基于opencv实现象棋识别及棋谱定位python源码+数据集(含代码注释)人工智能课程设计.zip

import cv2 as cv import numpy as np import os from scipy import misc, ndimage import sys # 获取象棋的棋盘坐标 def get_cell(input_path): img = cv.imread(input_path) gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY) rectKernel = cv.getStructuringElement(cv.MORPH_RECT, (5, 5)) gradX = cv.morphologyEx(gray, cv.MORPH_BLACKHAT, rectKernel) ret1, th1 = cv.threshold(gradX, 35, 255, cv.THRESH_BINARY_INV) edges = cv.Canny(th1, 50, 150, apertureSize=3) lines = cv.HoughLines(edges, 1, np.pi/180, 143) zx = [] sx = [] for line in lines: rho, theta = line[0] a = np.cos(theta) b = np.sin(theta) x0 = a*rho y0 = b*rho x1 = int(x0 + 1000*(-b)) y1 = int(y0 + 1000*(a)) x2 = int(x0 - 1000*(-b)) y2 = int(y0 - 1000*(a)) if (x1 > 10 and x2 > 10 and abs(x1 - x2) < 10): # print("x", (x1, x2, y1, y2)) sx.append((x1, x2, y1, y2)) elif (y1 > 10 and y2 > 10 and abs(y1 - y2) < 10): # print("y", (x1, x2, y1, y2)) zx.append((x1, x2, y1, y2)) # cv.line(img,(x1,y1),(x2,y2),(0,0,255),2) sx = sorted(sx, key=lambda x: x[0]) zx = sorted(zx, key=lambda x: x[2]) sxFilter = [] if len(sx) > 0: sxFilter.append(sx[0]) tempsx = sx[0] for i in range(len(sx))[1:]: # print(sx[i][0],tempsx[0]) if (sx[i][0] - tempsx[0]) > 20: tempsx = sx[i] sxFilter.append(sx[i]) zxFilter = [] if len(zx) > 0: zxFilter.append(zx[0]) tempzx = zx[0] for i in range(len(zx))[1:]: # print(zx[i][0],tempzx[0]) if (zx[i][2] - tempzx[2]) > 20: tempzx = zx[i] zxFilter.append(zx[i]) return sxFilter, zxFilter # 获取 棋盘中棋子的位置信息 def get_chess_position(img_path): img = cv.imread(img_path) gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY) rectKernel = cv.getStructuringElement(cv.MORPH_RECT, (13, 13)) gradX = cv.morphologyEx(gray, cv.MORPH_BLACKHAT, rectKernel) questionCnts = [] ret1, th1 = cv.threshold(gradX, 170, 255, cv.THRESH_BINARY) grad2 = cv.morphologyEx(th1, cv.MORPH_DILATE, rectKernel) grad2 = cv.morphologyEx(grad2, cv.MORPH_DILATE, rectKernel) # 找到每一个圆圈轮廓 cnts = cv.findContours(grad2, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)[0] for i in range(len(cnts)): x, y, w, h = cv.boundingRect(cnts[i]) print(x, y, w, h) if (abs(w - h) > 20): continue # questionCnts.append(cnts[i]) questionCnts.append((x, y, w, h)) return questionCnts # 棋子的匹配加载 path = '/Users/apple/Desktop/python/recognition-chinese-chess/Dataset/train/' name_dict = {"b_c.png": "黑方车", "b_j.png": "黑方将", "b_m.png": "黑方马", "b_p.png": "黑方炮", "b_s.png": "黑方士", "b_x.png": "黑方象", "b_z.png": "黑方卒", "r_c.png": "红方车", "r_j.png": "红方帅", "r_m.png": "红方马", "r_p.png": "红方炮", "r_s.png": "红方士", "r_x.png": "红方象", "r_z.png": "红方兵"} if os.path.exists(path): files = os.listdir(path) print(files) sys.exit() def get_match(old_img): if len(files) > 0: # print(files) check_score = 0 check_img = "" for fileName in files: tempImg = cv.imread(path + fileName) ''' ①TM_SQDIFF是平方差匹配；TM_SQDIFF_NORMED是标准平方差匹配。利用平方差来进行匹配, 最好匹配为0.匹配越差, 匹配值越大。 ②TM_CCORR是相关性匹配；TM_CCORR_NORMED是标准相关性匹配。采用模板和图像间的乘法操作, 数越大表示匹配程度较高, 0 表示最坏的匹配效果。 ③TM_CCOEFF是相关性系数匹配；TM_CCOEFF_NORMED是标准相关性系数匹配。将模版对其均值的相对值与图像对其均值的相关值进行匹配, 1 表示完美匹配, -1 表示糟糕的匹配, 0 表示没有任何相关性(随机序列)。 总结：随着从简单的测量(平方差) 到更复杂的测量(相关系数), 我们可获得越来越准确的匹配(同时也意味着越来越大的计算代价)。 ''' result = cv.matchTemplate(cv.resize( old_img, None, fx=0.8, fy=0.8, interpolation=cv.INTER_CUBIC), tempImg, cv.TM_CCOEFF_NORMED) # print(result) (_, score, _, _) = cv.minMaxLoc(result) if check_score == 0 or score > check_score: check_score = score check_img = fileName # print(cv.minMaxLoc(result),name_dict[check_img]) return name_dict[check_img] else: print('this path not exist') return None input_path = './4.jpg' sxFilter, zxFilter = get_cell(input_path) questionCnts = get_chess_position(input_path) img = cv.imread(input_path) img_show = img.copy() gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY) chess_array = [[] for i in range(10)] if(len(sxFilter) == 9 and len(zxFilter) == 10): # 竖线 直线 print("---------------------") for i in range(len(zxFilter)): for j in range(len(sxFilter)): x1, x2, y1, y2 = zxFilter[i] x3, x4, y3, y4 = sxFilter[j] check = False print(zxFilter[i], "-", sxFilter[j], i, j) for k in range(len(questionCnts)): x, y, w, h = questionCnts[k] # 获取中心坐标 centre_x = x+w/2 centre_y = y+h/2 # 范围之内匹配成功 print(centre_x, "-", centre_y) if(abs(centre_x-x3) < 25 and abs(centre_y-y1) < 25): print("命中!") check = True im = img[y:y + h, x:x + w] orc = get_match(im) if orc is not None: chess_array[i].append(orc) else: print("未识别...") if not check: chess_array[i].append("空空空") else: print("识别出错了!") for i in range(10): for j in range(9): print(chess_array[i][j], end="\t") print() for i in range(len(sxFilter)): line = sxFilter[i] cv.line(img, (line[0], line[2]), (line[1], line[3]), (0, 0, 255), 2) # print(line) cv.putText(img, str(i), (line[0], 500), cv.FONT_HERSHEY_SIMPLEX, 0.75, (255, 0, 0), 2) for i in range(len(zxFilter)): line = zxFilter[i] cv.line(img, (line[0], line[2]), (line[1], line[3]), (0, 0, 255), 2) # print(line) cv.putText(img, str(i), (300, line[2]), cv.FONT_HERSHEY_SIMPLEX, 0.75, (255, 0, 0), 2) res = np.hstack((img_show, img)) showIMG = cv.resize(res, None, fx=0.7, fy=0.6, interpolation=cv.INTER_CUBIC) cv.imshow("TEST", showIMG) k = cv.waitKey(0) if k == 27: cv.destroyAllWindows() cv.destroyAllWindows()