dockerfile/openalpr-base
1
0
Fork 0
mirror of https://github.com/kerberos-io/openalpr-base.git synced 2025-10-06 10:26:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

Increased recognition speed for plate region analysis by 2.5x.

Browse Source 
Reduced number of thresholded images from 10 to 4.  Accuracy (based on current benchmarks) is about the same, but recognition speed for a single plate region goes from ~88ms to ~40ms.
This commit is contained in:
Matt Hill
2014-02-18 22:49:25 -06:00
parent 646c2d6660
commit bcdb010cc5
1 changed files with 28 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

63
src/openalpr/utility.cpp
View File

@@ -112,7 +112,7 @@ void displayImage(Config* config, string windowName, cv::Mat frame)
vector<Mat> produceThresholds(const Mat img_gray, Config* config)
{
const int THRESHOLD_COUNT = 10;
const int THRESHOLD_COUNT = 4;
//Mat img_equalized = equalizeBrightness(img_gray);
timespec startTime;
@@ -120,44 +120,37 @@ vector<Mat> produceThresholds(const Mat img_gray, Config* config)
vector<Mat> thresholds;
//#pragma omp parallel for
for (int i = 0; i < THRESHOLD_COUNT; i++)
thresholds.push_back(Mat(img_gray.size(), CV_8U));
int i = 0;
for (int i = 0; i < THRESHOLD_COUNT; i++)
{
if (i <= 2) //0-2
{
int k = ((i%3) * 5) + 7; // 7, 12, 17
if (k==12) k = 13; // change 12 to 13
//#pragma omp ordered
adaptiveThreshold(img_gray, thresholds[i], 255, ADAPTIVE_THRESH_MEAN_C, THRESH_BINARY_INV , k, 3);
}
else if (i <= 6) //3-6
{
int k = i%2; // 0 or 1
int win = 18 + (k * 4); // 18 or 22
//#pragma omp ordered
NiblackSauvolaWolfJolion (img_gray, thresholds[i], WOLFJOLION, win, win, 0.05 + (k * 0.35));
bitwise_not(thresholds[i], thresholds[i]);
}
else if (i <= 9) //7-9
{
int k = (i%3) + 1; // 1,2,3
//#pragma omp ordered
NiblackSauvolaWolfJolion (img_gray, thresholds[i], SAUVOLA, 12, 12, 0.18 * k);
bitwise_not(thresholds[i], thresholds[i]);
}
}
// Adaptive
//adaptiveThreshold(img_gray, thresholds[i++], 255, ADAPTIVE_THRESH_MEAN_C, THRESH_BINARY_INV , 7, 3);
//adaptiveThreshold(img_gray, thresholds[i++], 255, ADAPTIVE_THRESH_MEAN_C, THRESH_BINARY_INV , 13, 3);
//adaptiveThreshold(img_gray, thresholds[i++], 255, ADAPTIVE_THRESH_MEAN_C, THRESH_BINARY_INV , 17, 3);
// Wolf
int k = 0, win=18;
//NiblackSauvolaWolfJolion (img_gray, thresholds[i++], WOLFJOLION, win, win, 0.05 + (k * 0.35));
//bitwise_not(thresholds[i-1], thresholds[i-1]);
NiblackSauvolaWolfJolion (img_gray, thresholds[i++], WOLFJOLION, win, win, 0.05 + (k * 0.35));
bitwise_not(thresholds[i-1], thresholds[i-1]);
k = 1; win = 22;
NiblackSauvolaWolfJolion (img_gray, thresholds[i++], WOLFJOLION, win, win, 0.05 + (k * 0.35));
bitwise_not(thresholds[i-1], thresholds[i-1]);
//NiblackSauvolaWolfJolion (img_gray, thresholds[i++], WOLFJOLION, win, win, 0.05 + (k * 0.35));
//bitwise_not(thresholds[i-1], thresholds[i-1]);
// Sauvola
k = 1;
NiblackSauvolaWolfJolion (img_gray, thresholds[i++], SAUVOLA, 12, 12, 0.18 * k);
bitwise_not(thresholds[i-1], thresholds[i-1]);
k=2;
NiblackSauvolaWolfJolion (img_gray, thresholds[i++], SAUVOLA, 12, 12, 0.18 * k);
bitwise_not(thresholds[i-1], thresholds[i-1]);