered()'), self.close )
|
||||
self.connect( self.actionOpenImage, SIGNAL('triggered()'), self.openImage )
|
||||
self.connect( self.actionOpenTemplate, SIGNAL('triggered()'), self.openTemplate )
|
||||
|
@ -428,7 +446,8 @@ class MainWindow(QMainWindow):
|
|||
self.connect( self.actionDelete, SIGNAL('triggered()'), self.removeTemplateBox )
|
||||
self.connect( self.actionZoom, SIGNAL('triggered()'), self.zoom )
|
||||
self.connect( self.actionUnzoom, SIGNAL('triggered()'), self.unzoom )
|
||||
self.connect( self.actionFindMatchingTemplate, SIGNAL('triggered()'), self.findMatchingTemplate )
|
||||
self.connect( self.actionFindMatchingTemplateByOffset, SIGNAL('triggered()'), self.findMatchingTemplateByOffset )
|
||||
self.connect( self.actionFindMatchingTemplateByText, SIGNAL('triggered()'), self.findMatchingTemplateByText )
|
||||
self.toggleImageBoxes()
|
||||
QTimer.singleShot( 1000, self.setup )
|
||||
self.updateTitle()
|
||||
|
@ -446,17 +465,44 @@ class MainWindow(QMainWindow):
|
|||
self.uiToolDock.setWidget( self.uiTool )
|
||||
|
||||
#rpc.session.login( 'http://admin:admin@127.0.0.1:8069', 'g1' )
|
||||
def sceneMouseMoved(self, pos):
|
||||
self.updatePosition( pos )
|
||||
|
||||
def findMatchingTemplateByOffset(self):
|
||||
self.findMatchingTemplate( 'offset' )
|
||||
|
||||
def findMatchingTemplateByText(self):
|
||||
self.findMatchingTemplate( 'text' )
|
||||
|
||||
def findMatchingTemplate(self, type):
|
||||
if type == 'offset':
|
||||
title = _('Template search by offset')
|
||||
else:
|
||||
title = _('Template search by text')
|
||||
|
||||
if not self.recognizer.image:
|
||||
QMessageBox.information( self, title, _('No image opened. Please open an image to find a matching template.') )
|
||||
return
|
||||
|
||||
def findMatchingTemplate(self):
|
||||
if not rpc.session.logged():
|
||||
if not self.login():
|
||||
return
|
||||
templates = TemplateStorageManager.loadAll()
|
||||
result = self.recognizer.findMatchingTemplate( templates )
|
||||
time = QTime()
|
||||
time.start()
|
||||
if type == 'offset':
|
||||
result = self.recognizer.findMatchingTemplateByOffset( templates )
|
||||
else:
|
||||
result = self.recognizer.findMatchingTemplateByText( templates )
|
||||
elapsed = time.elapsed()
|
||||
if not result['template']:
|
||||
QMessageBox.information( self, title, _('No template found for the current image. Took %d milliseconds') % elapsed )
|
||||
return
|
||||
self._template = result['template']
|
||||
self.scene.setTemplate(self._template)
|
||||
self.updateTitle()
|
||||
QMessageBox.information( self, _('Template parameters'), _('Template found with offset (%d, %d)') % (result['xOffset'], result['yOffset']) )
|
||||
QMessageBox.information( self, title, _('Template found with offset (%d, %d) in %d milliseconds') % (result['xOffset'], result['yOffset'], elapsed) )
|
||||
|
||||
|
||||
def recognizerChanged(self, recognizer):
|
||||
rect = self.uiTool.box.rect
|
||||
|
@ -631,7 +677,13 @@ class MainWindow(QMainWindow):
|
|||
server = _('Press %s to login') % shortcut
|
||||
else:
|
||||
server = 'not logged in'
|
||||
self.statusBar().showMessage( server )
|
||||
self.uiServer.setText( server )
|
||||
|
||||
def updatePosition(self, pos):
|
||||
pos = self.uiView.mapToScene( self.uiView.mapFromGlobal( QCursor.pos() ) )
|
||||
pos = self.scene.mapPointToRecognizer( pos )
|
||||
position = _('(%.2f, %.2f)') % (pos.x(), pos.y())
|
||||
self.uiPosition.setText( position )
|
||||
|
||||
def updateActions(self):
|
||||
# Allow deleting if there's a TemplateBox selected
|
||||
|
|
|
@ -18,10 +18,10 @@
|
|||
<x>0</x>
|
||||
<y>48</y>
|
||||
<width>709</width>
|
||||
<height>417</height>
|
||||
<height>439</height>
|
||||
</rect>
|
||||
</property>
|
||||
<layout class="QHBoxLayout" >
|
||||
<layout class="QHBoxLayout" name="horizontalLayout_2" >
|
||||
<item>
|
||||
<layout class="QVBoxLayout" >
|
||||
<item>
|
||||
|
@ -41,7 +41,7 @@
|
|||
<x>0</x>
|
||||
<y>20</y>
|
||||
<width>70</width>
|
||||
<height>377</height>
|
||||
<height>399</height>
|
||||
</rect>
|
||||
</property>
|
||||
</widget>
|
||||
|
@ -50,7 +50,58 @@
|
|||
</layout>
|
||||
</item>
|
||||
<item>
|
||||
<widget class="QGraphicsView" name="uiView" />
|
||||
<layout class="QVBoxLayout" name="verticalLayout" >
|
||||
<item>
|
||||
<widget class="QGraphicsView" name="uiView" />
|
||||
</item>
|
||||
<item>
|
||||
<widget class="QFrame" name="frame" >
|
||||
<property name="sizePolicy" >
|
||||
<sizepolicy vsizetype="Fixed" hsizetype="Expanding" >
|
||||
<horstretch>0</horstretch>
|
||||
<verstretch>0</verstretch>
|
||||
</sizepolicy>
|
||||
</property>
|
||||
<property name="frameShape" >
|
||||
<enum>QFrame::VLine</enum>
|
||||
</property>
|
||||
<property name="frameShadow" >
|
||||
<enum>QFrame::Raised</enum>
|
||||
</property>
|
||||
<layout class="QHBoxLayout" name="horizontalLayout" >
|
||||
<property name="margin" >
|
||||
<number>0</number>
|
||||
</property>
|
||||
<item>
|
||||
<widget class="QLabel" name="uiServer" >
|
||||
<property name="sizePolicy" >
|
||||
<sizepolicy vsizetype="Fixed" hsizetype="Preferred" >
|
||||
<horstretch>0</horstretch>
|
||||
<verstretch>0</verstretch>
|
||||
</sizepolicy>
|
||||
</property>
|
||||
<property name="text" >
|
||||
<string>TextLabel</string>
|
||||
</property>
|
||||
</widget>
|
||||
</item>
|
||||
<item>
|
||||
<widget class="QLabel" name="uiPosition" >
|
||||
<property name="sizePolicy" >
|
||||
<sizepolicy vsizetype="Fixed" hsizetype="Preferred" >
|
||||
<horstretch>0</horstretch>
|
||||
<verstretch>0</verstretch>
|
||||
</sizepolicy>
|
||||
</property>
|
||||
<property name="text" >
|
||||
<string>(0, 0)</string>
|
||||
</property>
|
||||
</widget>
|
||||
</item>
|
||||
</layout>
|
||||
</widget>
|
||||
</item>
|
||||
</layout>
|
||||
</item>
|
||||
</layout>
|
||||
</widget>
|
||||
|
@ -96,23 +147,14 @@
|
|||
<property name="title" >
|
||||
<string>&Actions</string>
|
||||
</property>
|
||||
<addaction name="actionFindMatchingTemplate" />
|
||||
<addaction name="actionFindMatchingTemplateByOffset" />
|
||||
<addaction name="actionFindMatchingTemplateByText" />
|
||||
</widget>
|
||||
<addaction name="menuFile" />
|
||||
<addaction name="menuEdit" />
|
||||
<addaction name="menu_Actions" />
|
||||
<addaction name="menuView" />
|
||||
</widget>
|
||||
<widget class="QStatusBar" name="statusbar" >
|
||||
<property name="geometry" >
|
||||
<rect>
|
||||
<x>0</x>
|
||||
<y>465</y>
|
||||
<width>709</width>
|
||||
<height>22</height>
|
||||
</rect>
|
||||
</property>
|
||||
</widget>
|
||||
<widget class="QToolBar" name="toolBar" >
|
||||
<property name="geometry" >
|
||||
<rect>
|
||||
|
@ -239,9 +281,14 @@
|
|||
<string>Show Feature Boxes</string>
|
||||
</property>
|
||||
</action>
|
||||
<action name="actionFindMatchingTemplate" >
|
||||
<action name="actionFindMatchingTemplateByOffset" >
|
||||
<property name="text" >
|
||||
<string>Find Matching Template</string>
|
||||
<string>Find Matching Template by &Offset</string>
|
||||
</property>
|
||||
</action>
|
||||
<action name="actionFindMatchingTemplateByText" >
|
||||
<property name="text" >
|
||||
<string>Find Matching Template by &Text</string>
|
||||
</property>
|
||||
</action>
|
||||
</widget>
|
||||
|
|
Loading…
Reference in New Issue