基于python+opencv的茶叶嫩芽识别与采摘点定位方法

try: from PIL import Image except ImportError: import Image from PIL import Image import numpy as np import cv2 as cv from skimage import measure from sklearn.svm import SVC # 非线性 分类 SVM tea = cv.imread('img_4.png') pic = 'img_4.png' # 茶叶图片 img = Image.open(pic) # img.show() # 显示原始图像 img_arr = np.asarray(img,np.float64) # 选取茶叶嫩芽背景图像上的关键点RGB值(10个) bg_RGB = np.array( [[143,150,133],[103, 86, 76,],[90, 133, 118],[166, 186 ,152,],[29, 51, 32], [65, 56, 34],[74 ,111, 95],[119 ,102, 67],[75 ,80, 4],[149, 153, 134]] ) # 选取茶叶嫩芽上的关键点RGB值(10个) tea_RGB = np.array( [[110, 139, 43],[79, 130, 0],[146, 171, 36],[196, 218, 106],[107, 145, 31], [164, 197, 28],[76 ,108, 0],[132, 170, 10],[103, 129, 5],[104 ,130, 29]] ) RGB_arr = np.concatenate((bg_RGB,tea_RGB),axis=0) # 按列拼接 # 背景用0标记，茶叶嫩芽用1标记 label = np.append(np.zeros(bg_RGB.shape[0]),np.ones(tea_RGB.shape[0])) # 原本 img_arr 形状为(m,n,k),现在转化为(m*n,k) img_reshape = img_arr.reshape([img_arr.shape[0]*img_arr.shape[1],img_arr.shape[2]]) svc = SVC(kernel='poly',degree=3) # 使用多项式核，次数为3 svc.fit(RGB_arr,label) # SVM 训练样本 predict = svc.predict(img_reshape) # 预测测试点 bg_bool = predict == 0. # 为背景的序号(bool) bg_bool = bg_bool[:,np.newaxis] # 增加一列(一维变二维) bg_bool_3col = np.concatenate((bg_bool,bg_bool,bg_bool),axis=1) # 变为三列 bg_bool_3d = bg_bool_3col.reshape((img_arr.shape[0],img_arr.shape[1],img_arr.shape[2])) # 变回三维数组(逻辑数组) # img_arr[bg_bool_3d] = 255. # 将背景像素点变为白色 img_arr[bg_bool_3d] = 0. # 将背景像素点变为黑色 img_split = Image.fromarray(img_arr.astype('uint8')) # 数组转image # img_split.show() # 显示分割之后的图像 img_split.save('split_tea.jpg') # 保存 # 再依次进行二值化、中值滤波、腐蚀、空洞填充，求最大连通区域等形态学操作 # 最后利用二值化图像求出最大连通区域 # 再原嫩芽图像上画出茶叶嫩芽的最小外包矩形和中心点坐标； # 将其转化为二值图像(省略此步骤直接调用mask二值图像) tea_s = cv.imread('split_tea.jpg') bgr= cv.cvtColor(tea_s, cv.COLOR_HSV2BGR) # cv.imshow('1',bgr) gray=cv.cvtColor(bgr,cv.COLOR_BGR2GRAY) # cv.imshow('2',gray) ret, thresh = cv.threshold(gray, 0, 255, cv.THRESH_BINARY + cv.THRESH_OTSU) # cv.imshow("thresh", thresh) # 中值滤波：进行嫩芽降噪 median = cv.medianBlur(thresh,3) cv.imshow('median',median) # 开操作：去掉目标特征外的孤立点 kernel = cv.getStructuringElement(cv.MORPH_RECT, (3,3)) opening = cv.morphologyEx(median,cv.MORPH_OPEN,kernel) cv.imshow('open',opening) # 闭操作：去掉目标特征内的孔 kernel = np.ones((3,3),np.uint8) closing = cv.morphologyEx(opening,cv.MORPH_CLOSE,kernel) cv.imshow('clos',closing) cv.imwrite('SVM_clos.jpg', closing) # 求出最大连通区域 # image:二值图像 threshold_point:符合面积条件大小的阈值 def remove_small_points(image, threshold_point): img = cv.imread(image, 0) img_label, num = measure.label(img, connectivity=2, return_num=True) # 输出二值图像中所有的连通域 props = measure.regionprops(img_label) # 输出连通域的属性，包括面积等 resMatrix = np.zeros(img_label.shape) for i in range(1, len(props)): if props[i].area > threshold_point: tmp = (img_label == i + 1).astype(np.uint8) resMatrix += tmp # 组合所有符合条件的连通域 resMatrix *= 255 return resMatrix threshold_point=400 maxconnect = remove_small_points('SVM_clos.jpg', threshold_point) cv.imshow("maxconnect", maxconnect) # 再闭操作一次：去掉目标特征内的孔 kernel = np.ones((5,5),np.uint8) closing1 = cv.morphologyEx(maxconnect,cv.MORPH_CLOSE,kernel) cv.imshow('clos1',closing1) closing1=closing1.astype( np.uint8 ) # 在第一次分割嫩芽图像上框选出最大连通区域的嫩芽 # canny边缘检测 def canny_demo(image): t = 100 print(type(image[0][0])) canny_output = cv.Canny(image, t, t * 2) cv.imshow("canny_output", canny_output) # cv.imwrite("svm_canny_output.png", canny_output) return canny_output # 调用 binary = canny_demo(closing1) k = np.ones((3, 3), dtype=np.uint8) binary = cv.morphologyEx(binary, cv.MORPH_DILATE, k) cv.imshow('binary',binary) # 轮廓发现 contours, hierarchy = cv.findContours(binary, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) for c in range(len(contours)): rect = cv.minAreaRect(contours[c]) cx, cy = rect[0] box = cv.boxPoints(rect) box = np.int0(box) cv.drawContours(tea,[box],0,(0,255,0),2) cv.circle(tea, (np.int32(cx), np.int32(cy)), 2, (255, 0, 0), 2, 8, 0) cv.drawContours(tea, contours, c, (0, 0, 255), 1, 8) print(cx,cy) # 图像显示 cv.imshow("contours_analysis", tea) # cv.imwrite("svm_contours_analysis.png", pic) # # 最终在原图像上框选出每个嫩芽，并标注中心点，输出矩阵框左上角和右下角的坐标，描绘嫩芽轮廓 cv.waitKey(0) cv.destroyAllWindows()