OpenCV虾苗计数

项目部署：

实现原理：

          看代码应该能懂。

完整代码：

'''
/-----resources - 存储源文件
    ∟run
       ∟result - 存储处理结果

虾苗识别 - 自动计算面积阈值
'''
import matplotlib.pyplot as plt
import numpy as np
import cv2
from tqdm import tqdm
import os
import threading
import time


# 文件重命名
def rename(prefix):
    rootpath = r'resources'
    files = os.listdir(rootpath)
    filetype = '.jpg'

    index_name = 0
    for file in files:
        oldname = os.path.join(rootpath, file)

        num_bit = 1
        index = index_name
        while(index // 10 != 0):
            num_bit += 1
            index = index // 10

        newname = os.path.join(rootpath, str(index_name).zfill(prefix)) + filetype

        os.rename(oldname, newname)
        print(newname)

        index_name += 1


# rename(6)


class ShrimpIdent:
    def __init__(self):
        self.resPath = "resources"
        self.relPath = "run/result"
        self.filesPath = [(self.resPath + '/' + f) for f in os.listdir(self.resPath)]  # 待处理图像的相对路径
        self.relsPath = [(self.relPath + '/' + f) for f in os.listdir(self.resPath)]  # 处理结果保存的相对路径
        self.chartsPath = self.relPath + '/' + "charts.jpg"
        self.charts = {}
        # self.charts = {'0.jpg': 10356, '1.jpg': 8755, '2.jpg': 12821, '3.jpg': 21769, '4.jpg': 12821, '5.jpg': 12821,
        #                '6.jpg': 9475, '7.jpg': 9006, '8.jpg': 10171, '9.jpg': 10391}

    # 显示初始图像
    def showOriginalImg(self, img):
        cv2.imshow("win", img)

        # b, g, r = cv2.split(img)  # plt 输出格式
        # img2 = cv2.merge([r, g, b])
        # plt.imshow(img2)
        # plt.show()

    # 显示处理后的图像
    def showProcessedImg(self, img, count):
        cv2.imshow("win", img)

        # b, g, r = cv2.split(img)
        # img = cv2.merge([r, g, b])
        # plt.imshow(img)
        # plt.show()
        print("count:", count)

    # 计算图像阈值
    def computeAreasThreshold(self, areas):
        areas_set = list(set(areas))
        areas_set.sort()

        areas_num = []

        for i in range(len(areas_set)):
            if areas_set[i] == 0:
                continue
            areas_num.append(areas.count(areas_set[i]))

        areas_thre_index = 0
        areas_thre = 0

        for i in range(len(areas_num)):
            if areas_num[i] > max(areas_num) * 0.1:
                areas_thre = i

        return areas_thre

    # 识别目标并计数
    def computeNumberShrimp(self, pic_path):
        img = cv2.imread(pic_path)
        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        thre, edge = cv2.threshold(gray, 100, 255, cv2.THRESH_BINARY)
        edge = 255 - edge
        contours, hierarchy = cv2.findContours(edge, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
        cv2.drawContours(edge, contours, -1, (0, 0, 0xff), 5)

        areas = []
        for cont in contours:
            area = cv2.contourArea(cont)
            areas.append(area)

        areas_thre = self.computeAreasThreshold(areas)

        count = 0
        for cont in tqdm(contours):  # 进度条（tqdm）
            area = cv2.contourArea(cont)  # 计算包围性状的面积
            if area == 0.0 or area >= areas_thre:
                continue

            count = count + 1
            rect = cv2.boundingRect(cont)  # 提取矩形坐标
            cv2.rectangle(img, rect, (0, 0, 0xff), 1)  # 绘制矩形
            y = 10 if rect[1] < 10 else rect[1]  # 防止编号到图片之外
            cv2.putText(img, str(count), (rect[0], y), cv2.FONT_HERSHEY_COMPLEX, 0.4, (0, 255, 0), 1)  # 在左上角写上编号

        return img, count

    def run(self):
        stime = time.time()

        self.startThread(self.filesPath, self.savefor, self.paintcharts)

        etime = time.time()
        s = etime - stime
        m, s = divmod(s, 60)
        h, m = divmod(m, 60)
        print("耗时:  %d:%02d:%02d" % (h, m, s))
        print("END")

    # 多线程(number of runs, child thread func, main thread func)
    def startThread(self, nors, cthreadfunc, mthreadfunc):
        pthread_list = []  # 子线程列表
        for r in nors:
            pthread_list.append(
                threading.Thread(target=cthreadfunc, args=(r, self.relsPath[self.filesPath.index(r)],)))  # 子线程调用函数

        # 启动多线程
        for item in pthread_list:
            item.start()

        flag = [1] * len(pthread_list)
        while True:  # 所有子线程执行结束后，再执行主线程
            if sum(flag) == 0:
                mthreadfunc()  # 子线程结束后调用函数
                break
            else:
                for p in pthread_list:
                    flag[pthread_list.index(p)] = p.isAlive()
                time.sleep(2)

    # 保存结果图像
    def savefor(self, pic_path, rel_path):
        img, count = self.computeNumberShrimp(pic_path)
        self.charts.update({pic_path.split('/')[-1]: count})  # 图表参数
        cv2.putText(img, str(count), (100, 100), cv2.FONT_HERSHEY_COMPLEX, 4, color=(0, 0, 255))  # 向图片中写入识别个数
        cv2.imwrite(rel_path, img)

    # 绘制数量统计图表
    def paintcharts(self):
        print("********** 打印结果 **********")
        x_filename = np.array(list(self.charts.keys()))  # x轴
        y_count = np.array(list(self.charts.values()))  # y轴
        charts_size = len(x_filename)

        plt.bar(range(charts_size), y_count, color='lightsteelblue')
        plt.plot(range(charts_size), y_count, marker='o', color='coral')
        plt.xticks(range(charts_size), x_filename)
        plt.xlabel('Pic Name')
        plt.ylabel("Number of shrimps")
        # plt.legend()
        # plt.show()
        plt.savefig(self.chartsPath)


if __name__ == "__main__":
    shi = ShrimpIdent()
    shi.run()
    # shi.paintcharts()

待检测图片：

输出图片：

细节： 

统计图：