ImageRecognition/DoublePointerCountALGO分割算法.cpp

#define _CRT_SECURE_NO_WARNINGS
#include <iostream>
#include <fstream>
#include <string>
#include <math.h>
#include <cmath>
#include <opencv2/imgproc.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/highgui/highgui.hpp>
#include "opencv2/imgproc/types_c.h"
#include <onnxruntime_cxx_api.h>
#include "DoublePointerCountALGO.h"
#include <vector>
#include <algorithm>

using namespace cv;
using namespace std;
using namespace Ort;

void DoublePointerCountALGO::Init(bool isCuda)
{
	string model_path = "models/double_pointer_count.onnx";
	std::wstring widestr = std::wstring(model_path.begin(), model_path.end());
	sessionOptions.SetGraphOptimizationLevel(ORT_ENABLE_BASIC);
	ort_session = new Session(env, widestr.c_str(), sessionOptions);
	size_t numInputNodes = ort_session->GetInputCount();
	size_t numOutputNodes = ort_session->GetOutputCount();
	AllocatorWithDefaultOptions allocator;
	for (int i = 0; i < numInputNodes; i++)
	{
		input_names.push_back(ort_session->GetInputName(i, allocator));
		Ort::TypeInfo input_type_info = ort_session->GetInputTypeInfo(i);
		auto input_tensor_info = input_type_info.GetTensorTypeAndShapeInfo();
		auto input_dims = input_tensor_info.GetShape();
		input_node_dims.push_back(input_dims);
	}                                                                                                                                                             
	for (int i = 0; i < numOutputNodes; i++)
	{
		output_names.push_back(ort_session->GetOutputName(i, allocator));
		Ort::TypeInfo output_type_info = ort_session->GetOutputTypeInfo(i);
		auto output_tensor_info = output_type_info.GetTensorTypeAndShapeInfo();
		auto output_dims = output_tensor_info.GetShape();
		output_node_dims.push_back(output_dims);
	}
	this->inpHeight = input_node_dims[0][2];
	this->inpWidth = input_node_dims[0][3];
	this->outHeight = output_node_dims[0][2];
	this->outWidth = output_node_dims[0][3];
}

int DoublePointerCountALGO::detect(Mat& srcimg)
{
	vector<float> input_image_ = { 1, 3, 512, 512 };
	Mat dstimg;
	Size resize_size(input_image_[2], input_image_[3]);
	resize(srcimg, dstimg, resize_size, 0, 0, cv::INTER_LINEAR);
	int channels = dstimg.channels();
	input_image_.resize((this->inpWidth * this->inpHeight * dstimg.channels()));

	for (int c = 0; c < channels; c++)
	{
		for (int i = 0; i < this->inpHeight; i++)
		{
			for (int j = 0; j < this->inpWidth; j++)
			{
				float pix = dstimg.ptr<uchar>(i)[j * 3 + 2 - c];
				input_image_[(c * this->inpHeight * this->inpWidth + i * this->inpWidth + j)] = (pix / 255.0 - mean[c]) / stds[c];
			}
		}
	}

	array<int64_t, 4> input_shape_{ 1, 3, this->inpHeight, this->inpWidth };
	auto allocator_info = MemoryInfo::CreateCpu(OrtDeviceAllocator, OrtMemTypeCPU);
	Value input_tensor_ = Value::CreateTensor<float>(allocator_info, input_image_.data(), input_image_.size(), input_shape_.data(), input_shape_.size());

	// 开始推理
	vector<Value> ort_outputs = ort_session->Run(RunOptions{ nullptr }, &input_names[0], &input_tensor_, 1, output_names.data(), output_names.size());
	float* pred = ort_outputs[0].GetTensorMutableData<float>();
	Mat result(outHeight, outWidth, CV_32FC1, pred);
	result = 2 - result;
	result *= 255;
	result.convertTo(result, CV_8UC1);
	Mat binary;
	threshold(result, binary, 150, 255, THRESH_BINARY);//二值化阈值处理
	//形态学变换
	Mat Sobel_Y_thres;
	Mat element = cv::getStructuringElement(MORPH_RECT, Size(1, 1));
	morphologyEx(binary, Sobel_Y_thres, cv::MORPH_OPEN, element, Point(-1, -1), 2);
	Mat result_line_image = creat_line_image(Sobel_Y_thres, 256, 130);
	/*namedWindow("矩形", WINDOW_NORMAL);
	imshow("矩形", result_line_image);
	waitKey(0);*/
	return get_meter_reader(result_line_image);
	//return result;
}

Mat DoublePointerCountALGO::creat_line_image(const Mat& circle, int Radius, int RingStride)
{
	Mat rectangle;
	float theta;
	int rho;

	rectangle = Mat::zeros(Size(Radius * pi * 2, RingStride), CV_8UC1);
	int nl = rectangle.rows;   // number of lines					
	int nc = rectangle.cols * rectangle.channels();   // total number of elements per line
	for (int j = 0; j < nl; j++) {
		// get the address of row j
		uchar* data = rectangle.ptr<uchar>(j);
		for (int i = 0; i < nc; i++) 
		{
			theta = pi * 2.0 / LINE_WIDTH * float(i + 1);
			rho = (Radius - j - 1);  
			int position_y = (float)circle_center[0] + rho * (float)std::cos(theta) + 0.5;
			int position_x = (float)circle_center[1] - rho * (float)std::sin(theta) + 0.5;
			data[i] = circle.at<uchar>(position_y, position_x);	
		}
	}
	return rectangle;
}

//像素值提取及累加
int DoublePointerCountALGO::get_meter_reader(const Mat& image)
{
	Mat histogram = Mat::zeros(Size(256, 1), CV_8UC1);
	int rows = LINE_HEIGH;  	 //输入图像的行数
	int cols = LINE_WIDTH; 		 //输入图像的列数
	int sum_horsum = 0;          //水平方向列相加和
	int long_pointer = 0;
	int short_pointer = 0;
	int METER_RANGE = 9;

	//按矩形计算像素值
	vector<int>num1;
	vector<int>num_cols;
	vector<int>num3;
	vector<int>num_pixel;
	for (int c = 0; c < cols; c++)
	{
		int versum = 0;
		for (int r = 0; r < rows; r++)
		{
			int index = int(image.at<uchar>(r, c));
			versum += index;
		} 

		if (versum != 0)
		{
			sum_horsum += versum;
			num1.push_back(c);             //列索引
			num3.push_back(sum_horsum);    //像素累加
		}
		if (versum == 0)
		{
			//列索引
			int maxValue1 = 0;
			for (auto v : num1)
			{
				if (maxValue1 < v) maxValue1 = v;
			}
			if (maxValue1 != 0)
			{
				num_cols.push_back(maxValue1);
			}
			vector<int>().swap(num1);

			//像素
			int maxValue2 = 0;
			for (auto v : num3)
			{
				if (maxValue2 < v) maxValue2 = v;
			}
			if (maxValue2 != 0)
			{
				num_pixel.push_back(maxValue2);
			}
			sum_horsum = 0;
			vector<int>().swap(num3);
		}
	}

	//标记长短指针
	auto firstValue1 = num_pixel.front();
	auto lastValue1 = num_pixel.back();
	//长短指针取值
	auto firstValue2 = num_cols.front() - 30;
	auto lastValue2 = num_cols.back() - 30;
	//赋值于长短指针
	if (firstValue1 < lastValue1)
	{
		short_pointer = firstValue2;
		long_pointer = lastValue2;
	}else 
	{
		short_pointer = lastValue2;
		long_pointer = firstValue2;
	}

	//计算表盘读数
	float rectangle_value = 1650;
	float short_result_ratio = (1.0 * short_pointer / rectangle_value);
	float long_result_ratio = (1.0 * long_pointer / rectangle_value);
	float short_result_value = (1.0 * short_result_ratio * METER_RANGE) - 4; 
	if (short_result_value < 0)
	{
		short_result_value += 9;
	}
	float long_result_value = (1.0 * long_result_ratio * METER_RANGE) - 4;
	if (long_result_value < 0)
	{
		long_result_value += 9;
	}
	//四舍五入取整
	if (short_result_value > 0)
	{
		short_result_value = short_result_value - int(short_result_value) >= 0.5 ? int(short_result_value) + 1 : int(short_result_value);
	}else
	{
		short_result_value = -short_result_value - int(-short_result_value) >= 0.5 ? -(int(-short_result_value) + 1) : -int(-short_result_value);
	}
	if (long_result_value > 0)
	{
		long_result_value = long_result_value - int(long_result_value) >= 0.5 ? int(long_result_value) + 1 : int(long_result_value);
	}else
	{
		long_result_value = -long_result_value - int(-long_result_value) >= 0.5 ? -(int(-long_result_value) + 1) : -int(-long_result_value);
	}
	//cout << "short_result_ratio:" << short_result_ratio << "  long_result_ratio:" << long_result_ratio << endl;
	cout << "short_result_value:" << short_result_value << "  long_result_value:" << long_result_value << endl;
	int result = (short_result_value * 10) + long_result_value;
	cout << "读数：" << result <<  endl;
	return result;
}