Files
openalpr-base/src/openalpr/detection/detectorcuda.cpp
2015-03-18 19:00:30 -04:00

144 lines
4.2 KiB
C++

/*
* Copyright (c) 2013 New Designs Unlimited, LLC
* Opensource Automated License Plate Recognition [http://www.openalpr.com]
*
* This file is part of OpenAlpr.
*
* OpenAlpr is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License
* version 3 as published by the Free Software Foundation
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "detectorcuda.h"
#ifdef COMPILE_GPU
using namespace cv;
using namespace std;
namespace alpr
{
DetectorCUDA::DetectorCUDA(Config* config) : Detector(config) {
if( this->cuda_cascade.load( config->getCascadeRuntimeDir() + config->country + ".xml" ) )
{
this->loaded = true;
printf("--(!)Loaded CUDA classifier\n");
}
else
{
this->loaded = false;
printf("--(!)Error loading CUDA classifier\n");
}
}
DetectorCUDA::~DetectorCUDA() {
}
vector<PlateRegion> DetectorCUDA::detect(Mat frame, std::vector<cv::Rect> regionsOfInterest)
{
Mat frame_gray;
cvtColor( frame, frame_gray, CV_BGR2GRAY );
vector<PlateRegion> detectedRegions;
for (int i = 0; i < regionsOfInterest.size(); i++)
{
Mat cropped = frame_gray(regionsOfInterest[i]);
vector<PlateRegion> subRegions = doCascade(cropped, regionsOfInterest[i].x, regionsOfInterest[i].y);
for (int j = 0; j < subRegions.size(); j++)
detectedRegions.push_back(subRegions[j]);
}
return detectedRegions;
}
vector<PlateRegion> DetectorCUDA::doCascade(Mat frame, int offset_x, int offset_y)
{
if (frame.cols > config->maxDetectionInputWidth)
{
// The frame is too wide
this->scale_factor = ((float) config->maxDetectionInputWidth) / ((float) frame.cols);
if (config->debugDetector)
std::cout << "Input detection image is too wide. Resizing with scale: " << this->scale_factor << endl;
}
else if (frame.rows > config->maxDetectionInputHeight)
{
// The frame is too tall
this->scale_factor = ((float) config->maxDetectionInputHeight) / ((float) frame.rows);
if (config->debugDetector)
std::cout << "Input detection image is too tall. Resizing with scale: " << this->scale_factor << endl;
}
int w = frame.size().width;
int h = frame.size().height;
vector<Rect> plates;
equalizeHist( frame, frame );
resize(frame, frame, Size(w * this->scale_factor, h * this->scale_factor));
//-- Detect plates
timespec startTime;
getTimeMonotonic(&startTime);
float maxWidth = ((float) w) * (config->maxPlateWidthPercent / 100.0f) * this->scale_factor;
float maxHeight = ((float) h) * (config->maxPlateHeightPercent / 100.0f) * this->scale_factor;
Size minSize(config->minPlateSizeWidthPx * this->scale_factor, config->minPlateSizeHeightPx * this->scale_factor);
gpu::GpuMat cudaFrame, plateregions_buffer;
Mat plateregions_downloaded;
cudaFrame.upload(frame);
int numdetected = cuda_cascade.detectMultiScale(cudaFrame, plateregions_buffer, (double) config->detection_iteration_increase, config->detectionStrictness, minSize);
plateregions_buffer.colRange(0, numdetected).download(plateregions_downloaded);
for (int i = 0; i < numdetected; ++i)
{
plates.push_back(plateregions_downloaded.ptr<cv::Rect>()[i]);
}
if (config->debugTiming)
{
timespec endTime;
getTimeMonotonic(&endTime);
cout << "LBP Time: " << diffclock(startTime, endTime) << "ms." << endl;
}
for( unsigned int i = 0; i < plates.size(); i++ )
{
plates[i].x = (plates[i].x / scale_factor) + offset_x;
plates[i].y = (plates[i].y / scale_factor) + offset_y;
plates[i].width = plates[i].width / scale_factor;
plates[i].height = plates[i].height / scale_factor;
}
vector<PlateRegion> orderedRegions = aggregateRegions(plates);
return orderedRegions;
}
}
#endif