YOLO进行骑行非机动车检测与目标提取

#yolo算法提取非机动车并进行截图 import numpy as np import time import cv2 import os import matplotlib as plt import skimage labelsPath = r"E:\AI_projects\object_detect\YOLOV4(have_done)\coco.names" weightsPath = r"E:\AI_projects\object_detect\YOLOV4(have_done)\yolov3.weights" configPath = r"E:\AI_projects\object_detect\YOLOV4(have_done)\yolov3.cfg" #读取图片并 img_path=r'E:\AI_projects\object_detect\YOLOV4(have_done)\dataset\motor2.jpg' # image=skimage.io.imread(img_path) # image=plt.image.imread(img_path) image0 = cv2.imread(img_path) image0=cv2.resize(image0,(1200,800)) image=image0 #将可能含有4通道(250,250,4)带有透明度的图像转换成3通道(250,250,3),模型只接受3通道图像 image=cv2.cvtColor(image, cv2.COLOR_BGR2RGB) #锐化 kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]], np.float32) image = cv2.filter2D(image, -1, kernel=kernel) #读取coco80种物体标签 LABELS = None with open(labelsPath,'rt') as f: LABELS = f.read().rstrip('\n').split("\n") #初始化随机的颜色，这里主要是为后续检测到不同的对象，进行不同颜色的边框画图 np.random.seed(42) COLORS = np.random.randint(0, 255, size=(len(LABELS), 3),dtype="uint8") net = cv2.dnn.readNetFromDarknet(configPath, weightsPath) (H, W) = image.shape[:2] #计算图片的blob值，输入神经网络，进行前向反馈预测图片，只不过net.forward里面是ln， 神经网络的所有out层 ln = net.getLayerNames() ln = [ln[i[0] - 1] for i in net.getUnconnectedOutLayers()] blob = cv2.dnn.blobFromImage(image, 1 / 255.0, (416, 416),swapRB=True, crop=False) net.setInput(blob) start = time.time() layerOutputs = net.forward(ln) end = time.time() print("[INFO] YOLO took {:.6f} seconds".format(end - start)) #我们过滤到置信度小于0.5的对象，提取大于0.5置信度的对象， #分别计算每个检测对象的BOX、置信度以及label ID，并保存在如下先前建立的初始化参数list里面 boxes = [] confidences = [] classIDs = [] for output in layerOutputs: for detection in output: scores = detection[5:] classID = np.argmax(scores) confidence = scores[classID] if confidence > 0.5: box = detection[0:4] * np.array([W, H, W, H]) (centerX, centerY, width, height) = box.astype("int") x = int(centerX - (width / 2)) y = int(centerY - (height / 2)) boxes.append([x, y, int(width), int(height)]) confidences.append(float(confidence)) classIDs.append(classID) #对于每个对象，Yolo4 会框出 3 个左右的区域，但是我们只需要显示出最合适的区域。 #非最大值抑制算法NMS，就是搜索出局部最大值，将置信度最大的框保存，其余删除。 big_set=[] idxs = cv2.dnn.NMSBoxes(boxes, confidences, 0.5,0.4) if len(idxs) > 0: for i in idxs.flatten(): (x, y) = (boxes[i][0], boxes[i][1]) (w, h) = (boxes[i][2], boxes[i][3]) color = [int(c) for c in COLORS[classIDs[i]]] cv2.rectangle(image, (x, y), (x + w, y + h), color, 2) text = "{}: {:.4f}".format(LABELS[classIDs[i]], confidences[i]) cv2.putText(image, text, (x, y - 5), cv2.FONT_HERSHEY_SIMPLEX,0.5, color, 2) big_set.append([x,y,w,h,LABELS[classIDs[i]],confidences[i]]) cv2.imshow("Original Image", image0) cv2.imshow("Image", image) cv2.waitKey(0) person=[] motorbike=[] for i in range(len(big_set)): if big_set[i][4]=='motorbike': motorbike.append(big_set[i]) if big_set[i][4]=='person': person.append(big_set[i]) # print('person:',person) # print('motorbike:',motorbike) if len(person)>0 and len(motorbike)>0 : for i in range(len(person)): for j in range(len(motorbike)): Ax=person[i][0] Ay=person[i][1] Aw=person[i][2] Ah=person[i][3] A1x=motorbike[j][0] A1y=motorbike[j][1] A1w=motorbike[j][2] A1h=motorbike[j][3] if (abs(Ax+Aw/2-A1x-A1w/2)<10 and abs(Ay+Ah/2-A1y-A1h/2)<200): x=min(Ax,A1x) y=min(Ay,A1y) xw=max(Ax+Aw,A1x+A1w) yh=max(Ay+Ah,A1y+A1h) dst=image0[y-20:yh+20,x-20:xw+20] cv2.imshow('imge',dst) cv2.waitKey(0) cv2.destroyAllWindows()