【Toybrick RK3399Pro AI开发板试用体验】opencv 识别可视范围的人类

2019-9-2 18:05:44 2298

0 ` 本帖最后由 moonyuan 于 2019-9-3 00:48 编辑在很多场合，例如节能建筑智能会议室需要人走关灯在传统的PIR红外识别，误触发/不触发概率高，使用摄像头识别比传统方式可靠那么接下来实现利用opencv 识别当前可视范围的是否存在人类首先加载caffe网络模型 net = dnn::readNetFromCaffe("./MobileNetSSD_deploy.prototxt.txt", "./MobileNetSSD_deploy.caffemodel"); 复制代码读取图片进行灰度转换，因为模型训练时使用300300，所以图片需要重新resize image = imread("messi.jpg"); Mat gray; cvtColor(image,gray,CV_BGR2GRAY); resize(gray, gray, Size(300, 300)); 复制代码* 将blob输入到网络，并向前推进， Mat blob; blob = blobFromImage(result,0.007843, Size(300, 300), 127.5); net.setinput(blob); Mat detections; detections = net.forward(); Mat detectionMat(detections.size[2], detections.size[3], CV_32F, detections.ptr()); for (int i = 0; i < detectionMat.rows; i++){ float confidence = detectionMat.at(i, 2); if (confidence > 0.1){ size_t objectClass = (size_t)(detectionMat.at(i, 1)); printf("n current image classification : %s, possible : %.2f ", CLASSES[objectClass].c_str(), confidence); if((confidence <= 0.1) \|\| strcmp("person",CLASSES[objectClass].c_str())){ continue; } int xLeftBottom = static_cast(detectionMat.at(i, 3) * gray.cols); int yLeftBottom = static_cast(detectionMat.at(i, 4) * gray.rows); int xRightTop = static_cast(detectionMat.at(i, 5) * gray.cols); int yRightTop = static_cast(detectionMat.at(i, 6) * gray.rows); ostringstream ss; ss << confidence; String conf(ss.str()); Rect object((int)xLeftBottom, (int)yLeftBottom, (int)(xRightTop - xLeftBottom), (int)(yRightTop - yLeftBottom)); rectangle(result, object, Scalar(0, 255, 0)); String label = String(CLASSES[objectClass]) + ": " + conf; int baseLine = 0; Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine); rectangle(result, Rect(Point(xLeftBottom, yLeftBottom - labelSize.height), Size(labelSize.width, labelSize.height + baseLine)), Scalar(255, 255, 255), CV_FILLED); putText(result, label, Point(xLeftBottom, yLeftBottom), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 0, 0)); } } 复制代码最后把图像输出 imshow("结果", result); 复制代码我们看到图像的认为可以没独立识别出来再识别一下其他的图输出结果结果都是能够识别到 94%以上的相似率，后续经过修改定可以放到指定的场景稳定准确的工作 ` 0
2019-9-2 18:05:44　　评论淘帖0 举报相关推荐 • 【Toybrick RK3399Pro AI开发板试用体验】opencv 人脸识别 3295 • 【Toybrick RK3399Pro AI开发板试用体验】Toybrick RK3399Pro安装ROS系统 5320 • 【Toybrick RK3399Pro AI开发板试用体验】编译安装OpenCV 3.4.7 3496 • 【Toybrick RK3399Pro AI开发板试用体验】opencv试运行 2287 • 【Toybrick RK3399Pro AI开发板试用体验】3399PRO 开箱 4433 • 【Toybrick RK3399Pro AI开发板试用体验】安装debian rootfs到rk3399pro 3927 • 【Toybrick RK3399Pro AI开发板试用体验】ACT Ⅵ：Linux下静态人脸识别 python+opencv 2495 • 【Toybrick RK3399Pro AI开发板试用体验】ACT VIII：试用结贴 1793 • 【Toybrick RK3399Pro AI开发板试用体验】开发环境搭建 2360 • 【Toybrick RK3399Pro AI开发板试用申请】AI图形识别+语音识别 2705 1 个讨论