fastcv 教程2：图像识别直线
Fastcv使用说明
本文件 所在目录包含了高通下使用fastcv静态库以及sdk的例程，由于fastcv使用交叉编译以及静态库，所以问文件生成略大，大约8M，所以不用往开发板推送动态库。
Fastcv 包含fastcv.h 和libfastcv.a的静态库。
Api接口地址:
https://developer.qualcomm.com/docs/fastcv/api/modules.html
1 霍夫检测直线的原理：
YUV->Rgb888->单通道gray->高斯滤波->边缘检测->霍夫变换->结果
在fastcv下对应的函数是
//yuv—rgb888
fcvColorYCbCr422PseudoPlanarToRGB888u8 (ybuf,chrombuf,640,480,0,0,rgb,0);
int file_fd,gray_fd;
int write_size;
#if rgb
if ((file_fd = open("outimg.img",O_CREAT | O_RDWR | O_APPEND,777)) == -1)
{
printf("create a picture file error!\n");
}
write_size=write(file_fd,rgb,sizeof(rgb));
printf("write_size = %d\n",write_size);
close(file_fd);
#endif
//Rgb565ConvertBmp((char*)outimg,640,480, "test.bmp");
#if bmp
rgb888bmp24((char*)rgb,"bmp24.bmp",640,480);
#endif
memcpy(readBuff,rgb,640*480*3);
//这里将rgb 转换bgr
for(i = 0; i < (640*480); i++)
{
temp = *(readBuff + i*3);
*(readBuff + i*3) = *(readBuff + i*3 + 2);
*(readBuff + i*3 + 2) = temp;
}
// myColorRGB888ToGray (readBuff, 640, 480);
//灰度转化
fcvColorRGB888ToGrayu8(readBuff, 640, 480,0, grayimg,0);
memset(readBuff,0,sizeof(readBuff));
//高斯滤波
fcvFilterGaussian3x3u8 (grayimg, 640, 480,readBuff,0);
//边缘检测
fcvFilterCanny3x3u8(readBuff,640,480,cannyBuff,0,15);
//将生成的边缘检测保存成bmp 进行查看
RgbgrayConvertBmp((char *)cannyBuff,640,480,"canny.bmp");
fcvHoughLineu8(cannyBuff,640,480,0,0.9,100,&linebuf, lineout);
原图yuv格式：
边缘后：
结果：
~ # ./fastcv_server
FastCV version 1.7.1
read 614400
file canny.bmp open success
Image size=307200, file size=308278, width=640, height=480
RgbgrayConvertBmp over
check line 5
check line START: 0.000000 2.117236
check line end: 640.000000 2.158375
check line START: 1.954749 0.000000
check line end: 2.007480 480.000000
check line START: 96.000000 381.228149
check line end: 608.000000 291.381042
check line START: 32.000000 399.263794
check line end: 640.000000 290.721497
check line START: 32.000000 391.076508
check line end: 64.000000 386.329926
说明：在结果中，反馈出来5条线，前两条线是边缘检测会生成两条对比线，水平和垂直的各一条，红框中是水平的。正好对应坐标（0，2）---（640，2）和（1.9，0）--（2.0，480）
下图1-2 1’-2’ 1”-2”分别对应剩下的三条，受环境影响如果照片不够清晰，干扰太多就会出现这种情况，将一条线识别成多条线的可能
我们将fcvHoughLineu8(cannyBuff,640,480,0,0.9,100,&linebuf, lineout);种的0.9 变成0.1
fcvHoughLineu8(cannyBuff,640,480,0,0.1,100,&linebuf, lineout);后会出现7条直线
前两条依旧是对比线。
~ # ./fastcv_server
FastCV version 1.7.1
read 614400
file canny.bmp open success
Image size=307200, file size=308278, width=640, height=480
RgbgrayConvertBmp over
check line 7
check line START: 0.000000 2.117236
check line end: 640.000000 2.158375
check line START: 1.954749 0.000000
check line end: 2.007480 480.000000
check line START: 32.000000 400.639923
check line end: 640.000000 287.543457
check line START: 32.000000 392.301483
check line end: 512.000000 308.120605
check line START: 576.000000 302.354980
check line end: 640.000000 291.278809
check line START: 608.000000 299.494934
check line end: 640.000000 289.787903
check line START: 96.000000 378.592285
check line end: 128.000000 383.339355
